IAEA HUMAN HEALTH SERIES No.

36
The complexity of a combined SPECT/CT system requires
rigorous quality control procedures. However, the information
provided by clinical examinations needs to be complemented
with an understanding of potential problems arising from the
combined imaging procedure. This publication presents an
overview of quality control procedures in SPECT and SPECT/CT
and describes the pitfalls and image artefacts that can occur.

IAEA HUMAN HEALTH SERIES IAEA HUMAN HEALTH SERIES

No. 36

SPECT/CT Atlas of Quality Control and Image Artefacts

SPECT/CT Atlas of
Quality Control and
Image Artefacts

INTERNATIONAL ATOMIC ENERGY AGENCY

VIENNA
ISBN 978–92–0–103919–4
ISSN 2075–3772 1
 RELAT ED P U B L I C ATI ON S

IAEA HUMAN HEALTH SERIES PUBLICATIONS

QUALITY ASSURANCE FOR PET AND PET/CT SYSTEMS
The mandate of the IAEA human health programme originates from Article II of
IAEA Human Health Series No. 1
its Statute, which states that the “Agency shall seek to accelerate and enlarge the STI/PUB/1393 (145 pp.; 2009)
contribution of atomic energy to peace, health and prosperity throughout the world”. ISBN 978–92–0–103609–4 Price: €32.00
The main objective of the human health programme is to enhance the capabilities of
IAEA Member States in addressing issues related to the prevention, diagnosis and
QUALITY ASSURANCE FOR SPECT SYSTEMS
treatment of health problems through the development and application of nuclear
IAEA Human Health Series No. 6
techniques, within a framework of quality assurance. STI/PUB/1394 (249 pp.; 2009)
Publications in the IAEA Human Health Series provide information in the areas ISBN 978–92–0–103709–1 Price: €52.00
of: radiation medicine, including diagnostic radiology, diagnostic and therapeutic nuclear
medicine, and radiation therapy; dosimetry and medical radiation physics; and stable
QUALITY ASSURANCE PROGRAMME FOR COMPUTED TOMOGRAPHY:
isotope techniques and other nuclear applications in nutrition. The publications have a
DIAGNOSTIC AND THERAPY APPLICATIONS
broad readership and are aimed at medical practitioners, researchers and other
IAEA Human Health Series No. 19
professionals. International experts assist the IAEA Secretariat in drafting and reviewing STI/PUB/1557 (171 pp.; 2012)
these publications. Some of the publications in this series may also be endorsed or co- ISBN 978–92–0–128910–0 Price: €48.00
sponsored by international organizations and professional societies active in the relevant
fields.
PET/CT ATLAS ON QUALITY CONTROL AND IMAGE ARTEFACTS
There are two categories of publications in this series: IAEA Human Health Series No. 27
STI/PUB/1642 (101 pp.; 2014)
IAEA HUMAN HEALTH SERIES ISBN 978–92–0–101014–8 Price: €48.00
Publications in this category present analyses or provide information of an
advisory nature, for example guidelines, codes and standards of practice, and quality
assurance manuals. Monographs and high level educational material, such as graduate
texts, are also published in this series.

IAEA HUMAN HEALTH REPORTS

Human Health Reports complement information published in the IAEA Human
Health Series in areas of radiation medicine, dosimetry and medical radiation physics,
and nutrition. These publications include reports of technical meetings, the results of
IAEA coordinated research projects, interim reports on IAEA projects, and educational
material compiled for IAEA training courses dealing with human health related subjects.
In some cases, these reports may provide supporting material relating to publications
issued in the IAEA Human Health Series.

All of these publications can be downloaded cost free from the IAEA web site:
http://www.iaea.org/Publications/index.html
Further information is available from:
Marketing and Sales Unit
International Atomic Energy Agency
Vienna International Centre
PO Box 100
1400 Vienna, Austria

Readers are invited to provide their impressions on these publications.

Information may be provided via the IAEA web site, by mail at the address given above,
or by email to:
[email protected].

www.iaea.org/books
 SPECT/CT ATLAS OF
QUALITY CONTROL
AND IMAGE ARTEFACTS
 The following States are Members of the International Atomic Energy Agency:

AFGHANISTAN GERMANY PAKISTAN

ALBANIA GHANA PALAU
ALGERIA GREECE PANAMA
ANGOLA GRENADA PAPUA NEW GUINEA
ANTIGUA AND BARBUDA GUATEMALA PARAGUAY
ARGENTINA GUYANA PERU
ARMENIA HAITI PHILIPPINES
AUSTRALIA HOLY SEE POLAND
AUSTRIA HONDURAS PORTUGAL
AZERBAIJAN HUNGARY
QATAR
BAHAMAS ICELAND
REPUBLIC OF MOLDOVA
BAHRAIN INDIA
BANGLADESH INDONESIA ROMANIA
BARBADOS IRAN, ISLAMIC REPUBLIC OF RUSSIAN FEDERATION
BELARUS IRAQ RWANDA
BELGIUM IRELAND SAINT LUCIA
BELIZE ISRAEL SAINT VINCENT AND
BENIN ITALY THE GRENADINES
BOLIVIA, PLURINATIONAL JAMAICA SAN MARINO
STATE OF JAPAN SAUDI ARABIA
BOSNIA AND HERZEGOVINA JORDAN SENEGAL
BOTSWANA KAZAKHSTAN SERBIA
BRAZIL KENYA SEYCHELLES
BRUNEI DARUSSALAM KOREA, REPUBLIC OF SIERRA LEONE
BULGARIA KUWAIT SINGAPORE
BURKINA FASO KYRGYZSTAN SLOVAKIA
BURUNDI LAO PEOPLE’S DEMOCRATIC SLOVENIA
CAMBODIA REPUBLIC SOUTH AFRICA
CAMEROON LATVIA SPAIN
CANADA LEBANON
SRI LANKA
CENTRAL AFRICAN LESOTHO
SUDAN
REPUBLIC LIBERIA
CHAD LIBYA SWEDEN
CHILE LIECHTENSTEIN SWITZERLAND
CHINA LITHUANIA SYRIAN ARAB REPUBLIC
COLOMBIA LUXEMBOURG TAJIKISTAN
CONGO MADAGASCAR THAILAND
COSTA RICA MALAWI TOGO
CÔTE D’IVOIRE MALAYSIA TRINIDAD AND TOBAGO
CROATIA MALI TUNISIA
CUBA MALTA TURKEY
CYPRUS MARSHALL ISLANDS TURKMENISTAN
CZECH REPUBLIC MAURITANIA UGANDA
DEMOCRATIC REPUBLIC MAURITIUS UKRAINE
OF THE CONGO MEXICO UNITED ARAB EMIRATES
DENMARK MONACO UNITED KINGDOM OF
DJIBOUTI MONGOLIA GREAT BRITAIN AND
DOMINICA MONTENEGRO NORTHERN IRELAND
DOMINICAN REPUBLIC MOROCCO UNITED REPUBLIC
ECUADOR MOZAMBIQUE
OF TANZANIA
EGYPT MYANMAR
UNITED STATES OF AMERICA
EL SALVADOR NAMIBIA
ERITREA NEPAL URUGUAY
ESTONIA NETHERLANDS UZBEKISTAN
ESWATINI NEW ZEALAND VANUATU
ETHIOPIA NICARAGUA VENEZUELA, BOLIVARIAN
FIJI NIGER REPUBLIC OF
FINLAND NIGERIA VIET NAM
FRANCE NORTH MACEDONIA YEMEN
GABON NORWAY ZAMBIA
GEORGIA OMAN ZIMBABWE

The Agency’s Statute was approved on 23 October 1956 by the Conference on the Statute of the
IAEA held at United Nations Headquarters, New York; it entered into force on 29 July 1957.
The Headquarters of the Agency are situated in Vienna. Its principal objective is “to accelerate and enlarge
the contribution of atomic energy to peace, health and prosperity throughout the world’’.
 IAEA HUMAN HEALTH SERIES No. 36

SPECT/CT ATLAS OF
QUALITY CONTROL
AND IMAGE ARTEFACTS

INTERNATIONAL ATOMIC ENERGY AGENCY

VIENNA, 2019
 COPYRIGHT NOTICE

All IAEA scientific and technical publications are protected by the terms of
the Universal Copyright Convention as adopted in 1952 (Berne) and as revised
in 1972 (Paris). The copyright has since been extended by the World Intellectual
Property Organization (Geneva) to include electronic and virtual intellectual
property. Permission to use whole or parts of texts contained in IAEA publications
in printed or electronic form must be obtained and is usually subject to royalty
agreements. Proposals for non-commercial reproductions and translations are
welcomed and considered on a case-by-case basis. Enquiries should be addressed
to the IAEA Publishing Section at:

Marketing and Sales Unit, Publishing Section

International Atomic Energy Agency
Vienna International Centre
PO Box 100
1400 Vienna, Austria
fax: +43 1 26007 22529
tel.: +43 1 2600 22417
email: [email protected]
www.iaea.org/publications
© IAEA, 2019

Printed by the IAEA in Austria

December 2019
STI/PUB/1860

IAEA Library Cataloguing in Publication Data

Names: International Atomic Energy Agency.
Title: SPECT/CT atlas of quality control and image artefacts / International Atomic
Energy Agency.
Description: Vienna : International Atomic Energy Agency, 2019. | Series: IAEA
human health series, ISSN 2075–3772 ; no. 36 | Includes bibliographical
references.
Identifiers: IAEAL 19-01276 | ISBN 978–92–0–103919–4 (paperback : alk. paper) |
ISBN 978–92–0–159219–4 (pdf)
Subjects: LCSH: Single-photon emission computed tomography. | Tomography —
Atlases. | Diagnostic imaging. | Quality control.
Classification: UDC 616–073(084) | STI/PUB/1860
 FOREWORD

In recent years, hybrid single photon emission computed tomography (SPECT)/computed tomography (CT)
imaging has become increasingly prevalent in nuclear medicine and diagnostic radiology. To accurately interpret
SPECT/CT images, it is important to understand the principles of image formation and the biological distribution
of the radiopharmaceutical, and also to understand the types of image artefact that can arise from these imaging
systems. This publication presents an overview of quality control procedures in SPECT and SPECT/CT and
describes the pitfalls and artefacts that can occur in these imaging modalities.
A series of case studies of image artefacts are presented that can occur, with descriptions of their causes and
the steps that can be taken to avoid their recurrence. Of the 67 case studies in this publication, some are actual
clinical cases, while others use phantoms in order to yield valuable data. The atlas is intended to be used as a guide
on how to take appropriate quality control measures and to assist with problem analyses and prevention. It is hoped
that this publication will be especially useful for medical physicists, physicians, nuclear medicine and diagnostic
radiology professionals, and service engineers.
The IAEA wishes to express its appreciation to all those who contributed to the drafting and review of
this publication, particularly J.C. Dickson (United Kingdom), S. Holm (Denmark), O. Mawlawi (United States
of America) and C.C. Robilotta (Brazil). The IAEA officer for this publication was G.L. Poli of the Division of
Human Health.
 EDITORIAL NOTE
Although great care has been taken to maintain the accuracy of information contained in this publication, neither the IAEA nor
its Member States assume any responsibility for consequences which may arise from its use.
This publication does not address questions of responsibility, legal or otherwise, for acts or omissions on the part of any person.
Guidance provided here, describing good practices, represents expert opinion but does not constitute recommendations made on
the basis of a consensus of Member States.
The use of particular designations of countries or territories does not imply any judgement by the publisher, the IAEA, as to the
legal status of such countries or territories, of their authorities and institutions or of the delimitation of their boundaries.
The mention of names of specific companies or products (whether or not indicated as registered) does not imply any intention to
infringe proprietary rights, nor should it be construed as an endorsement or recommendation on the part of the IAEA.
The IAEA has no responsibility for the persistence or accuracy of URLs for external or third party Internet web sites referred to
in this book and does not guarantee that any content on such web sites is, or will remain, accurate or appropriate.
 CONTENTS

1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1. Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2. Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3. Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.4. Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2. SPECT AND SPECT/CT SYSTEMS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2.1. Emission tomography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2.1.1. Basic design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.1.2. Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.3. Image reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.4. Corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2. Computed tomography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.1. Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.2. Image reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3. CT based attenuation correction for SPECT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3.1. Attenuation correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3.2. Models for attenuation correction calculation from CT data . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4. Benefits of SPECT/CT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4.1. Functional mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4.2. Quantitative SPECT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.5. Commercially available SPECT/CT systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.5.1. General Electric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.5.2. Mediso . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.5.3. Philips . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.5.4. Siemens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3. QUALITY CONTROL OF SPECT AND SPECT/CT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

3.1. Routine periodic planar, SPECT and SPECT/CT quality control tests . . . . . . . . . . . . . . . . . . . . . . 21
3.2. SPECT and SPECT/CT testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2.1. Planar uniformity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2.2. Peaking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2.3. Pixel size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.4. Centre of rotation and multiple head registration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.5. SPECT uniformity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.6. SPECT resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2.7. SPECT image contrast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2.8. Slice thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2.9. System volume sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.2.10. Sensitivity and uniformity with rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3. CT testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3.1. Tube warm up and air calibrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3.2. Uniformity, noise, linearity, resolution and artefacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3.3. CT dose assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.4. SPECT/CT tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.4.1. SPECT/CT registration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.4.2. SPECT/CT image quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4. ARTEFACTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4.1. System related artefacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

4.1.1. Non-uniformity of detector response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.1.2. Crystal hydration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.1.3. Poor correction maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.1.4. Centre of rotation offset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.1.5. Energy peaking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.1.6. Head tilt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.1.7. Mismatched sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.1.8. Phantom related artefacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.2. Acquisition related artefacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.2.1. Matrix and pixel size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.2.2. Number of projections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.2.3. Counts per projection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.2.4. Acquisition arc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.2.5. Collimator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.2.6. Radius of rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2.7. Orbit effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.3. Reconstruction artefacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.3.1. Filtered back projection streak artefacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.3.2. Filters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.3.3. Iterations and subsets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.3.4. Use of resolution modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.3.5. Chang attenuation correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.4. Patient related artefacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.4.1. Voluntary patient motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.4.2. Involuntary patient motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.4.3. Postural and positional artefacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.4.4. Incomplete injection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.4.5. Radiopharmaceutical cross-contamination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.4.6. Patient attenuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.5. CT issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.5.1. CT slice thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.5.2. CT helical pitch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.5.3. Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.5.4. Slow rotation time CT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.6. SPECT/CT artefacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.6.1. Beam hardening artefacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.6.2. Contrast CT: Negative contrast . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
4.6.3. Truncation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
4.6.4. Misregistration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
4.7. Quantitative issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.7.1. Iterations and subsets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.7.2. Corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.7.3. Iodine contrast in 99mTc- and 111In-SPECT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4.7.4. Local effects of iodine contrast in 99mTc-SPECT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
4.7.5. ECG gating: Arrhythmia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121

REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
ABBREVIATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
CONTRIBUTORS TO DRAFTING AND REVIEW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
 1.  INTRODUCTION

1.1. BACKGROUND

Single photon emission computed tomography (SPECT) is a modality whose roots were established in the
early 1960s following the work of Kuhl and Edwards [1]. However, it was not until the advent of systems using
the Anger gamma camera in the 1970s that SPECT was established as a technology that would be recognizable
today [2–4]. Little changed from these original gamma camera tomographic systems until the work of Hasegawa [5]
in the 1990s, which combined a SPECT and a computed tomography (CT) system in what is known as a hybrid
SPECT/CT scanner. This was nearly ten years before the inception of the first positron emission tomography/
computed tomography (PET/CT) scanner. The adoption of SPECT/CT systems into routine clinical use initially
proved to be slow. However, all major manufacturers have now released SPECT/CT systems with various
specifications and capabilities.

1.2. OBJECTIVE

The complexity of a combined SPECT/CT system requires rigorous quality control procedures. Similarly, the
wealth of information provided by combined SPECT/CT examinations needs to be complemented with a thorough
understanding of potential pitfalls and artefacts arising from the combined imaging procedure. SPECT/CT, as with
any other imaging modality, is acceptable for routine clinical and research applications only if technical pitfalls can
be avoided prospectively; if artefacts from incorrect or suboptimal acquisition procedures can be recognized and,
whenever possible, corrected retrospectively; and if the resulting image information can be interpreted correctly,
which entails an appreciation of variants of the represented image information.

1.3. SCOPE

This publication presents an overview of quality control procedures in SPECT and SPECT/CT and describes
the pitfalls and image artefacts that can occur. Clearly, it is impossible for any book to cover all potentially
encountered image artefacts at every site that uses a SPECT or SPECT/CT system. However, an understanding
of the source of these artefacts will help to support improved SPECT/CT operations along with the high quality
standards required for an increased acceptance of this modality. Guidance provided here, describing good practices,
represents expert opinion but does not constitute recommendations made on the basis of a consensus of Member
States.

1.4. STRUCTURE

This publication builds on the IAEA Quality Control Atlas for Scintillation Camera Systems [6], published
in 2003, and compliments the PET/CT Atlas on Quality Control and Image Artefacts [7], published in 2014.
Section 2 provides a brief overview of the basic principles of emission tomography, CT and the use of CT images
for attenuation correction (AC) of SPECT images, and concludes with a survey of SPECT and SPECT/CT systems
currently available. Section 3 describes the required quality control procedures for the operation of SPECT and
SPECT/CT systems in routine clinical practice, and additional literature is provided in the references. Section 4
presents the case studies of artefacts from different sources, ranging from hardware malfunctions to patient induced
artefacts. Each case study includes some background, a description and some guidance on how to interpret and
understand the nature of the artefacts and ways to avoid them.

1
 2.  SPECT AND SPECT/CT SYSTEMS

2.1. EMISSION TOMOGRAPHY

2.1.1. Basic design

SPECT is a modality whose roots were established in the early 1960s following the work of Kuhl and
Edwards [1]. However, it was not until the advent of systems using the Anger gamma camera [2–4] that SPECT
was recognized as a novel imaging technology (see Fig. 1(a)). Typically, SPECT systems use a gamma camera
detector capable of rotating around the object of interest, and projection images are acquired at multiple angles
to provide the information necessary to reconstruct a three dimensional distribution of activity. Until recently,
little had changed from these original gamma camera tomographic systems (see Fig. 1(b)), although configurations
of two, three and even four detector heads were developed to speed up the process of SPECT data acquisition
(see Fig. 1(c)). Based on some of these designs, SPECT systems specific to different areas of nuclear medicine
have been developed — namely, nuclear cardiology and the brain (see Fig. 1(d)).
More recently, equipment manufacturers have focused on novel SPECT systems that differ from the standard
gamma camera SPECT design. There has been a move towards hybrid SPECT/CT systems, and replacing the
scintillation detector based gamma camera systems with detectors based on semiconductor technology. The
introduction of SPECT/CT systems has been driven, at least initially, for the attenuation correction of myocardial
perfusion studies and basic anatomical mapping, with more recent developments focusing on better anatomical
mapping and quantitative evaluation of SPECT studies. Furthermore, imaging devices are now available that

FIG. 1. (a) An original gamma camera design (courtesy of R. Jaszczak and reproduced with permission courtesy of IOP Publishing)
and (b) a modern equivalent (courtesy of GE Healthcare). (c) Multidetector designs (courtesy of Siemens Healthineers) have led to
(d) application specific devices (courtesy of Mediso Medical Imaging Systems).

2
either replace the photomultiplier tube of a scintillation detector with pin or silicon drift diodes, or that replace the
complete detector with designs based on semiconductor materials such as cadmium zinc telluride (CZT). The use
of semiconductor designs has several advantages. Compared to a typical scintillation detector, solid state systems
offer much more compact imaging devices, which can lead to improved patient comfort and a reduced scanner
footprint. There are also additional benefits in performance that come with these systems  [8,  9]. For example,
improvements in energy resolution offer clear benefits in dual radionuclide scanning, while high count rate
performance provides advantages in first-pass and dynamic imaging. Other improvements, such as discontinuing
the use of photomultiplier tubes, also offer the possibility of combining SPECT with magnetic resonance (MR) in
hybrid SPECT/MR scanners (see Fig. 2 for examples of current solid state based SPECT imaging systems).

2.1.2. Acquisition

A number of parameters are required to define the SPECT acquisition. As with standard gamma camera
imaging, the number of acquired counts determines the noise properties of the image. Since there is a limit to
the activity that can be administered to a patient, count and noise levels are controlled by imaging time. SPECT
requires images acquired at multiple angles, so the noise level at each of these angles needs to be considered. If the
time per projection is overextended, the acquisition could become prohibitively long, and the radiopharmaceutical
distribution might change. Conversely, very short projection times can result in low counts and high image noise
level, which might destabilize certain reconstructions [10]. Another factor that determines the level of image noise
is the pixel size used. Typically, SPECT acquisitions use either 64 × 64 or 128 × 128 matrices with some zoom
factor, resulting in pixel sizes between approximately 2 mm and 8 mm. With spatial resolution of modern systems

FIG. 2. SPECT systems with semiconductor technology: (a) D-SPECT cardio dedicated cardiac system (courtesy of Spectrum
Dynamics Medical SA); (b) Discovery 670 CZT all purpose SPECT/CT system (courtesy of GE Healthcare); and (c) Veriton wholebody
system (courtesy of Spectrum Dynamics Medical SA).

3
in the range of 8–10 mm, there is no benefit to acquire with smaller pixel sizes, while larger pixel sizes are avoided
because of sampling and partial volume issues.
To obtain SPECT data, images from multiple angles are acquired by rotating the detector(s) around the
object of interest. Traditionally, SPECT requires images obtained from 360 degrees of arc, which for multiple head
systems can be split between each detector, for example 180 degrees of arc per detector for dual headed systems
(see Fig. 3(a)). However, a split acquisition approach requires well matched detector sensitivities to avoid bias
artefacts. In nuclear cardiology, because of the left anterior sided position of the heart, images using 180 degrees
of arc are commonly acquired between right anterior oblique (RAO) to left posterior oblique (LPO) positions.
This provides optimal resolution and limits attenuation losses over the heart area compared to a full 360 degree
arc acquisition, but it does lead to distortion in the right posterior part of the reconstructed data due to the distance
dependent resolution of the collimator not being balanced out by an opposed view (see Fig. 3(c)). Fortunately,
this area is not diagnostically helpful in typical cardiac studies. To make this type of acquisition more efficient
on dual detector systems, the two detectors are positioned into a 90 degree or L-mode configuration to form a
right angle (see Fig. 3(b)). One additional rotational option available in many SPECT systems is the choice of
‘step-and-shoot’, or continuous acquisition mode. In the step-and-shoot mode, the detectors are stationary during
the acquisition of projection data, with a short quick movement used to move between different projection angles.
In continuous mode, data are being acquired while the detector moves at a constant speed, with the data binned into
angular projections as the detectors rotate. Continuous mode therefore slightly reduces the acquisition time, at the
cost of some blurring of the projection data.
As with planar imaging, collimation plays a key role in the performance and acquisition of SPECT systems.
Spatial resolution in SPECT is often inferior to that of planar imaging. This is due to the increased distance between

FIG. 3. (a) The standard position for SPECT on a dual detector system is for each detector opposed to each other with each detector
travelling through 180 degrees to acquire over 360 degrees of arc. (b) In nuclear cardiology, a right angle (L-mode) approach is
commonly used to acquire data over 180 degrees of arc. (c) Acquiring over 180 degrees of arc can lead to distortion in areas distant
to the acquired projections.

4
the individual detectors and the source in SPECT, and the additional resolution losses from the tomographic
reconstruction process. To limit this problem, non-circular orbits are frequently used, with the patient–collimator
distance minimized for each projection angle. This works well, although it is frequently avoided in head imaging,
where collisions between the patient and the detector could cause significant harm. In this application, circular
orbits are used instead, with the patient couch elevated so that the patient–collimator distance is minimized and
relatively consistent around the head. Furthermore, higher resolution collimators are preferred where available,
with increased injected activities used to compensate for the loss in system sensitivity. There has also been a
move to more novel collimator designs created for specific applications. In brain imaging, for example, fan beam
collimators have been used to improve the spatial resolution of SPECT imaging without reducing the sensitivity.
More recently, developments in slit-slat and multiple pinhole collimators [11–13] have led to the introduction of
clinical and pre-clinical systems for SPECT.
Similarly in the field of nuclear cardiology, the use of novel collimator designs has also been explored. Fan
beam collimators have again been popular, with designs adopting an off-centre plane [14] or using variable focal
planes [15]. More recently, the use of ultra high sensitivity collimators has been introduced in an effort to drastically
improve patient throughput  [16,  17]. Typically, such systems have shown that diagnostic quality images can be
produced at a quarter of the time or administered activity compared to standard SPECT systems [18]. The typical
trade-off of using high sensitivity collimators is a corresponding loss of spatial resolution. However, developments
in tomographic reconstruction algorithms have led to new approaches that model these losses, allowing for the
recovery of spatial resolution (see Section 2.1.4). This combination of high sensitivity collimators and resolution
model reconstruction algorithms allows for the use of imaging protocols with shortened acquisition times (or
reduced administered activities) while still maintaining spatial resolution.
A consequence of the use of shortened acquisition times is that it offers greater potential for the implementation
of dynamic SPECT. Acquiring dynamic SPECT data allows the user to look at the changes in the distribution of the
radiopharmaceutical over time. Until recently, with standard SPECT acquisitions typically taking 20–30 minutes,
the available temporal resolution was only useful for applications looking at relatively slow tracer kinetics or
displacement studies  [19,  20]. However, with the introduction of high sensitivity SPECT systems the potential
to achieve the temporal resolution required to track first-pass kinetics and to perform kinetic modelling can be
realized [21].

2.1.3. Image reconstruction

To form a three dimensional image, the individually acquired images obtained from different angles undergo
a mathematical reconstruction process. The two different reconstruction processes typically available in clinical
systems are filtered back projection (FBP) and iterative reconstruction [22, 23].
The process of FBP is shown in Fig. 4. Taking an object with two point sources of activity, Fig. 4 shows
that the image or projection acquired at each angle will be different. In an anterior view, the two point sources are

FIG. 4. (a) The acquisition of projection data from a source distribution and (b) its subsequent back projection to form the original
image.

5
represented as two individual objects, while in an oblique view, two objects will be seen, albeit with a different
distance between the sources. In the lateral view, however, the image will show only one object, with a signal
equal to the sum of the two source activities. Prior to reconstruction, only projection data are available, and the
distribution of source activity is unknown. By back projecting the projection data onto a blank image matrix and
summing contributions from each back projection, the image of two point sources starts to become resolved.
In Fig. 4, only one transaxial slice is represented to demonstrate the reconstruction process. In clinical SPECT,
however, because a two dimensional planar image is acquired at each projection angle, it is possible to reconstruct
a stack of consecutive multiple slides to form a three dimensional object.
One of the consequences of the back projection process is that the resultant image is a blurred version of the
original object. To address this, a ramp filter is applied before the back projection process. For ideal mathematical
projection data, FBP yields an exact solution. However, because SPECT is noisy, a modification of the ramp filter
is required, usually by applying a low pass filter such as Butterworth or Hann filters. The type and form of low pass
filter depend on the spatial resolution requirements and noise characteristics of the data.
Iterative reconstruction takes a different approach to the reconstruction process (see Fig. 5). In this method, an
algorithm predicts what the image might be, for example suggesting that 100 counts are in each voxel. An estimate
is then made of the projections that would result from this predicted image. In its simplest form, this would be a
basic forward projection step of image into projection images, but it can also consider factors such as attenuation
of photons or resolution losses (see Section 2.1.4). A comparison is then made between the acquired projection
images and the estimated forward projections. If the two image sets are identical or similar enough to within a
pre-set acceptable difference, the reconstruction algorithm can stop, and the estimated image can be assumed to
be a good representation of the true image. If there are significant differences between estimated and acquired
projection data, then an updated estimate is generated based on the ratio of the acquired and estimated data. This
process can then go through multiple iterations until a good representation of the object has been achieved. One
such process is maximum likelihood expectation maximization (MLEM) [24, 25].
One of the issues with MLEM is that during the iterative process the estimated image can become
increasingly noisy (see Fig. 6). Smoothing can be used to control this noise, albeit with the consequential loss in
spatial resolution. Alternatively (and more commonly performed), this effect can be accomplished by stopping
the iterative reconstruction process earlier. As can be seen in Fig. 6, at 20 iterations the image looks suitable for
diagnosis. However, a consequence of stopping the iterative reconstruction early is that the true value of uptake of
the radiopharmaceutical is not reached. Furthermore, the level of convergence to the true value can vary in different
parts of the image, which means the relative accuracy will change across the image. A possible solution is to use
maximum a posteriori reconstruction, which uses prior information from, for example, CT or magnetic resonance
imaging to regulate and control the iterative process more effectively [26, 27]. Such reconstruction algorithms are
being implemented in PET/CT systems, and are being introduced into SPECT/CT systems [28]. An alternative is to

FIG. 5. Diagrammatic description of iterative reconstruction.

6
 FIG. 6. Formation of image with increasing numbers of iterations.

use FBP algorithms for imaging techniques where accurate quantification is required, although this technique has
artefacts of its own, and it is difficult to implement correction algorithms into the reconstruction.
Another issue with MLEM is the speed of the reconstruction process. It is mathematically demanding to
perform MLEM reconstruction, particularly at 120 or 128  projections, with each containing 128 × 128 pixels.
This becomes even more challenging when reconstructing electrocardiogram (ECG) gated acquisitions, with up
to 16 bins for each projection. Ordered subset expectation maximization (OSEM) is commonly used to speed up
the reconstruction process, typically by a factor of ten [29]. In this technique, instead of using the complete set of
projections, only a subset of estimated and acquired projections is compared at a time in a series of subiterations.
Hence with a typical OSEM reconstruction of 2  iterations and 20  subsets using a 120  projection acquisition,
20 subiterations of 6 angle comparisons will be made for each full iteration. The resultant image is found to be
equivalent to 40 (= 2 × 20) full iterations of MLEM. As with FBP, following MLEM or OSEM reconstruction,
post-reconstruction filtering is often applied to produce images with acceptable noise characteristics.

2.1.4. Corrections

On gamma camera based SPECT systems, there can be significant mechanical stress on the detector-to-gantry
mounts as the system rotates during the SPECT acquisition. As a result, the mechanical centre of the field of view
(FOV) from the perspective of the detector can be different to the reconstructed centre of the image originally
assumed by the reconstruction software. For example, on a dual detector system, with the detectors in an anterior and
posterior position, the centre of the FOV is likely to be in concordance with the assumed centre of the reconstructed
image (see Fig. 7(a)). However, when both detectors are in a lateral position, the force of gravity may position the
mechanical centre slightly lower than the software assumed centre (see Fig. 7(b)), while detectors in a 90 degree
configuration (see Fig. 7(c)) demonstrate a mixed picture with one detector in concordance, and the second detector
in discordance with the assumed centre of rotation (COR). As the detectors rotate to different positions during a
SPECT acquisition, the gravitational effect on the detectors and the resulting shift away from the assumed COR
will change. This effect can be accounted for by a COR correction (see Section  3.2.4), which characterizes the
difference in the mechanical and image centres of rotation, and is applied prior to image reconstruction. Several
COR corrections are often required depending on the number of collimator and detector configurations available,
with each providing slightly different mechanical challenges to the gantry.
Other corrections can be applied during the reconstruction process. Correction for the attenuation of photons
within the body is commonly applied in SPECT. For applications in which the attenuating material can be assumed
to be constant in its density, calculated attenuation based on defined boundaries and a linear attenuation coefficient
can be used  [30]. Application of this technique is most frequently seen in neurological SPECT studies. An
alternative method for areas with non-uniform attenuation is to use transmission imaging for AC. Such systems can

7
use radionuclide transmission sources, or now more commonly CT, to provide tomographic maps of attenuation.
Figure 8(a) shows how, for a point source, the projections from the estimated image can be derived using ‘forward
projection’ in the iterative reconstruction. By adding the information from an attenuation map (see Fig. 8(b)), it can

FIG. 7. Mechanical stress on a dual detector system. (a) When the detectors are in the anterior/posterior position, the projection of
the mechanical COR corresponds with the software COR. (b) With detectors in the lateral position, this might not be the case. (c) In a
90 degree configuration, the differences in mechanical stresses can require a different COR correction.

FIG. 8. Estimates of projections in iterative reconstruction (a) without attenuation, (b) including the effect of attenuation and
(c) including the effects of both attenuation and resolution losses.

8
be seen how the estimated projections might change depending on the position of the source. This is how AC can
be incorporated into the iterative reconstruction. A similar technique can be used to model resolution losses into
the reconstruction. Figure 8(c) shows how, instead of a single point representation of the source, the probability of
where the count could occur can be spread depending on the source-to-collimator distance — a distance affecting
the spatial resolution of the image. Corrections for other physical effects, such as scatter, can be incorporated
into the reconstruction process [31], although simpler multiple energy window methodologies are currently more
commonly performed before the reconstruction process [32, 33].

2.2. COMPUTED TOMOGRAPHY

2.2.1. Acquisition

CT is similar to SPECT in terms of the reconstruction process. Using a rotating X ray tube detector system,
projections are acquired at multiple angles and reconstructed to form three dimensional images. The main difference
between the techniques is that the projections in CT use the measurement of the attenuation of an external X ray
radiation source to form the image, whereas in emission tomography an external detector measures the distribution
of in vivo radioactivity. Thus, the radiation fluence in CT is much more intense and much shorter in duration than
in SPECT because the radiation source can be switched on and off at will, and the intensity of the beam can also
be selected and collimated to the volume slices of interest. Nor does it dwell before and after imaging, which is the
case in emission imaging.
The hardware involved in CT is quite different to that in SPECT. Opposite to the X ray source is an array
of X ray detectors, typically 16–64  channels covering 2–4  cm in axial width and a few degrees in arc in the
axial direction (see Fig. 9(a)). The array is arranged to capture multiple slices of information simultaneously
(see Fig. 9(b)), albeit with a much reduced axial extent compared to SPECT detectors. Slice configurations can use
some or all channels to produce a multitude of slice options. The intense radiation used in CT allows for the rapid
continuous rotation of the source and detector, which typically sit on a slip ring to allow multiple uninterrupted
revolutions. Following each revolution, the patient couch can be moved to increase the amount of the patient
scanned, using what is known as axial mode. Alternatively, the imaging couch can move continuously during the
gantry rotation to acquire tomographic data in helical mode.

FIG. 9. (a) Schematic diagram of a typical CT system, showing an X ray source detector array pair with arc. (b) The detector array is
organized over a few degrees of arc in the axial direction and can acquire multiple slices of data simultaneously. A subgroup of eight
channels in this example can be chosen to limit the axial extent, if required.

9
 On account of the continuous motion and high photon flux (typically five orders of magnitude greater than
seen in nuclear medicine), CT scans involve many more projections in the reconstruction. Furthermore, unlike in
SPECT, the reconstructed matrix size in CT is typically fixed to an array of 512 × 512 pixels within a reconstructed
slice. If smaller pixels are required, or if there is a lot of empty space in the image, a smaller fan beam (collimated
field size) is used. In the axial plane, the array of detectors can be configured to produce more or fewer slices per
rotation (see Fig. 9(b)), depending on the slice thickness required and level of noise. CT acquisitions have a number
of parameters which define the type and quality of the image (see Fig. 10). The following subsections describe the
reasons and consequences of parameter choices.

2.2.1.1. Tube voltage

The potential difference across the CT tube determines the maximum energy of X ray photons produced.
Typically in a range of discrete values between 80 kV and 140 kV, higher potential differences are chosen when
more penetrating radiation is required (i.e. for large patients), with smaller voltages used with smaller and paediatric
patients. The contrast in the image depends on the tube voltage chosen (see Section 2.3.1), with lower voltages
creating larger differences in attenuation and therefore better contrast. A consequence of using higher tube voltages
is that, for a given tube current, the radiation output and dose rate from the tube increases (almost by the square of
voltage), thus increasing patient radiation doses (for a fixed rotation time). On some modern systems, tube voltage
can be automatically chosen by the CT system, based on the preview scan.

2.2.1.2. Tube current

The current across the tube determines (for a given kV) the number of X ray photons produced. As mA
increases, the number of photons produced increases linearly, which results in more photons being detected and

FIG. 10. An example of a CT acquisition configuration screen, showing the different parameters.

10
therefore an improved image quality resulting from higher signal to noise ratios, but also a higher patient radiation
dose. On many CT systems, the mA can be modulated to decrease automatically the radiation output in areas or
projections with low attenuation. For example, when scanning the chest, the values of attenuation are less than
when scanning the pelvis, so it is possible to reduce the photon flux (i.e. mA) when scanning the lung region
compared to the pelvic region. Similarly, for most patients, the thickness in the anterior–posterior direction is less
than that when acquiring lateral projection data. It is possible therefore to vary the mA used at different angles
depending on patient thickness. Tube current varies between 10 mA and 700 mA, with currents above 500 mA only
(and seldomly) used for special applications.

2.2.1.3. Rotation time

In addition to the tube current, the total number of X ray photons detected for a given projection is determined
by the dwell time at that angle, which is given as the rotation time. A complete 360 degree gantry rotation typically
takes between 0.3 s and 2 s. The rotation time used often depends on the application: in nuclear cardiology, where
there is a need to ‘freeze’ cardiac motion, the shortest rotations times are used with 180 degree reconstructions.
For most other applications, rotation times of around 0.5–1.0 s are used. The total number of photons reaching the
detector is proportional to the product of tube current and exposure time (referred to as milliampere-seconds, mAs),
with the values of the two components being determined by the temporal resolution required.

2.2.1.4. Scanning mode

Scanning mode is normally used to describe whether the scanner table moves following each rotation of the
gantry (axial mode), or if it moves continuously during the rotation motion to scan the patient in helical mode. If
helical mode is selected, a further parameter needs to be defined. The pitch determines how fast the table moves
during a rotation. Higher pitches (wider spirals) offer shorter scanning durations and lower radiation doses at the
cost of poorer axial spatial resolution and higher noise. A third scanning mode is the cine or dynamic mode, which
produces multiple images at a given body position and is typically used for tracking CT contrast over time or for
dealing with motion, such as respiration.

2.2.2. Image reconstruction

Reconstruction of CT data offers similar options to those used in SPECT. Up until 2008, almost all CT
scanners only utilized FBP for image reconstruction. The technique in CT is similar to that used in SPECT, with
projections from the rotating X ray tube detector pair undergoing filtering and back projection to produce a three
dimensional stack of two dimensional images. There are, however, some subtle differences. One difference is that
CT requires a series of blank scans or air calibrations to reconstruct the attenuation map. An air calibration or blank
scan is simply a scan that is performed without an object in the CT FOV. Typically performed at the beginning of
the day for a series of kV, mA and detector configurations [34, 35], air calibrations are pulled into the reconstruction
process so that the attenuation from the patient can be calculated as a ratio to the signal measured with nothing in
the FOV.
The geometry involved in CT is also different from that used in SPECT. In CT, a fan beam geometry is used
in which a point source of X rays irradiates an arc of detectors to yield non-parallel projection data for a given
image angle (see Fig. 11(a)). Furthermore, with the axial extent of the detector array now reaching up to 16 cm
in specialist volumetric scanners, data are also no longer parallel in the axial direction either (see Fig. 11(b)). For
ease of reconstruction, it is normal to group parallel projections before the reconstruction starts (see Fig. 11(c)).
An additional challenge in reconstruction is when data are acquired helically. In this instance, interpolation of
projection data is used to allow the reconstruction of a stack of transaxial slices (see Fig. 11(d)).
The second option for reconstruction, as in SPECT, is iterative reconstruction. This technique was the original
form of reconstruction available for CT. However, because of computational and image quality issues, the technique
was quickly replaced by FBP. More recently, an improvement in computational power and algorithm developments
has led to the re-emergence of the technique. Iterative reconstruction in CT is basically similar to that used in
emission tomography. Following a first estimate from FBP, a series of iterations, typically involving forward and
back projection, are used to calculate the best match between an image estimate and the measured projection

11
FIG. 11. Geometry involved in CT. (a) Most CT scanners work with a fan beam geometry. (b) With extended axial extent. (c) For
reconstruction parallel projections, data are rebinned prior to reconstruction. (d) Helical scans require interpolation to derive slice
information from the acquired data.

data. The advantages over standard FBP is that by including a model of the noise process and the imaging system
(e.g. X ray source and detector array), image artefacts and noise can be potentially reduced, thereby improving
overall image quality. The success in reducing image noise allows lower dose CT protocols to be adopted. Each
manufacturer has its unique ‘black box’ solution to iterative reconstruction, taking subtly different approaches.
A good general description of iterative reconstruction in CT and the approaches followed by the manufacturers is
given by Willemink [34]. A fuller, more mathematical description is given in the review by Beister et al. [36].

2.3. CT BASED ATTENUATION CORRECTION FOR SPECT

2.3.1. Attenuation correction

In order to obtain reliable quantitative images in SPECT, it is necessary to determine how attenuation by
the overlying tissues and organs influence acquired data. Even without a need for quantification, the relationship

12
between observed activities in tissues of different density and at different depths in the body is strongly influenced
by attenuation and the inherent inconsistency of non-corrected projections. Historically in SPECT (as well as in
PET), AC was based on transmission sources (either lines or points). In PET, the correction can be performed in
the projection space before using FBP because the correction factor for a given projection element depends only
on the properties of the attenuating object and not on the source localization. In SPECT, however, the correction
also depends on the actual source depth. Therefore, the implementation in SPECT requires that the correction
be included in an iterative reconstruction loop. Single photon emitters which have been used for attenuation
measurements are 153Gd and 133Ba. However, due to the limited photon flux available from such sources, the
acquisition time is considerable; and even when simultaneous acquisition could be used, the necessary corrections
for the ‘cross-contamination’ between emission and transmission counts would significantly contribute to the noise.
In this regard, AC in SPECT using transmission sources has not been widely adopted in clinical use.
With the introduction of CT to SPECT systems, the adoption of AC has become more widespread. The
information provided in a CT  measurement is basically that of tissue specific attenuation and, compared to the
results from transmission sources, this ‘μ map’ is essentially noiseless. The reconstructed CT image contrast owes
its anatomical information to the fact that different tissues have slightly different composition and density, and
therefore different (linear) attenuation coefficients. The standard Hounsfield unit (HU) of CT numbers is calculated
from a linear scaling of the reconstructed attenuation coefficients, assigning the values −1000 to air and 0 to
water [37]. However, photon attenuation is strongly dependent on the photon energy, and therefore a ‘translation’
or ‘downscaling’ of the μ map is needed between the effective CT energy and the relevant photon energy used for
emission imaging. For PET, one such translation for each CT energy suffices, since all nuclides used in PET have
the same annihilation photon energy of 511 keV. For SPECT, the need for downscaling to individual single photon
energies adds to the complexity.
The energy dependency of (mass) attenuation coefficients for all elements is different but well known [38, 39],
and the mass attenuation of a compound material or mixture is determined by its elemental composition alone, each
element having its own dependency on energy. Overall, the mass attenuation cross-section at the energies relevant
to CT and SPECT is dominated by the sum of Compton scattering σc and photoelectric absorption τ, while pair
production κ cannot occur (see Fig. 12).
Compton scattering is almost independent of the atomic number (Z) and has a low dependency of the photon
energy (E), while photoelectric absorption depends strongly on both Z (power of 3–4) and E (power of −3). At
effective CT energies of 50–80 keV, and for the soft tissue components, Compton scatter alone is of significance,
while for bone the photo absorption is important too (see Fig. 13).

2.3.2. Models for attenuation correction calculation from CT data

Each pixel of a CT image represents a mixture of elements, and the same composite value (in HU) can in
principle result from different underlying mixtures of unknown elemental composition and densities. Therefore, the
problem of energy translation does not have a strictly unique mathematical solution free of assumptions. Empirical
approaches have been shown to solve the problem adequately in most situations. However, exceptions can lead to
artefacts, which are covered in subsequent sections.
It is evident from Fig. 13 that a simple scaling with one common HU-to-μ value factor cannot account for
the difference in materials, even among homogeneous pixels. A segmentation of the CT volume into the two major
components of bone and non-bone (soft tissue), assigning a single SPECT μ value to each of the two domains, on
the other hand cannot itself reflect the fact that both types of tissue can still have very different densities (e.g. in the
lung volume the density is only ~0.3 g/cm3).
The common way to resolve this is to use a pixel wise scaling via a bilinear function (see Fig. 14). It represents
the μ value as a continuous and unique function of HU and, therefore, does not require image segmentation. The
method generally assumes the presence only of air (−1000 HU), water or soft tissue (~0 HU) and bone (~1000 HU).
The lower part of the curve (line) intercepts at zero attenuation and therefore represents a simple scaling
corresponding to the density of a mixture of air and soft tissue. The upper curve (line) represents a mixture of soft
tissue and bone. The separation point corresponds to pure soft tissue (of density ~1 g/cm3) and the slope accounts
for an increasing amount of compact bone (~1.9 g/cm3) in the mixture. Originally proposed only for use with the
highest available CT energy (voltage 140 kV) to minimize the scaling factors and the effects of their potential
uncertainties, it has later been generalized to variable CT energy as well as to variable SPECT photon energy [40].

13
FIG. 12. Diagram with (E,Z) dependency of mass attenuation coefficients. The two ‘curtain edges’ are formed by curves representing
the points in an (E,Z) diagram where two of the interaction cross-sections τ, σ and κ are equal. The horizontal dashed line (Z ≈ 7)
shows that, for soft tissue, Compton scattering is dominating between 20 keV and 25 MeV. The vertical line at 140 keV shows that, for
99m
Tc, Compton scattering is the more important effect up to Z = 30. The line at 511 keV (PET) is shown for comparison.

99mTc 131 I
1000
Bone
Soft Tissue
100
Ti
Fe
μ (cm−1)

10 Iodine contrast
Ba-contrast

1 Au
Hg

0.1
10 100
E (keV)
FIG. 13. Linear attenuation coefficients (cm−1) for some materials of interest in SPECT/CT. At low energy (mainly photo absorption),
compact bone and soft tissue differ by an order of magnitude owing to differences in Z; at higher energies (Compton scattering), they
differ only by the density factor (<2). Two contrast media (iodine and barium), two typical metals for prostheses (titanium and iron)
and elements for dental work (mercury and gold) are shown (see also Table 1, Section 2.3.2). In CT contrast materials (iodine and
barium) and in metallic implants or dental work, photo absorption is dominant for CT as well as for 140 keV photons.

14
FIG. 14. Translation of HU to linear attenuation coefficients at 140 keV for SPECT (based on Ref. [40]). Example of an implementation
of the bilinear method for three CT energies (given as kV) and conversion to SPECT (140 keV).

Once the μ  map is known, it can be included as part of the system matrix in the loop of an iterative
reconstruction (e.g. attenuation weighted OSEM). Violation of the assumptions by the presence of other materials
(e.g. contrast agents and metal) can result in image artefacts, which are shown and discussed in later sections.
Various proprietary extensions of the conversion methods have been introduced to reduce these effects. Table 1
shows linear attenuation values at a CT effective energy of 80 keV, at 140 keV (99mTc) and 364 keV (131I), as well as
the CT to SPECT ratios, the ‘downscaling factor’, for some tissues and materials of interest.

TABLE 1. LINEAR ATTENUATION COEFFICIENTS FOR CT AND SPECT

μ (cm−1) Ratio (downscaling factor)

Density
Material
(g/cm3)
80 keV 140 keV 364 keV 80/140 80/364

Soft tissue 1.06 0.193 0.162 0.116 1.20 1.67

Compact bone 1.92 0.428 0.295 0.198 1.45 2.17

Ti (Z = 22) 4.5 1.83 0.794 0.427 2.30 4.27

Fe (Z = 26) 7.8 4.64 1.67 0.770 2.77 6.03

Iodine contrast 1.3 1.23 0.401 0.151 3.07 8.15

BaSO4 contrast 1.5 1.61 0.502 0.174 3.22 9.27

Au (Z = 79) 19.3 107 42.7 4.98 2.50 21.4

Hg (Z = 80) 13.6 71.5 30.8 3.58 2.32 20.0

Note: Mass attenuation values are calculated using XCOM [38] and multiplied by density to provide linear attenuation
values. The composition and density of soft tissue and bone are from Ref. [41]. Values for the four elements shown
(iron, gold, mercury and titanium) are for pure elemental material. The μ values at 80 keV for gold and mercury are
calculated as a mean of the values above and below the K edge. The values for the contrast agents are for the material
before administration and dilution in the patient: the iodine contrast is an aqueous solution of a commercial product
containing 300 mg/mL of iodine; and the barium contrast is a suspension in water containing 600 mg/mL of BaSO4.

15
2.4. BENEFITS OF SPECT/CT

2.4.1. Functional mapping

A key benefit of SPECT/CT is the ability to map function derived from nuclear medicine imaging onto
the anatomical information from CT. With the exception of neurological applications, where the poor soft tissue
contrast of CT is limiting, the advantages of SPECT/CT are very clear.
In oncology, many radiopharmaceuticals are specific for tumour physiology, which can be extremely
beneficial, particularly for agents that are also used for radionuclide therapy. An obvious consequence of this
is that these tracers also provide limited anatomical information, which makes the localization of metastatic
or nodal disease difficult. An example of the benefits of functional mapping is shown in Fig. 15. In this
123
I-metaiodobenzylguanidine (mIBG) study, the SPECT images suggest that the uptake is in the liver, but it is not
completely clear which areas of the liver actually show this uptake. Once the CT images are provided, not only is
it possible to determine the regions of the affected liver tissue, but this can be done with great anatomical accuracy
on the CT image.
In cardiology, the use of myocardial perfusion SPECT to assess coronary artery disease is an extremely
powerful examination that helps cardiologists to distinguish between reversible and irreversible ischemia.
Although helpful, the data do not offer information on the vessels that might be causing the ischemia. A diagnosis
of irreversible ischemia typically results in the patient being referred to the interventional suite for an invasive
fluoroscopic angiography to confirm the location of the occlusion and to explore whether an interventional
technique can be performed to clear the blockage. With SPECT/CT, there is a non-invasive alternative. Information
on vessel occlusion can be explored using CT coronary angiography, which together with myocardial SPECT
can potentially provide information about functional and anatomical deficits in one patient visit. Unfortunately,
CT coronary angiography requires a high specification CT scanner, which is only available in a small number
of clinical SPECT/CT systems. An example of a combined cardiac perfusion emission scan with a CT coronary
angiography investigation performed in the same visit is shown in Fig. 16.
A functional mapping application of SPECT/CT that has seen significant success is in the assessment of
orthopaedic injury. Bone SPECT, with agents such as 99mTc-methylene diphosphonate (MDP) or hydroxyethylene
diphosphonate (HDP) is very good at highlighting areas of high bone turnover that are caused by fractures or
mechanical stresses. It also provides some anatomical information to localize these injuries. However, the superior
anatomical detail offered by CT provides better localization than SPECT alone, with the exquisite spatial resolution
also providing additional information on the type of bone injury. Combining bone SPECT and CT has therefore
proved incredibly helpful in the assessment of bone fractures and mechanical stresses (see Fig. 17).

2.4.2. Quantitative SPECT

An advantage of accurate AC provided from CT is the possibility of performing quantitative SPECT. One of
the earliest applications of quantitative SPECT/CT was in the field of nuclear cardiology to assess reversible and
irreversible ischemia, with visual interpretation typically involved in the assessment of relative perfusion deficits.
This generally works well. However, in larger patients the attenuation from breast tissue in females or from the
diaphragm in males can lead to apparent reductions in uptake which are not real and purely the consequence of
photon attenuation. This creates subsequent problems when the data are used to derive bullseye plots and compared
to normal databases. Calculated AC such as those described by Chang  [30] are ineffective for this application
because the assumption of uniform attenuation throughout the body is clearly not valid in the thoracic region, where
soft tissue, lung and bone attenuation is present. Applying AC with CT on a multimodality SPECT/CT system can
clearly be helpful, as can be seen in Fig. 18. In this example, the inferior wall of the myocardium highlighted in the
slice data shows relatively homogenous uptake in the AC data (upper slice of each pair), while when no attenuation
correction (NAC) is applied, there appears to be reduced perfusion in the lower slice of each pair. This is also
observed in the bullseye or polar representation on the right of Fig. 18, where comparisons to normal databases
are typically made. On the right bullseye plot (no correction), there appears to be reduced perfusion which is less
significant in the left bullseye plot (with correction).
In neurology, the benefit of CT based attenuation correction (CTAC) for quantitative analysis is less clear.
Although quantification in clinical, neurological SPECT is performed, particularly in dopamine transporter and

16
FIG. 15. 123I mIBG SPECT/CT study of the lower thorax/upper abdomen. The SPECT only study highlights uptake of the
radiopharmaceutical in the liver. With the addition of CT, the affected area of the liver becomes clearer and the SPECT uptake is also
shown to correspond with features on the CT.

17
 FIG. 16. Combined emission perfusion scan and CT coronary angiography performed in the same imaging session.

FIG. 17. SPECT, CT and fused SPECT/CT of an orthopaedic bone injury.

regional cerebral blood flow imaging, a calculated Chang approach is typically used for AC. This method can
work well; however, difficulties with highly attenuating head supports, and defining brain outlines, can lead to
inaccuracies and variability, which can be overcome using a low dose CT for AC.
A more recent application of quantitative SPECT is being implemented in oncology imaging. With the
expansion of radionuclide therapies, there is growing interest in using quantitative SPECT to determine activity in
tumours or organs so that dosimetric estimates can be performed. After some form of calibration, it is possible to
use CTAC together with other corrections (e.g. scatter) to determine the activity concentration in these areas [42].

18
FIG. 18. Cardiac SPECT/CT study with upper slice of each pair of slices showing CTAC data, while the lower slice of each pair
represents data with no AC applied. At the bottom, AC (left) and NAC data (right) are shown as bullseye plots. The yellow arrows
indicate AC recovering attenuated counts (left) and diminished in the NAC data (right).

By applying the techniques over a series of images taken at different time intervals, it is then possible to determine
the accumulated activity and therefore the organ or tumour dose. Until recently, this technique had been performed
with users applying their own corrections and calibrations. However, manufacturers are now integrating information
and corrections into their own algorithms to provide direct activity concentration values in SPECT images with
greater ease.

19
2.5. COMMERCIALLY AVAILABLE SPECT/CT SYSTEMS

Since its inception by Hasegawa in the early 1990s (see also Ref. [43]), all the major manufacturers have
released SPECT/CT systems with various specifications and capabilities. A description and comparison of some of
these systems can be found in Seret et al. [44].

2.5.1. General Electric

In 1999, General Electric was the first of the major manufacturers to release a clinical SPECT/CT system
with its Millennium VG Hawkeye product. The system used a low mA (2.5 mA) dental X ray tube and fan beam
detector system placed on the same slip gantry as the gamma camera detectors. Rotating at a slow speed of 16 s per
rotation, this 140 kV system provided axially acquired data with a 1 cm slice thickness using a photon flux (mAs)
adequate for AC of SPECT data and localization of SPECT features. In the following years, updates were released
providing 2 × 5 mm and the 4 × 5 mm slice thicknesses with a choice of kV (120 kV or 140 kV) and the possibility
of performing high pitch helical scans in order to reduce scan time and dose.
General Electric has since released the Optima 640 and a range of systems in the Discovery 670 series with
varying degrees of functionality. The Optima 640 is a SPECT system with a four slice CT scanner, offering slice
thicknesses of 2.5 mm and 5 mm, a tube current range of 10–30 mA, two tube voltages (120 kV and 140 kV) and
the possibility of axial and helical scanning. The Discovery 670 series uses 8 and 16 slice CT subsystems with a
wider range of tube voltages, rotation times, tube current and slice thicknesses, and includes options such as CT
iterative reconstruction and semiconductor (CZT) detector technology. This series is designed for a wide multitude
of SPECT/CT applications with the exception of cardiac CT. For combined cardiac SPECT and CT applications,
General Electric has produced the Discovery 570c, which incorporates a CZT multipinhole cardiac SPECT system
with a 64 slice CT subsystem capable of acquiring CT coronary angiography studies.

2.5.2. Mediso

The SPECT/CT product from Mediso is the AnyScan system, which offers the possibility of SPECT, CT and
PET in the same gantry. In terms of the SPECT/CT system, the AnyScan provides a 16 slice CT system with a wide
range of tube voltages, currents, pitches and slice thicknesses.

2.5.3. Philips

The first SPECT/CT product from Philips was the Precedence SPECT/CT. Based on the Skylight SPECT,
the two models Precedence 6 and Precedence 16 are based on traditional 6 and 16 slice CT systems, respectively.
Following the Precedence, Philips released the Brightview  XCT. This system used the standard conventional
Brightview platform and added a flat panel cone beam CT subsystem with a 14 cm axial FOV and fast (12 s) and
slow (60 s) rotation speeds. Using a tube voltage of 120 kV and current range of 5–20 mA, this low dose approach
is in some ways similar to the initial strategy taken by General Electric with the Hawkeye systems, although with
smaller slice thicknesses (0.33 mm) and the capability of acquiring a 40 cm SPECT FOV in three gantry rotations
by using 180 degree reconstructions.

2.5.4. Siemens

The three SPECT/CT products from Siemens are the Symbia T2, T6 and T16 with 2, 6 and 16  slice CT
subsystems, respectively. The two slice system has more limited functionality compared to the T6 and T16;
and because of this, it is more limited in the clinical applications for which it can be used. The T6 and T16 are
effectively identical albeit with a different number of slices that can be acquired per rotation, and can be used for a
wide range of clinical uses with the exception of advanced cardiac applications. Siemens have updated the product
series to include the Symbia Intevo and Intevo Excel xSPECT systems, which automatically draw in corrections
and quantitative reconstruction methods to provide quantitative values of uptake in the image in terms of kBq/cm3
or standardized uptake value (SUV). These systems also allow the possibility to use iterative reconstruction of
CT data.

20
 3.  QUALITY CONTROL OF SPECT AND SPECT/CT

A quality control programme for SPECT and SPECT/CT systems is necessary to ensure optimal clinical
image quality. Elements of such a programme should include system calibrations and corrections as well as testing
of system performance characteristics at an interval consistent with that of a potential change in, or failure of, that
particular characteristic. These tests should be inclusive of both SPECT and CT (if applicable) components of the
system.
Several organizations, such as the IAEA, the American Association of Physicists in Medicine (AAPM),
the European Association of Nuclear Medicine (EANM) and the National Electrical Manufacturers Association
(NEMA), have independently developed recommendations for gamma camera/SPECT and SPECT/CT quality
control and testing. Although these recommendations have many similarities in what should be tested, they also
have differences in the testing method and frequency. Overall, there is consensus that testing should be performed
for both the planar as well as the tomographic imaging modes of a system and that the planar tests should be
conducted under both intrinsic as well as extrinsic (i.e. with collimators in place) conditions. A list of tests specific
to planar/SPECT and CT components of a hybrid system and their frequency as recommended by the different
organizations is presented in Table 2, Section 3.1. In addition, a brief description of only the SPECT tests in
the recommendations is provided in the following sections. (No such descriptions are provided for planar tests,
since this publication is specific to SPECT imaging.) More detailed description of these tests can be found in the
references given.
Prior to any testing, updated system calibrations and corrections should be performed by a medical physicist
or a service engineer to ensure optimal system condition. Depending on the gamma camera make and model, these
calibrations and corrections should include at a minimum:

—— Detector energy correction and spatial distortion (non-linearity) maps;

—— Intrinsic detector (including radionuclide specific) and system (also known as extrinsic, with collimator)
uniformity maps;
—— Peaking (radionuclide photopeak energy window centring);
—— Detector high voltage/photomultiplier tube gain adjustments (tuning);
—— SPECT COR/multiple head registration (MHR); and SPECT/CT FOV registration.

3.1. ROUTINE PERIODIC PLANAR, SPECT AND SPECT/CT QUALITY CONTROL TESTS

There are a number of gamma camera quality control tests that are performed at various frequencies to ensure
the proper operation of the scanner. These tests are divided between planar, SPECT, SPECT/CT and CT related
tests. A list of these tests and their frequencies as determined by various organizations is provided in Table 2.
A description of each of the SPECT and SPECT/CT tests is in Sections 3.2–3.4.

3.2. SPECT AND SPECT/CT TESTING

Of the group of tests specific to planar mode operation, some quality control tests are also essential for
optimal SPECT and SPECT/CT imaging. These tests include planar uniformity, peaking and pixel size.

3.2.1. Planar uniformity

Uniformity is a measure of a gamma camera’s response to a uniform source of irradiation. It is quantified

intrinsically as the degree of uniformity exhibited by the detector itself or extrinsically as the uniformity exhibited
by the combination of the detector and collimator together. Depending on the particular gamma camera make
and model, it may be more convenient to acquire either intrinsic or extrinsic floods on a daily basis. If intrinsic
floods are acquired daily, then periodic (e.g. weekly or monthly) acquisition of extrinsic floods for the various
collimators should be specified to monitor for collimator damage. If extrinsic floods are acquired daily, then

21
TABLE 2. QUALITY CONTROL TESTS AND THEIR FREQUENCIES

IAEA AAPM NEMA EANM

[45] TG177 [46] NU 1-2018 [47] [48, 49]

System checks

Physical inspection A A+Y A

Safety interlocks A+Y D

Collimator and detector head mountings D D

and collimator damage

Detector head shielding leakage A A+Y A A

Basic computer timing ½Y

Computer timing in dynamic acquisition ½Y

ECG gated acquisition ½Y

Multiple window spatial registration A + ½Y A + Y* A A + ½Y/Y

Computer monitors A+Y

Good practice checks

Energy calibration of pulse height analyser D D

Background count rate D D

Flood field uniformity and sensitivity D D D

Film processing (if available) D

Planar tests

Peaking W

Intrinsic uniformity for 99mTc A+W A+M A A + W/M

Intrinsic uniformity other isotopes ½Y Q

Intrinsic resolution and linearity A + ½Y A+Y A A + ½Y

Intrinsic hydration ½Y A+Y *

Intrinsic count rate A + ½Y A+Y A *

Energy resolution A+Y A A

System uniformity A + ½Y

System resolution A+W A+W A A

System planar sensitivity A + ½Y A+Y A

System count rate with scatter A

22
TABLE 2. QUALITY CONTROL TESTS AND THEIR FREQUENCIES (cont.)

IAEA AAPM NEMA EANM

[45] TG177 [46] NU 1-2018 [47] [48, 49]

Whole body system resolution A + Y* A A+Y

Whole body scanning constancy *

Collimator penetration and scatter A

Collimator hole alignment A *

SPECT

Centre of rotation A + W/M M* A W/M

Pixel size A + ½Y A + ½Y

SPECT uniformity A + ½Y A+Y

SPECT resolution in air A + ½Y A+Y A A + ½Y

SPECT resolution with scatter A + ½Y A+Y A *

SPECT image contrast A + ½Y A+Q ½Y

Slice thickness A + ½Y

System volume sensitivity A

Detector to detector sensitivity A A

Sensitivity/uniformity with rotation A

Head to head registration A

SPECT/CT

SPECT/CT image quality A+Q A

SPECT/CT spatial registration A+Y A M

CT

CT number accuracy/linearity D A+M

CT dose assessment A+Y

CT image quality assessment D D

Note: A — acceptance; D — daily; W — weekly; M — monthly; Q — quarterly; ½Y — half yearly; Y — yearly; * — optional.

periodic (e.g. weekly or monthly) acquisition of an intrinsic flood should be specified, as collimator responses may
mask subtle intrinsic detector non-uniformities.
Intrinsic uniformity is usually measured using a point source (~37 MBq of 99mTc) placed at a distance
equivalent to 5 FOV from the detector surface, to expose the entire detector to a uniform photon flux (within 1%);

23
while extrinsic uniformity is usually measured using a 57Co sheet source placed directly onto the detector head.
In both cases, data can be acquired for a total of 5–10 million counts using a matrix size of 256 × 256 pixels. On
some systems, floods (intrinsic or extrinsic) of both scanner heads can be acquired at the same time. It is important
to remember that, depending upon the initial activity, a new 57Co sheet source may have sufficient 56Co and 58Co
impurities to cause high dead time, and thus suboptimal floods, owing to their high energy gamma emissions that
easily penetrate through the septa of most collimators used clinically. Fortunately, the half-lives of 56Co (77 d) and
58
Co (71 d) are much shorter than 57Co (272 d), so their effect will dissipate over time.
Intrinsic and extrinsic uniformities are evaluated using visual assessment, which involves comparing the
flood images to earlier ones or to expert observer judgement as to whether or not the image quality is acceptable.
Alternatively, NEMA uniformity quantitative results may be calculated and compared to manufacturer’s
specifications or to established limits based on initial or past results. These calculations include integral and
differential uniformity (IU, DU) for the useful and central fields of view (UFOV, CFOV), where IU and DU are
calculated as:

max− min high − low

IU = × 100 (%) DU = × 100 (%)
max + min high + low

where min and max are the minimum and maximum counts in pixels lying within the UFOV or CFOV, and high
and low are the maximum and minimum count difference in any five contiguous pixels in the X and Y directions
within the UFOV or CFOV. The CFOV is the area defined by 75% of the UFOV linear dimensions (see Fig. 19).

FIG. 19. Extrinsic uniformity results from a General Electric Millennium gamma camera.

3.2.2. Peaking

Peaking is the ability to centre an energy window over a radionuclide’s gamma or X ray emission photopeak
that is used for imaging. Peaking should be performed intrinsically using a point source of about 37 MBq at a
distance equivalent to at least 5 FOVs from the detector to expose the entire detector to a uniform photon flux.
Peaking should be performed for all radionuclides used clinically. Peak shifts should be ≤ ±3% of the photopeak
energy for 99mTc and 57Co and ≤ ±5% for other radionuclides. Peaking is automated on some systems, while other
systems require manual positioning of energy windows. On most systems, the energy spectrum is displayed as a
histogram of counts versus energy with energy window limits shown as vertical lines. It can be useful to look at this
spectrum prior to imaging patients. A narrower energy window provides better scatter rejection but also reduces the
number of unscattered counts recorded in the image. Flood images acquired off-peak are characterized by visible
tubes (see Fig. 20), while an on-peak image shows no such tubing effect (see Fig. 21).

24
FIG. 20. Flood images of 99mTc (140 keV photopeak; energy window of 20%) acquired off-peak, using a photopeak centred at 126 keV
(top row) and 154 keV (bottom row) for both heads.

99m
FIG. 21. Energy spectrum and flood image of Tc (140 keV photopeak) acquired on-peak showing a uniform count density across
the image.

25
3.2.3. Pixel size

This test is used to verify the accuracy of the pixel size. In SPECT, this value has implications for quantitative
evaluation applicable for AC as well as determining organ sizes from resultant tomographic images. Pixel size is
determined using point sources or line sources (according to NEMA) placed at known locations along the X and Y
directions and imaged in planar mode. The centroid of these sources is then determined and the distance between
them calculated. This distance is then divided by the number of pixels that separate the point sources to determine
the pixel size in the X and Y direction. It is essential that the distance between the point sources (typically ≥10 cm)
be accurate to within 1 mm.

3.2.4. Centre of rotation and multiple head registration

The COR measurement verifies the alignment between the projection of the SPECT system axis of rotation
(AOR) and the centre pixel column(s) in the acquired projection image matrices. The MHR test, on the other hand,
is used to ensure that each detector head in a multiple detector SPECT system samples the same volume — which
means that the image of a point source on one detector is exactly at the same location on a mirror image of a second
detector placed at 180 degrees apart from the first detector. It is necessary, however, that the detector heads always
be parallel to the AOR (without head tilt).
Both of these tests are conducted as a verification of system calibrations for COR and MHR. The calibration
of both COR and MHR is usually combined in one test to be performed for each scanner model according to the
manufacturer’s procedure manual. The correction is collimator dependent, and on some SPECT systems, separate
calibrations are required for all detector configurations (e.g. 180 and 90 degrees). The 90 degree configuration is
typically used only for cardiac SPECT imaging.
The verification tests are described in Refs [45, 47] and use point sources placed at pre-set locations within
the FOV of the scanner. A SPECT scan of the sources is then acquired, and the offset locations (mean, maximum
and standard deviation) of the point sources from the COR is determined for each angle. This process is repeated
for each detector head (see Fig. 22). Manufacturers have pre-set tolerance values for these offsets, and if the value
is exceeded, the test results in a failed outcome. In such cases, a service call should be placed to rectify the issue
before the scanner can be placed for clinical use.

3.2.5. SPECT uniformity

The purpose of this test is to evaluate the tomographic image uniformity. It is conducted using a SPECT
performance phantom (e.g. Jaszczak phantom). The inclusion of the cold spheres and rods sections in the phantom
is optional for this test; however, it is good practice, since they can potentially be used for the resolution and
contrast (image quality) tests described later. The phantom is filled with about 370 MBq of 99mTc and mixed well
before positioning at isocentre. Tomographic phantom imaging is then performed with 120–128  views, using a
128 × 128 matrix and the smallest possible radius of rotation. The number of counts per view varies between
different standards but is typically in the range of 240  000–500  000. Upon completion, image reconstruction is
performed with all manufacturer corrections, including AC using FBP with sharp filter [45] or a Hann/Butterworth
filter using an order and cut-off frequency that would be used for clinical studies [46].
The analysis of tomographic uniformity (see Fig. 23) also varies between the different standards
and ranges from visual inspection [46] to the calculation of contrast of an image artefact as a measure of
non-uniformity (see Ref. [45], which also includes a visual assessment of the images).

3.2.6. SPECT resolution

The purpose of this test is to evaluate the tomographic image resolution. The different methods performed
depend on the standard used and range from: the visual inspection of the rods section of a SPECT performance
phantom to determine the minimum size rods [46]; to an evaluation of the radial, tangential and axial full width
at half maximum (FWHM) of the point sources in air used for the COR test described in Section 3.2.4 [45, 47];
or determining the FWHM of three line sources in a water phantom [47]. Reference [45] also includes a visual
assessment of the resolution (see Fig. 24). In any one of these tests, it should be noted that resolution is highly

26
dependent on the radius of rotation used and that the obtained results in air are considerably different from those
measured in water due to scattering effects.

FIG. 22. COR calibration results using five point sources.

FIG. 23. Images of the uniform section of a SPECT performance phantom showing ring artefacts (left) and good image
uniformity (right).

27
FIG. 24. Solid spheres (left) and rods sections (right) of a SPECT performance phantom showing the different size spheres and rods.
In the left image, the fourth sphere (15.9 mm) is visible. The image on the right shows that the third (9.5 mm) or fourth (7.9 mm) rods
section is visible.

3.2.7. SPECT image contrast

The purpose of this test is to evaluate the tomographic image contrast. A SPECT performance phantom is
also used here with the solid spheres in place (see Fig. 24). Phantom data acquisition and image reconstruction are
performed as described in the previous sections. Image contrast is reported using visual assessment as the smallest
visible sphere.

3.2.8. Slice thickness

This test is unique to Ref. [45] and is used to test the thickness of a tomographic slice at the centre of the
FOV. The test uses a point source in air as described in Section 3.2.4. Analysis is performed by drawing a line
profile through the point source along the axial direction and the FWHM of the profile determined. The results of
this test are highly dependent on the radius of rotation, which should be set to 15 cm or, if this is not possible, then
the smallest radius possible. A circular orbit of rotation should be used.

3.2.9. System volume sensitivity

This test is unique to the NEMA standard [47]. The system volume sensitivity is the total system sensitivity
to a uniform concentration of activity in a specific cylindrical phantom. This measurement is dependent on detector
configuration, collimator type, radionuclide type, energy window setting and source configuration, among other
factors. The test is conducted using a uniform cylindrical phantom 20  cm in diameter. The phantom is injected
with a known amount of activity and positioned at isocentre. SPECT acquisition with at least 120 views, but no
more than 128, should be acquired, with each view acquired for 10 s. Field uniformity corrections or any other
mechanisms that alter the number of counts in these projection images are to be disabled (with the exception
of decay correction). The total count rate can then be determined by summing the total number of counts in all
projections and dividing it by the total acquisition time. The system volume sensitivity can then be determined by
dividing the count rate by the activity concentration.

3.2.10. Sensitivity and uniformity with rotation

This test is unique to Ref. [45] and is used to determine variations in system sensitivity as a function of
angular position of the detector. The test requires imaging a flood source of about 185  MBq that can be safely
attached to the collimator face. A tomographic scan is then acquired with different rotation speeds while keeping the

28
same number of views. Analysis is then performed by calculating the mean and standard deviation of the number of
counts per view for the different speeds. Any variation in sensitivity exceeding 1% as a function of angle (view) for
each speed can be attributed to a lack of magnetic shielding or mechanical drive problems.

3.3. CT TESTING

In addition to SPECT quality control testing, a hybrid SPECT/CT system requires that the CT component
of the system undergo a similar set of quality control tests. If the CT component is to be used for diagnostic
CT scans, then a complete evaluation of the CT scanner should be performed by a medical physicist, and the results
reported [35].

3.3.1. Tube warm up and air calibrations

CT quality control consists of several tests, the first of which is an X ray tube warm up procedure. This is
necessary to bring the X ray tube to its optimal operating temperature to stabilize temperatures of both the cathode
and anode and aid in extending the lifetime of the X ray tube. This procedure takes about one minute to finish.
Tube warm up is followed by an air calibration (or fast calibration) procedure, which comprises a series
of calibrations and system tests, such as checking the operation of the data acquisition system converter board,
collimator calibration, gantry balancing, mylar window cleanliness, focal spot position, and finally an array of
detector gain calibrations performed at different X ray tube voltages based on user predefined settings. The overall
procedure takes about 20–30 minutes depending on scanner manufacturer and model as well as the user predefined
voltage settings.

3.3.2. Uniformity, noise, linearity, resolution and artefacts

Following tube warm up and air calibration, a CT number uniformity, linearity, image resolution and low
contrast detectability test is then performed using a water filled phantom which includes inserts. The phantom,
which is supplied with the scanner, allows for all these tests to be performed in a single scan and is placed on
a holder and then positioned centrally in the FOV (see Fig. 25). A scan of the phantom is acquired with pre-set
parameters, but typically at 120 kV, the most common clinically used tube voltage. Regions of interest (ROIs) are
then drawn on the resultant images at different locations and slices to evaluate CT number uniformity (see Fig. 26),
linearity (accuracy), resolution (see Fig. 27) and slice thickness. The CT images are also evaluated for the presence
of artefacts (see Fig. 28). For uniformity, it is expected that the CT number in the ROIs does not change significantly
across the FOV. The standard deviation of CT numbers within the ROIs of the uniformity test are also used to assess

 
FIG. 25. Examples of CT phantoms.

29
 FIG. 26. CT image of the water filled phantom with ROIs drawn at different locations to determine the CT number uniformity.

image noise, and can be helpful in determining the effect different exposure parameters have on this measurement.
For linearity, the CT numbers for water, air and acrylic (all part of the phantom) should be within expected ranges.
In addition, the CT images of the phantom should be void of any artefacts.

3.3.3. CT dose assessment

In addition to the image quality evaluation, an assessment of radiation dose should be performed for the
CT system. The dosimetric quantity for CT is the computed tomography dose index (CTDI), which, although not an
actual dose figure, is a parameter which permits relative dose comparison among different scanners. Measurements
require a polymethyl methacrylate cylindrical CTDI phantom and an ionization chamber (see Fig. 29). This test
should be performed annually or following a CT X ray tube or detector array replacement. The CTDI phantom is
placed at isocentre in the FOV of the scanner with the ionization chamber placed in the central hole of the phantom.
A CT axial acquisition is then performed and the radiation exposure (or kerma) measured. This measurement is
repeated for four different peripheral positions of the ionization chamber within the phantom. A weighted average
of these five exposure (or kerma) measurements is used to calculate the CTDI value, which is then compared to the
value reported by the scanner. The process is repeated for clinically relevant CT acquisition parameter combinations
(kV and mAs). The measured and reported values should be similar to within manufacturers’ specified deviations.
In addition, a review of the predetermined dose notification levels that have been set as part of each
CT protocol may be performed. When a CT study is suspected to exceed the pre-set notification value, an alert

30
FIG. 27. CT image of the water filled phantom with ROIs drawn at different locations (white regions) corresponding to air, water and
acrylic to determine the CT number linearity.

message is generated before the exam is acquired in order to allow the user to modify the acquisition parameters.
This enables the user to reduce the patient dose, or override the alert if necessary. Furthermore, a dose alert
(different from a dose notification level) value can also be programmed into new CT scanners to trigger a message
when the cumulative dose at a location, plus the dose for the next planned and confirmed scan(s), is likely to
exceed a preprogrammed value. It is important to note that dose alert values are applicable across all protocols on
a scanner, while dose notifications are specific to protocols. These tools allow users to track patient CT doses and
ensure they are within specified limits.

3.4. SPECT/CT TESTS

3.4.1. SPECT/CT registration

For hybrid systems, a SPECT/CT registration test should be performed to ensure that the two scanners
are properly aligned. The motivation for SPECT/CT image registration is that spatial registration is critical for
accurate SPECT image reconstruction using CTAC and the display of fused images for clinical interpretation.
The standards of the AAPM [46] and the EANM [48, 49] are the only ones which describe this test. However, the
AAPM recommends that the manufacturer’s specific registration test phantom procedure and analysis tools be used
to evaluate the spatial registration accuracy (see Figs 30 and 31). Alternatively, the AAPM also suggests visually

31
 FIG. 28. CT image of the water filled phantom showing ring artefacts, which will necessitate a service call.

inspecting the SPECT/CT images of the SPECT performance phantom for spatial registration when acquired with
and without any table loading (placement of >50 kg on the patient couch to simulate a patient; see Fig. 32) [46].
An optional procedure in which point sources placed on the SPECT phantom that are visible on both SPECT and
CT images is also described. In this case, the alignment of the point sources on both images can be determined.
Currently, there are no generally accepted criteria for alignment accuracy between the SPECT and CT images;
however, an alignment mismatch that is less than the size of a SPECT image pixel should be acceptable.

3.4.2. SPECT/CT image quality

This test evaluates the overall SPECT/CT image quality when using CT based AC while all processing and
reconstruction parameters that are recommended by the manufacturer for routine use in clinical SPECT/CT imaging
(e.g. scatter corrections, iterative reconstruction parameters and resolution recovery, among others) are applied. The
test uses the same SPECT performance phantom described in Section 3.4.1 with the resolution and sphere insets
included except that, in this case, CTAC is applied. The phantom is first filled with about 740 MBq of activity and
placed at isocentre in the FOV. CT and SPECT scans of the phantom are then acquired and reconstructed using
standard clinical protocol parameters. The resultant images are then evaluated visually for uniformity, resolution
and contrast in a manner similar to standard SPECT imaging. Alternatively, a NEMA PET image quality phantom
could be used for this assessment, which is more applicable to body imaging.

32
FIG. 29. CTDI phantom with an ionization chamber placed at the centre of the phantom and positioned at isocentre in the FOV of the
scanner.

FIG. 30. SPECT/CT registration tool from a manufacturer, which includes ten point sources (right) seen in the fused SPECT and CT
images (left and middle).

33
FIG. 31. SPECT/CT registration tool (left) from a manufacturer, which includes six line sources seen in the SPECT and
CT images (right).

FIG. 32. Fused SPECT/CT images of a SPECT performance phantom that can be used for the assessment of the alignment of the
SPECT and CT scanners of the hybrid system.

34
 4.  ARTEFACTS

Clinically useful diagnosis from SPECT and SPECT/CT imaging depends on accurate interpretation
of the images and corresponding quantitative results. This means that nuclear medicine professionals should
be acquainted with the different patterns associated with images of different organs or parts of the body being
examined. However, possible image abnormalities or artefacts related to machine, user or patient factors might
not be recognized by these professionals, which could hinder optimal patient management. This section provides
examples of abnormal patterns and artefacts caused by errors owing to system, user and patient sources and their
manifestation on resultant SPECT and SPECT/CT images. This section also describes methods to resolve these
abnormalities and artefacts.

4.1. SYSTEM RELATED ARTEFACTS

In order to deliver good and quantifiable SPECT images, it is essential that artefacts related to system factors
should not be present. These artefacts can manifest themselves as rings, which can be caused by, for example,
detector non-uniform response, COR offset or the wrong uniformity correction map. Other factors such as
inadequate energy peaking, head tilt, head to head registration, mismatched sensitivity and mechanical problems
can also generate artefacts that might hamper both visual inspection and quantification of clinical data.

4.1.1. Non-uniformity of detector response

Background

The response of a scintillation camera when subject to a uniform flux of photons should be uniform over the
whole FOV. When non-uniformities are present, full or partial ring artefacts will appear in the tomographic slices.
These can be caused by one or more of, but not limited to, the following factors:

(a) Poor detector uniformity;

(b) Small differences in the responses from the photomultiplier tubes;
(c) Non-linearity of the detector and electronics;
(d) Defects in the collimator;
(e) Crystal hydration.

Gamma camera quality control tests (see Section 3 for more details) should be performed before SPECT
studies, so that most non-uniformity artefacts can be detected and corrected for prior to tomographic acquisitions.

Case A

Projections of a Jaszczak phantom filled with 99mTc were acquired for a radius of rotation of 17 cm and
using a 128 ×  128 matrix. The 128  projections were reconstructed with FBP and Chang-AC. Figure 33 shows
the reconstructed images presenting ring artefacts in three of the four slices in the uniform section, indicating the
presence of detector non-uniformities.

Guidance

Poor detector uniformity may result in ring artefacts in the reconstructed SPECT slices, even when not
noticeable in planar image, as the tolerances permitted are far less than those accepted in conventional planar
imaging.

35
 FIG. 33. SPECT images of a Jaszczak phantom showing ring artefacts in several slices, indicating detector non-uniformities.

Case B

SPECT image quality was tested using a SPECT performance phantom. Data were acquired with a low
energy ultra high resolution (LEUHR) collimator, 128 × 128 matrix, 128 projections and 16 cm radius of rotation.
FBP reconstructed slices presented ring artefacts, depicted in Fig. 34. Daily uniformity checks showed a degrading
trend of the uniformity parameters, although still within recommended values by the manufacturer.

Guidance

When daily uniformity parameters tend to degrade (e.g. for an ‘old’ camera) even when they are within the
manufacturer’s recommended limits, poor planar uniformity is amplified by the reconstruction process. SPECT
image quality tests should therefore be carried out in order to ensure that no damaging effects would affect the
clinical results.

Case C

A myocardial perfusion study was performed on a patient using 99mTc labelled methoxyisobutylisonitrile
(MIBI) and the stress images presented defects at the apex and anterior wall that were not consistent with the
patient’s clinical data (see Fig. 35). Daily routine tests were checked, and it was verified that one of the heads
presented non-uniform response over part of the FOV, as shown in the top row of Fig. 36. The system was

36
 FIG. 34. Ring artefacts present in some of the FBP reconstructed Jaszczak phantom images, obtained with an LEUHR collimator.

FIG. 35. 99mTc-MIBI myocardial perfusion results obtained with one detector of a dedicated cardiac system out of calibration.

recalibrated and the uniformity maps became normal, as indicated in the bottom row of Fig. 36. The patient study
was repeated and the new images were in accordance with his clinical information (see Fig. 37).

Guidance

It is necessary to check carefully, both visually and numerically, the daily operational tests before starting
clinical studies in order to avoid mistakes or repeating patient studies.

37
 FIG. 36. Uniformity maps for both heads before recalibration (top) and after recalibration (bottom).

FIG. 37. 99mTc-MIBI myocardial perfusion results obtained for the same patient as in Fig. 35 after recalibration of the faulty detector.

4.1.2. Crystal hydration

Background

Crystal hydration is a rare event in gamma camera systems. When it occurs, however, it can have a large
impact on both planar and SPECT images.

38
Case

After an incident involving a water leak in a SPECT scanning room, the SPECT system was tested and
hydration was found in both crystals of a dual head camera (see Fig. 38). SPECT data of a NEMA PET image
quality phantom is shown in Fig. 39. Data were acquired using a 128 × 128 matrix, 120 projections of 20 s and
reconstructed without AC.
On-peak images show that the hydration is only visible on head 2. The non-uniformity of these two close cold
spots form the half-ring artefact shown in the SPECT images. Once both crystals had been replaced, the system
functioned properly.

Guidance

Off-peak hydration tests should be carried out when physical factors, such as water leaks, affect the
environmental conditions in the room that houses the SPECT or SPECT/CT scanner. It is important to note that not
all hydration spots visible in off-peak flood images result in on-peak image artefacts.

4.1.3. Poor correction maps

Background

Nuclear medicine imaging systems have both intrinsic and extrinsic uniformity maps that are used to correct
for non-uniformities present in the detector and the collimated detector, respectively. The use of an appropriate
correction map is essential if non-uniformity artefacts are to be avoided. When liquid flood phantoms are not
mixed properly, or when collimators are exchanged between systems of the same scanner model, the uniformity
maps might introduce artefacts in the SPECT images. Furthermore, outdated correction maps that do not match the
current state of the detector can introduce image artefacts.

FIG. 38. Lower (126 keV) and upper (154 keV) off-peak (left and right) and on-peak (middle) flood images showing hydration spots in
both crystals of a dual head SPECT system.

39
 FIG. 39. Half-ring artefact caused by crystal hydration of a dual head SPECT camera.

Case A

SPECT uniformity test was performed with a Jaszczak phantom filled with 99mTc. FBP reconstructed and
uniformity corrected slices presented ring artefacts (see Fig. 40). It was verified that this defect was created by a
poorly mixed uniformity liquid flood phantom.

Case B

Uniformity corrected slices of a SPECT performance test phantom presented ring artefacts, as shown in
the top row of Fig. 41. When uniformity correction was excluded from the reconstruction process, the artefacts
disappeared (shown in bottom row). The loaded uniformity correction map was found to be outdated; and when the
proper map was used, ring artefacts did not appear.

Guidance

When creating tomographic uniformity maps using a flood tank, it is important that the radioactive solution
be properly mixed to ensure that the system correction map only corrects for varying detector performance. A prior
planar image of the phantom can help to ensure that the phantom is properly mixed. Where a choice of uniformity
maps is available, it is important that the correct map be chosen; and where collimators may be interchanged
between systems, any system uniformity correction should be for the desired detector–collimator combination.

4.1.4. Centre of rotation offset

Background

Ring artefacts that can occur owing to COR errors are not clearly seen on clinical images but result in a loss
of resolution or blurring in the reconstructed image. Therefore, evaluation using appropriate test phantoms, such as

40
 FIG. 40. Partial ring artefacts caused by a poorly mixed fillable flood phantom.

FIG. 41. Ring artefacts resulting from using an outdated uniformity map (top row) and reconstructed slices without uniformity
correction (bottom row).

multiple point sources, should be carried out according to manufacturer’s recommendations. COR testing aims to
guarantee correct alignment of the physical COR with the reconstruction COR.

Case

A SPECT performance phantom filled with 99mTc was imaged with a 128 × 128 matrix and 128 projections.
FBP reconstruction and Chang-AC were performed. Figure 42 depicts the slices presenting ring artefacts in all of
them, indicating offset error (i.e. a shift of the physical COR in relation to the centre of the matrix of reconstruction).

Guidance

A periodic check of COR offset is essential in order to ensure that no distortions or ring artefacts be introduced
in SPECT studies, even if planar uniformity maps do not present inhomogeneities.

4.1.5. Energy peaking

Background

Before acquiring a SPECT image, it is important to ensure that the correct radionuclide be selected and that
the camera be peaked correctly on the chosen radionuclide.

41
FIG. 42. Ring artefacts in a sequence of SPECT images of a performance test phantom owing to COR error. The centre of the rings
corresponds to the axis of rotation

Case

Two 99mTc-MIBI myocardial perfusion studies were performed. Patient (a) was acquired on-peak (centred at
140 keV) and at low off-peak (centred on 125 keV), as shown in Fig. 43(a). Patient (b) was acquired on-peak and
at high off-peak (centred on 155 keV), as shown in Fig. 43(b). If the energy window is placed too low, more scatter
radiation will be acquired in the projections, diminishing the contrast (see Fig. 43(a)). Conversely, with the energy
window placed high, less scattered radiation is present and borders are better defined, albeit at the cost of a lower
count rate. The resulting SPECT slices are shown in Fig. 44.
Due to the increase of scatter in the low off-peak images, resolution and contrast are compromised with
thicker myocardial walls and more non-myocardial uptake respectively. Using a high off-peak energy window
improves contrast — but with more noise — although this appears to be minimal in this example owing to high
acquired counts.

42
FIG. 43. Projection views of a myocardial perfusion study (a) imaged at 125 keV and 140 keV with the lower energy window used and
(b) energy peak and projection views at 155 keV and 140 keV using the higher energy window.

FIG. 44. Projection views of two myocardial perfusion studies: patient (a) imaged at low off-peak and on-peak; and patient (b)
imaged at high off-peak and on-peak. The centre of the energy window settings is also shown for each image.

Guidance

The energy window needs to be well centred in order to produce good contrast, high resolution and low noise
images. In the case of a large contribution of scattered radiation, the window can be shifted towards the high energy
end, with the acquisition time per projection increased in order to keep an acceptable noise level.

43
4.1.6. Head tilt

Background

COR artefacts typically present as ring artefacts that show throughout every slice of a tomographic dataset.
In some instances, however, a ring artefact might only be present in a small number of slices. While this may be
due to detector non-uniformity, another cause for this type of artefact is head tilt (i.e. detector sag in the axial
(Z) plane). With head tilt, the COR might only be correct for some, but not all, slices, giving the slice dependent
COR ring artefact.

Case A

Ring artefacts can be observed in reconstructed slices from a dual detector SPECT system (see Fig. 45). As
the uniformity maps for both detectors did not have noticeable inhomogeneities (see Fig. 46(a)), COR offsets were
checked and found to be within the manufacturer’s limits (see Fig. 46(b)). Once the head tilts were adjusted and a
new calibration performed, all slices were brought within tolerance, and ring artefacts were not seen in the resulting
uniformity transverse slices (see Fig. 47).

Case B

The reconstructed slices of a myocardial perfusion study showed abnormal shapes in all projections. It
was verified that the detectors were tilted, converging towards the gantry. The heads were repositioned and new
rest study was performed. The top row of Fig. 48 shows the tilted head slices, while the bottom row shows the
corrected ones.

Guidance

When ring artefacts are present in uniform phantom slices, with planar uniformity not showing
inhomogeneities and COR testing within acceptable limits, head tilt can be suspected as the cause of the artefact.

FIG. 45. Ring artefacts present in the slices further away from the centre along the AOR of a uniform cylinder.

44
FIG. 46. Acceptable uniformity and COR corrections suggesting that SPECT artefacts are not associated with uniformity or COR.

FIG. 47. Reconstructed slices of the uniform cylinder after adjusting COR deviations.

45
 FIG. 48. Myocardial perfusion study reconstruction with tilted head (top row) and with no tilt (bottom row).

Improved COR calibrations can resolve the issue. If this is not successful, the manufacturer’s engineer may need to
be called to fix the head tilt.

4.1.7. Mismatched sensitivity

In poorly performing systems, there may be a change in detector sensitivity as the scanner rotates around the
patient. It is expected that this could have an adverse effect on image quality.

Case

A Jaszczak phantom filled with a 99mTc solution was imaged while positioned on the imaging couch of a
SPECT/CT system (see Fig. 49). Because the couch attenuated the signal from posterior projections as demonstrated
in the figure, this acquisition simulates a potential change in sensitivity of a SPECT system with acquisition angle.
Images showing the uniform section of the phantom with different reconstructions, together with a vertical
profile through this slice with each reconstruction is shown in Fig. 50. Using FBP and OSEM reconstructions
without AC, the profile displays a slight (~15%) reduction in counts between the bottom edge of the phantom
compared to the top edge. A small difference can also be seen visually in the rim of these phantom images.
This would be expected given that the projection counts were reduced in these posterior stations. No significant
difference was seen between the iterative and FBP methods.
Once Chang-AC is applied, this difference is less apparent (but still visible because it does not account for
bed attenuation) in both the image and profile data. For reference, the data iteratively reconstructed with OSEM
and CTAC, accounting for the bed attenuation, are given as truth for the phantom data. It shows no difference
between upper and lower phantom rims, signifying that any difference found in the other reconstructions are real.
Interestingly, there is also a bias between CT and Chang-AC, which is likely augmented by the inappropriate
choice of a linear attenuation coefficient for this phantom.

46
FIG. 49. (A) Projection from an acquisition of a Jaszczak phantom clearly showing a reduction in counts from imaging couch
attenuation and (B) the change in total counts acquired for each projection.

FIG. 50. Uniform slice from the Jaszczak phantom reconstructed with four reconstructions: (left to right) OSEM with no correction
for attenuation; FBP with NAC; FBP with Chang-AC; and iterative CTAC. The profiles from each reconstruction are also shown.

47
Guidance

Although SPECT reconstructions are relatively tolerant of detectors with changing sensitivity at different
angles, all efforts should be made to ensure that detector sensitivity is not affected by detector angle, particularly if
quantification is to be performed.

4.1.8. Phantom related artefacts

Background

When a phantom is used to test the characteristics and performance of a SPECT system, high activity spots
may be present in the reconstructed slices if radioactivity adheres to the phantom edges or its internal structures.
Conversely, cold spots may indicate the presence of air bubbles in the mixture. In addition, ring patterns may
appear if there are defects in the phantom.

Case

A dual head SPECT system was testing using a Jaszczak phantom with a 128 × 128 acquisition matrix,
800 kcounts per projection, OSEM reconstruction with 8 subsets and 3 iterations, and a Butterworth filter order 5
and 0.7 cut-off frequency. Chang-AC was applied with μ = 0.12 cm−1. Figure 51 shows reconstructed images across
all slices of the phantom, with a ring located at half the radius in the resolution rods section that is not observed
in the uniform section of the phantom. This defect was caused by damage to the rods section when it was dropped
onto the floor.

Guidance

The condition of a SPECT testing phantom should be carefully checked when used to verify the image quality
of a SPECT system, as damage to the phantom can hamper the results of the test.

4.2. ACQUISITION RELATED ARTEFACTS

Several acquisition parameters can be selected during a SPECT image acquisition, including: collimator
type, number of projections, time or counts per projection, matrix size, FOV and zoom factor, which define the
pixel size, rotation direction, type of orbit, radius of rotation and detector head orientation. All of these parameters
should be optimized in order to minimize image artefacts. In general, data are acquired over 360 degrees, with the
number of projections chosen depending on the detector FOV and the linear sampling interval. In cardiac studies,
however, data are acquired with the detectors positioned in 90 degrees and over a 180 degree arc, from RAO to
LPO projections.

4.2.1. Matrix and pixel size

Background

Pixel size is determined by dividing the camera’s FOV by the matrix dimension. Images with large pixel size
are characterized by lower image resolution. Conversely, smaller pixel size produces images with potentially better
resolution at the expense of a higher noise level. Zoom factors can be applied during acquisition to increase the
image area occupied by the same region of an object in the unzoomed mode, resulting in a better resolution.

Case A

A 99mTc filled Jaszczak phantom was imaged with a dual head SPECT system with a low energy high resolution
(LEHR) collimator, a radius of rotation of 20 cm, 120 projections with 800 kcounts per projection and acquisitions

48
 FIG. 51. Ring artefacts in the rods section from damage to the SPECT performance phantom.

matrices equal to 64 × 64, 128 × 128, 256 × 256, and 128 × 128 with a zoom factor of 1.33. Image reconstruction
was performed with FBP with a ramp filter (cut-off at 0.5  Nyquist frequency) and Chang-AC (μ  =  0.12 cm−1).
Representative slices in the cold rods and spheres sections are shown in Fig. 52, where improvement in image
resolution with decreasing pixel size, at the expense of noise level, can be verified. More specifically, in the rods
the 256 × 256 acquisition and the 128 × 128 acquisition with a zoom factor of 1.33 resulted in four sections
resolved, in contrast to three sections in the 128 × 128 unzoomed acquisition.

Guidance

The appropriate choice of matrix size and filter cut-off frequency is essential to preserve the resolution of
the imaging system and to keep the details present in the object. When appropriate, zoom factor applied during
acquisition can be used to enhance spatial resolution. However, attention should be paid to the number of acquired
counts to maintain an acceptable image noise level.

Case B

A striatal phantom was imaged with a radius of rotation of 15 cm, 128 × 128 matrix, 120 projections and
a pixel size of 2.1 mm. Figure 53(a) shows a FBP Butterworth (cut-off at 0.45 Nyquist frequency and order 10)

49
FIG. 52. Cold rods (top row) and cold spheres (bottom row) sections reconstructed slices of a 99mTc filled Jaszczak phantom, with
different matrix size, pixel size and reconstruction filter cut-off fixed at 0.5 Nyquist frequency: (left to right) 64 × 64, 8.84 mm,
0.028 cycles/mm; 128 × 128, 4.42 mm, 0.057 cycles/mm; 256 × 256, 2.21 mm, 0.113 cycles/mm; and 128 × 128 with zoom
factor = 1.33, 3.32 mm, 0.075 cycles/mm.

FIG. 53. Striatal phantom imaged with different pixel sizes (spatial frequencies): (a) 2.4 mm (0.094 cycles/mm); (b) 2.7 mm
(0.083 cycles/mm); (c) 3.9 mm (0.058 cycles/mm); and (d) 4.8 mm (0.047 cycles/mm).

50
reconstructed slice, after Chang-AC, where both caudate nuclei and putamina can be distinguished (dimensions
7–15 mm). When the pixel size is enlarged from 2.1 mm to 4.8 mm (see Fig. 53(b–d)), the separation between the
four structures decreases until almost forming one sole structure. The images also show that the noise level in the
background has smoothed out during this process. Note that the cut-off frequency has been maintained fixed at
0.45 Nyquist frequency. So, in terms of cycles/mm, it increases with decreasing pixel dimension. This also has an
influence on spatial resolution, improving it with decreasing pixel size.

Guidance

Proper selection of pixel size according to the object of interest dimensions will preserve the information
collected. Note that filter cut-offs defined as a fraction of the Nyquist frequency may need to be altered to keep the
same noise and resolution characteristics.

4.2.2. Number of projections

Background

Increasing the number of projections improves image quality while increasing acquisition time for a fixed
dwell time per projection. Decreasing the number of projections degrades the image quality, particularly at the
edges of the FOV because of the poorer spatial sampling in this area. Although spatial resolution is a function of
distance to the collimator, at the centre of the FOV it is not highly dependent on the number of projections.

Case A

SPECT performance tests were obtained with a Jaszczak phantom, using a 128 × 128 matrix and different
angular samplings for a circular orbit with a radius of 20 cm. Figures 54–56 show reconstructed slices for 120,

FIG. 54. Matrix size: 128 × 128, 120 projections.

51
 FIG. 55. Matrix size: 128 × 128, 60 projections.

FIG. 56. Matrix size: 128 × 128, 30 projections.

52
 FIG. 57. Slices in the cold spheres and rods sections for all angular samplings, showing differences in resolution.

60 and 30 projections, respectively. Figure 57 illustrates corresponding slices in the cold spheres and the rods
sections for all angular samplings. The same number of counts per projection was used in all acquisitions, so total
counts in the images differ. The degrading effect in resolution with decreasing number of projections can be seen in
both the cold spheres and the rods sections of the phantom. The effect of variable count density on the visualization
of a non-uniformity ring artefact is also evident in some slices.

Case B

A striatal phantom was imaged at a radius of rotation of 15 cm, matrix with 3.3 mm pixel size and the same
total counts, for angular samplings of 30, 64, 90 and 128 projections. Figure 58 shows the results of Chang-AC FBP
reconstruction with a Butterworth filter cut-off frequency set at 0.45 Nyquist frequency and order 10. Streaks are
more prominent in the 30 projections slice than in the 64 projections slice and are not present in the higher sampled
images.

Guidance

An adequate choice of linear and angular samplings will preserve the imaging system’s resolution and
produce tomographic images with fewer distortions.

Case C

SPECT performance tests were obtained with a Jaszczak phantom, using a variety of angular and linear
samplings (matrix sizes) for a circular orbit with a radius of 20 cm. The same total counts were acquired in all
studies. Figure 59 shows reconstructed slices for cold spheres and rods sections for 64 × 64, 128 × 128 and
256 × 256 matrix sizes with 120 projections. In addition, 30 and 60 projections data were also acquired for the
128 × 128 matrix.

53
FIG. 58. Striatal phantom imaged using the same number of total counts but with different angular sampling: (a) 30, (b) 64, (c) 90
and (d) 128 projections. Butterworth filter cut-off frequency fixed at 0.45 Nyquist frequency and order 10.

The degrading effect in resolution with decreasing number of projections can be seen more clearly in the rods
section slices. It can also be noticed that better image resolution and less distortion are achieved when the number
of projections is close to the matrix size, as in the case of 128 × 128 matrix with 120 projections. Furthermore,
there is no advantage in using a larger matrix, 256 × 256, due to the spatial resolution of the system and the higher
noise level.

Guidance

An adequate choice of linear and angular samplings will preserve the imaging system’s resolution and produce
tomographic images with fewer distortions. Attention should be paid to the number of counts to be collected, as the
image noise level is affected by the matrix size.

Case D

A Jaszczak phantom was placed under a dedicated cardiac study camera, with the two heads positioned at
90 degrees, and projections were obtained with 32 × 32 × 32, 64 × 64 × 64 and 128 × 128 × 64 matrix size and
projections. FBP reconstruction with Chang-AC was performed, and the results for a slice in the rods section are
shown in Fig. 60. Geometric distortions are seen at the border region of the half of the phantom that was imaged,
while strong blurring is present at the other half that was not directly imaged.

54
FIG. 59. Jaszczak phantom acquired with same total counts: (left to right) 64 ×64 × 120, 128 × 128 × 30, 128 × 128 × 60,
128 × 128 × 120, 256 × 256 × 120.

FIG. 60. Rods section slices obtained with a dedicated cardiac SPECT scanner (90 degree detector configuration, using a 180 degree
arc acquisition from LPO to RAO) reconstructed with FBP and Chang-AC: (top row) 32 × 32, 32 projections; (middle row) 64 × 64,
64 projections; (bottom row) 128 × 128, 64 projections.

55
Guidance

In cardiac studies, incomplete 180  degree angular sampling is adopted despite the fact that it introduces
geometrical distortion, as the contrast is improved when compared to 360 degree acquisition and the study time is
shortened. It should be noted that the patient can be positioned in either supine or prone positions (see Section 4.4.3).

4.2.3. Counts per projection

Background

As with planar nuclear medicine imaging, reducing counts per projection will result in increased noise, while
longer scan times can reduce noise and improve image quality.

Case A

A 99mTc filled Jaszczak phantom was imaged using 128 projections, a 128 × 128 image matrix and a range
of 100–1200 kcounts per projection. Data were reconstructed using FBP with the same reconstruction filter and
Chang-AC (see Fig. 61). There is a clear difference between 100 and 1200 kcounts per projection, although in this
instance, the improvement in image resolution and contrast between 800 and 1200 kcounts per projection is limited.

Case B

Another 99mTc filled phantom shows no clear ring artefacts at low (100) kcounts per projection, but does show
a non-uniformity artefact at high (1200) kcounts per projection (see Fig. 62). The two sets of data were acquired
one after the other on the same system.

FIG. 61. Transaxial slices through the rods section of a Jaszczak phantom at different numbers of counts per projections.

56
Case C

A patient was referred for a 123I-ioflupane scan to distinguish between idiopathic Parkinson’s disease and
essential tremor. Three hours following the injection of 185 MBq of 123I-ioflupane, the patient was imaged for 40 s
per projection using a 3 mm voxel size and reconstructed using OSEM iterative reconstruction. Using Poisson
resampling techniques, it is possible to derive images of different counts levels. In Fig. 63, images with between
4 and 40 counts per second are shown. Once more, there is a clear difference between the lowest (4 s) and highest
(40 s) times per projection; although in this example, there is no clear benefit of imaging beyond 24 s per projection.

Guidance

The number of acquired counts per projection should be chosen to match acceptable noise levels for the
specific application. For quality control purposes, using a Jaszczak phantom with very low levels of noise is
preferred, particularly to identify underlying problems leading to artefacts, whereas for clinical imaging, a higher
level of noise may be acceptable. Nevertheless, whichever application is used, there comes a point where increasing
the acquired counts per projection brings very little extra benefit to the imaging task.

FIG. 62. Ring artefact caused by detector non-uniformity which is visible at high (1200) kcounts per projection but barely visible at
low (100) kcounts per projection.

57
 FIG. 63. 123I-ioflupane imaging with varying levels of acquired counts per projection.

4.2.4. Acquisition arc

Background

In general, tomographic slices are reconstructions of projections acquired over 360  degrees around the
patient. However, for organs localized mainly on one side of the body, such as the heart, partial arc acquisition is
performed in order to improve the spatial resolution of the image by excluding projections that are further away
from the organ. Acquisitions typically use the two detectors positioned in a fixed 90 degree configuration moving
from the RAO to LPO position.

Case A

A Jaszczak phantom was placed at the centre of a dedicated cardiac SPECT system with a fixed 90 degree
head configuration, and both 180 degree and complete 360 degree arc acquisitions were obtained. Figure 64 shows
the reconstructed slices from the half circle acquisition (top row) and those from the complete circle (bottom row).
The top row clearly shows the distortions from the incomplete data in the rods section of the phantom, as well as
improved contrast in proximal regions due to improved resolution when compared to the results in the bottom row.

Guidance

With the exception of imaging organs localized mainly on one side of the body, acquiring SPECT data over
a 360 degree arc should be performed in order to minimize image distortion and to improve image resolution and
contrast.

58
FIG. 64. Reconstructed slices of the rod section of a Jaszczak phantom from an LPO to RAO 180 degree acquisition (top row) and a
360 degree acquisition (bottom row).

4.2.5. Collimator

Background

The ability of an imaging system to distinguish two identical point sources is described by its spatial resolution.
The collimator is one of the most important components that defines the spatial resolution of a gamma camera. The
choice of collimator depends on the radionuclide, and the compromise of sensitivity and spatial resolution required
for the underlying application.

Case

A striatal phantom was imaged with 99mTc, using a pixel size of 3.3 mm, with 120 projections and a radius
of rotation of 15 cm. In this instance, the time of acquisition was kept constant. Data were reconstructed with FBP
using a Butterworth filter with cut-off at 0.45 Nyquist frequency and order 10, and Chang-AC. Figure 65 shows
the results when imaging was done using collimators of the type: (a) LEUHR; (b) LEHR; and (c) medium energy
general purpose (MEGP).
The loss of spatial resolution varies with the type of collimator used. An LEUHR collimator may be used
for optimal spatial resolution but with the penalty of reduced system sensitivity and potential image artefacts due
to reduced count density. MEGP collimators typically have wider holes to offset losses in sensitivity caused by
increased septal thickness and therefore have poorer spatial resolution than low energy collimators.

59
 FIG. 65. Striatal phantom imaged with different collimator types (a) LEUHR, (b) LEHR and (c) MEGP.

Guidance

The appropriate collimator should be used for the application and radionuclide being used. In modern systems,
resolution modelling in the reconstruction could be used to compensate for resolution losses on more sensitive
collimators (poorer spatial resolution). MEGP collimators are (as the name indicates) intended for nuclides with an
energy higher than 99mTc (e.g. 111In, 177Lu and to some extent 123I).

4.2.6. Radius of rotation

Background

Image resolution in SPECT degrades with distance from the detector head. Increasing the radius of rotation of
the camera heads will therefore cause a loss of spatial resolution.

Case

A striatal phantom was imaged with 99mTc, using an LEHR collimator and pixel size of 3.3 mm, with
120  projections. Five radii of rotation were used, with the resulting images shown in Fig. 66. In addition, the
FWHM of a line source placed centrally in the FOV of the SPECT scanner and imaged with increasing radius
of rotation is plotted in Fig. 67. The phantom images and plot clearly show the non-linear resolution losses with
increasing radius of rotation.

Guidance

All efforts should be taken to minimize the radius of rotation used for SPECT imaging. The use of non-circular
orbits is often advantageous, except when imaging the brain on account of its shape and safety considerations.

4.2.7. Orbit effects

Background

In the human body, the length of the anterior–posterior axis is usually smaller than the left–right one.
Consequently, circular orbits will collect anterior and posterior projections further away from the patient than the
lateral views. As illustrated in Section 4.2.6, the spatial resolution of a gamma camera degrades with the distance of
the object to the collimator. Therefore, to improve spatial resolution, non-circular orbits, such as elliptical or body
contoured, are preferable.

60
FIG. 66. Striatal phantom images with increasing radii of rotation showing degrading image resolution with increasing radius of
rotation.

FIG. 67. Variation of spatial resolution with radius of rotation for a parallel hole collimator.

61
Case

A heart and lung torso phantom was used, with an anterior–posterior length of 20 cm and a left–right length
of 29.6 cm. In order to verify the effects of the different acquisition orbits with a dual head SPECT camera, the
heart was replaced by a set of two thick (2.5 cm in diameter) and two thin (1.5 cm in diameter) rods, mounted at
different depths (3.0 cm and 11.4 cm from the anterior wall of the phantom), and two saline bags were attached on
each side, simulating patient arms (see Fig. 68). A solution of 1.26 GBq of 99mTc diluted in 4 L of water filled the
phantom cavity. Projections were collected with LEHR collimators, 256 × 256 matrix and 120 angular samplings
for both circular (radius of 26.5 cm) and body contour (anterior–posterior radius of 21.9 cm and lateral radius of
26.5 cm) orbits, with a total of 117.7 Mcounts and 115.3 Mcounts, respectively, and a zoom factor of 1.33.
Figure 69 shows the SPECT transverse and sagittal slices of the phantom inserts, depicting the rods at
different depths. Reconstruction was performed without AC, using OSEM (8 subsets and 5 iterations, Butterworth
filter with cut-off frequency at 0.7  Nyquist frequency and order 5) and FBP (ramp filter with cut-off frequency
at 0.6  Nyquist frequency). Figure 70 shows the transverse slices for both reconstructions. Some loss of spatial
resolution and contrast in the circular orbit acquisition is visible for both reconstructions. Note also that OSEM
preserves the geometry better than FBP, as well as dealing with the noise level in the image.
Four transverse slices were summed and count profiles were drawn for double lines passing through the
centres of both anterior and posterior rods (see Fig. 71). Body contour orbit data provided a more consistent result,
especially for the deeper structures, because the cameras were closer to the phantom at the anterior and posterior
projections, when the radius of rotation was 21.9 cm, than in the case of circular orbit, where this distance was
26.5 cm, equal to the lateral radius in body contour orbit.

FIG. 68. Torso phantom with rod inserts on the imaging table with two saline bags attached to each side simulating patient arms.
Thick yellow arrows indicate the thick rods, and thin green arrows indicate the thin rods.

62
 FIG. 69. SPECT transverse and sagittal slices depicting the thin and thick rod inserts in the thoracic phantom.

FIG. 70. FBP (top row) and OSEM (bottom row) reconstructed cold rod inserts transverse slices for both body contour and circular
orbits. Some loss of resolution and contrast is more noticeable in the FBP reconstructed images.

63
FIG. 71. OSEM reconstructed transverse slices for both circular (left) and body contour (right) orbits. Profiles were drawn for a
double line across the centres of the (a) anterior and (b) posterior rods.

Guidance

On account of the geometry of the human body and whenever feasible, body contour (or elliptical) orbits
should be used for tomographic acquisition to keep the camera as close as possible to the body, minimizing any
loss of spatial resolution and contrast.

64
4.3. RECONSTRUCTION ARTEFACTS

The formation of a three dimensional dataset from two dimensional projection data is achieved by
reconstruction, either by FBP or by iterative reconstruction methods such as OSEM (see Section  2.1.3). FBP is
a direct, fast and linear method, but often results in high noise levels, including streaking artefacts. Iterative
methods offer a better control of noise levels, but are non-linear and their results depend on the number of iterations
and subsets (for OSEM) used and the corrections included (e.g. resolution recovery). Quantitative results of
reconstructions may depend on the choice of method, filters and parameters for AC (e.g. by the Chang method).

4.3.1. Filtered back projection streak artefacts

Background

In FBP, data are back projected into image space to form the image. While this is a fast and direct method to
derive the tomographic data, it can cause streak artefacts — particularly where there are areas of high uptake in a
low uptake background.

Case A

This patient was referred to assess orthopaedic pain around the knee. SPECT/CT was performed following
the injection of 99mTc-HDP (see Fig. 72). Although the salient features of the FBP reconstructed data within the
knee are the same as those reconstructed using OSEM, the FBP data have clear streak artefacts coming from the
back projection process which are not visible with OSEM. The reconstruction filter and scaling of both images are
the same.

Case B

Myocardial perfusion SPECT and 123I-ioflupane SPECT images reconstructed using FBP are shown in
Fig. 73. In the two images, streak artefacts are still clearly visible within and outside the patient boundary, but they
do not adversely affect the diagnostic quality of the data.

Guidance

FBP reconstructions create streak artefacts in the reconstructed data. Although this form of reconstruction
does not necessarily produce image data of inferior diagnostic quality, the streak artefacts mean it is not preferred
for imaging small areas of high uptake in a low uptake background.

FIG. 72. Anterior projection (left) showing the position of a transaxial slice reconstructed using FBP (middle) and OSEM iterative
reconstruction (right).

65
 FIG. 73. Myocardial perfusion SPECT (left) and 123I-ioflupane SPECT (right) reconstructed using FBP.

4.3.2. Filters

Background

Image resolution in SPECT changes with the type of filter used and its corresponding parameters. Increasing
image smoothing reduces image noise but at the expense of degrading image resolution.

Case

A striatal phantom was imaged with 99mTc, using an LEHR collimator and pixel size of 3.3 mm, with
120 projections. Data were reconstructed with FBP using a Butterworth filter order 10 and with increasing cut-off
from 0.30 to 0.60 Nyquist frequency (decreasing smoothing) with a 0.15 increment. In all cases, Chang-AC was
applied. The results are shown in Fig. 74, which clearly shows that as cut-off frequency increases (decreasing
image smoothing), image resolution improves but at the expense of increasing image noise.

FIG. 74. Representative striatal phantom images reconstructed using FBP with a Butterworth filter order 10 with increasing cut-off
from 0.3 (left) to 0.6 Nyquist frequency (right), showing improvement of image resolution but at the expense of increasing image noise.

66
Guidance

The choice of filter type and associated cut-off frequency should be balanced between image resolution and
noise. Increasing image smoothing reduces image noise but degrades image resolution.

4.3.3. Iterations and subsets

Background

Iterative reconstruction using expectation maximization methods such as MLEM make many attempts at
matching an estimated image to the measured projections, using the differences in the current estimate to help to
improve the following estimate. As a consequence, the estimated image becomes a better representation of the true
activity distribution with increasing number of iterations.
This process can be accelerated by using OSEM, where the estimated image is only compared with a subset
of projections each time. With OSEM, the number of iterations and subsets are defined. The number of subsets
equals the number of projections divided by the subset size. One iteration is defined as a pass through all subsets.

Case

A thorax phantom with myocardial, liver and background compartments were filled with a solution of 99mTc
and imaged using a SPECT system with transmission AC. Figure 75 shows the image becoming more defined with

FIG. 75. Thorax phantom data reconstructed using MLEM reconstruction with between 1 and 150 iterations.

67
increasing numbers of MLEM iterations. At a high number of iterations, the noise becomes amplified — a known
effect of iterative reconstruction.
Figure 76 demonstrates how OSEM can speed up the iterative reconstruction process. The left column
shows a reconstruction with 30, 60 and 120  MLEM iterations. Sixty projections were collected in the SPECT
acquisition. A subset size of 4 therefore corresponds to 15 (= 60/4) subsets. The middle column clearly shows
that: 2 iterations of 15 subsets are equivalent to 30 MLEM iterations; 4 iterations of 15 subsets are equivalent to
60 MLEM iterations; and 8 iterations of 15 subsets are equivalent to 120 MLEM iterations — each leading to a
huge saving in reconstruction time. The right column shows that even a subset size of 2 giving 30 subsets can be
used. The equivalence of 1, 2 and 4 iterations with these 30 subsets correspond to similar image features derived
from 30, 60 and 120 MLEM iterations, respectively.

Guidance

The formation of tomographic data using iterative reconstruction is defined using iterations in the case of
MLEM and iterations and subsets with OSEM. The number of iterations has an impact on image quality and on
the quantitative accuracy of the activity distribution. With OSEM, significant savings in reconstruction time can be
achieved using subsets of projections and the corresponding much reduced numbers of iterations.

FIG. 76. Thorax phantom data reconstructed using MLEM and OSEM reconstruction showing the equivalence and improvement in
reconstruction times using OSEM with a subset size of 4 and 2 (corresponding to 15 and 30 subsets, respectively).

68
4.3.4. Use of resolution modelling

Background

Accounting for losses in spatial resolution within iterative reconstruction is an option available in most modern
SPECT/CT systems (see Section 2.1.4). Much of the focus on these algorithms is their ability to reduce noise in an
image, and in doing so provide an opportunity to reduce acquisition time and/or injected activity [50]. However,
by using resolution modelling which incorporates point spread function (PSF) information, it is also possible
to resolve small lesions that might be undetectable using standard reconstruction approaches. Unfortunately,
resolution modelling can also produce overshoot artefacts (also known as Gibbs artefacts), which can overestimate
activity and produce apparent cold spots in larger lesions. Such artefacts are particularly prevalent in areas of high
contrast (i.e. where there are large changes in activity).

Case A

A transaxial slice from a 99mTc filled NEMA image quality phantom reconstructed using OSEM with and
without resolution modelling is shown in Fig. 77. In the largest sphere, the overshoot effect of resolution modelling
is represented by overestimation of activity at the edge of the sphere and underestimation in the centre. In the
smaller sphere highlighted in the profile, the overshoot artefacts superimpose, artificially elevating the counts in
the sphere. Although potentially helpful in visualizing lesions, it may overestimate the relevance of the lesions.
Furthermore, if quantification is required, the use of resolution modelling can adversely affect the values recorded.

Case B

Figure 78 shows contrast recovery curves from a 99mTc filled NEMA image quality phantom highlighting
maximum sphere count, and contrast ratio (maximum sphere count to average count concentration) for spheres
in a phantom filled with a sphere to background ratio of 5:1. Clearly the use of resolution modelling increases
the maximum count in each of the spheres compared to standard reconstruction for all sphere sizes. In terms of
contrast, this difference is reduced given that the background counts increase, too. However, it can be seen that the
filling contrast ratio of 5:1 is overestimated using resolution modelling, while for standard reconstruction the result
in the largest sphere converges to 5:1. Note that for smaller spheres, the partial volume effect limits the recovery of
the true contrast.

Guidance

Although the use of resolution modelling within reconstructions can improve the noise characteristics of
the data and lead to better visualization of small features, overshoot artefacts from these reconstructions can

FIG. 77. A transaxial slice of a 99mTc filled NEMA image quality phantom (A) with resolution (PSF) modelling and (B) without
resolution modelling. (C) Profiles through each image highlighting the effects of resolution modelling. All reconstructions also include
correction for attenuation and scatter.

69
FIG. 78. Contrast recovery curves from a 99mTc filled NEMA image quality phantom: (left) maximum counts in each sphere of the
phantom with a sphere to background ratio of 5:1; (right) contrast ratios in each sphere represented as the maximum count to average
count concentration in the background compartment of the phantom. Reconstructions were corrected for attenuation (AC) and
corrected for attenuation and resolution losses using resolution modelling (AC + PSF).

lead to overestimated activity concentrations and ‘cold’ centres to larger features. Visualization of data with and
without resolution modelling can help to understand when these artefacts occur. It has to be noted, however,
that quantification with these techniques is prone to overestimate, or under certain circumstances underestimate,
activity. Reducing the numbers of iterations used in the reconstruction can help to mitigate these artefacts, although
resolution modelling typically takes many more iterations to converge compared to regular OSEM reconstruction.

4.3.5. Chang attenuation correction

4.3.5.1. Bad contouring

Background

SPECT images of uniform phantoms can be attenuation corrected using the Chang technique. This technique
requires that an ROI be defined around the boundary of the phantom. The correction then assumes a uniform
attenuation value in every pixel within the defined region and uses that to perform the AC.

Case

SPECT images of a 99mTc filled Jaszczak phantom were acquired and reconstructed using Chang-AC. Four
different boundaries were created (see Fig. 79). One of which matched the phantom boundary and is considered the
gold standard, one larger than the phantom, one smaller than the phantom and one that is offset with respect to the
phantom.
The effects of the different boundaries are very clear in the image. The images are improved when using
Chang-AC compared to NAC images. However, an overly large contour will produce central areas of the phantom
with relatively fewer counts, while boundaries that are too small will result in relatively elevated counts in central
regions. A shifted boundary introduces a gradient of counts across the images.

Guidance

Boundaries that are badly defined in Chang-AC will result in artefacts in the AC images. Care should be
exercised to ensure that object boundaries are defined correctly.

70
(a) (b)

(c) (d)

(e)

FIG. 79. SPECT phantom images with Chang-AC using different boundaries: (a) correct boundary; (b) without AC; (c) large
boundary, showing an enhancement of the edges with respect to the central region; (d) small boundary, showing increased counts in
the central region; and (e) offset boundary showing non-uniform counts across the images.

4.3.5.2. Effect of different μ values

Background

The Chang-AC approach assumes a constant attenuation coefficient within the imaged object. This constant
value is dependent on the tissue density of the object and the energy of the incident photon. For soft tissue or water
at 140 keV (99mTc), many camera systems use a default value of 0.11 cm−1, compared to the narrow beam definition
of 0.15 cm−1, in an attempt to correct for scattered photons.

Case

A cylindrical SPECT phantom filled with a solution of 99mTc was imaged in a SPECT scanner. The acquired
data were reconstructed using Chang-AC with different μ values in the range of 0.07–0.15 cm−1. The reconstructed
images are shown in Fig. 80, which shows that too low a value and the count density in central areas will be
underestimated, while too high a value and the count density will be overestimated. Interestingly in this example,
the optimal linear attenuation coefficient for image flatness is 0.09 cm−1. This is a consequence of the acrylic walls
of the phantom and also the scatter environment in the phantom, with the lower value giving a flatter profile, but
not necessarily a more accurate value of count density. A similar effect may be seen in brain imaging owing to
the higher density and therefore attenuation of the skull. It should also be noted that the narrow beam value of
0.15 cm−1 is not appropriate for this wide beam source arrangement.

Guidance

The use of an adequate μ  value for Chang-AC of SPECT images is essential for good quality and more
accurately quantitative images. However, because no phantom or human brain has the same attenuation throughout,
the value of linear attenuation coefficient chosen to create the flattest image may not result in the most accurate
value of count density.

71
FIG. 80. Images through the uniform area of a Jaszczak phantom using Chang-AC with different values of linear attenuation
coefficient.

4.3.5.3. Non-uniform attenuation

Background

Chang-AC assumes a uniform attenuation of photons within the imaged object. However, if the object is
composed of different materials with different attenuation coefficients then the application of the Chang technique
will result in errors.

Case

A uniform phantom containing three cylindrical inserts of different materials (air, water and teflon) was
imaged in a SPECT/CT scanner (see Fig. 81). The acquired data were reconstructed using CTAC as well as
Chang-AC with a μ  value of 0.11 cm−1. The CTAC SPECT image is used as the gold standard, since CT can
account for the different attenuating material during the SPECT image reconstruction. In addition, the data were
reconstructed without AC. None of the cylindrical inserts contained radioactivity and hence the resultant images
should show no counts in these regions. As can be seen from the images, the use of Chang-AC resulted in image
artefacts and large errors in quantification (see Section 4.7) particularly in the air insert. The image without AC also
showed more counts at the phantom boundaries as well as low counts in the central region which are characteristic
of uncorrected images.

Guidance

The use of Chang-AC should be restricted to objects of uniform attenuation; otherwise image artefacts will
appear in the resultant images.

72
FIG. 81. SPECT and CT images of a uniform phantom with three cylindrical inserts: (top left) CT image with inserts clockwise from
top of water, teflon and air; (top right) iterative reconstruction with CTAC; (bottom left) FBP with Chang-AC; (bottom right) FBP
with NAC.

4.4. PATIENT RELATED ARTEFACTS

Image artefacts may be generated not only by system related factors, but also by the patients themselves. This
section includes patient motion (voluntary as well as involuntary cardiac motion), injection problems or residual
of tracer from previous examinations or treatments, together with specific problems in positioning and related
attenuation effects.

4.4.1. Voluntary patient motion

Background

In general, compared to planar imaging, SPECT requires a greater number of acquired counts to produce
robust, low noise diagnostic images. Ethical limitations to the injected activity that can be given to a patient mean
that high level of counts need to be achieved by relatively long scan times — normally between 15 and 45 minutes.
This obviously places a greater demand on the patient to remain still during the acquisition of image data. Although
for most patients this is achievable, in some cases long scan duration results in patient motion. This can make the
reconstructed data difficult, or even impossible to use, for diagnosis.

Case A

A 50  year old male with drug resistant frontal lobe epilepsy was referred for an ictal cerebral blood flow
study to help localize an epileptogenic focus. Immediately following the onset of an epileptic seizure, 485 MBq of

73
99m
Tc-hexamethylpropyleneamineoxime (HMPAO) was injected, with imaging performed 2.5 hours later. Twenty
minutes after the start of imaging, the patient suffered another seizure, moved and left the scanner FOV at angles of
114 and 117 degrees, respectively. The consequences of the motion and empty projections reconstructed with FBP
and OSEM are shown in Fig. 82.
In Fig. 82(a), the artefact caused by the incomplete projection data is seen in the FBP reconstruction. With
iterative reconstruction, the study might not be recognized as being artefactual if the reconstructed data were shown
in isolation. However, the lack of counts in the lower left and the upper right of the transaxial image are caused by
the missing projection data.

Case B

A rest myocardial perfusion study was performed using 99mTc-MIBI. In the left panel of Fig. 83, the linogram
shows that the patient moved during the study. Motion correction was applied to the data to produce the linogram
in the right panel. The bullseye plots from uncorrected and corrected data show that on the uncorrected data, the
perfusion deficit is at a 9–10  o’clock position, representing the anterior wall of the heart. Once corrected, the
perfusion deficit moves to the apex of the heart (centre of the bullseye plot).

Guidance

During SPECT studies, it is important that the technologist fully explain the procedure and its duration prior
to patient set-up. When positioning the patient, it is equally important to ensure that the patient is comfortable, and
that appropriate immobilization straps are used to minimize any potential patient motion. It is important that the
doctor writing the report of the images be made aware that motion has occurred. Although it is always possible
to perform quality control for motion in projection or sinogram data, it may not always be possible to spot in the
reconstructed data that the patient has moved. Therefore, the projection images should be checked before releasing
the patient.

4.4.2. Involuntary patient motion

Background

A SPECT acquisition can take anywhere between 10 and 40 minutes. Consequently, any image of the
mediastinum will create a time averaged image of uptake within the heart, unless that acquisition is ECG gated.
However, because the heart spends most time in or close to the end diastolic phase, the myocardial uptake is
relatively well defined and not grossly affected by cardiac motion.

FIG. 82. Cerebral blood flow study showing an artefact caused by incomplete projection data: (a) FBP reconstruction; and
(b) iterative reconstruction (OSEM).

74
FIG. 83. A rest myocardial perfusion study: (left) sinogram and linogram showing patient movement during the SPECT acquisition
which, before correction, presents with a perfusion deficit in the patients’ anterior wall; (right) after motion correction of the sinogram
data, the sinogram and linogram now appears uninterrupted and the perfusion deficit moves to the apex.

Since the myocardium is relatively thin in the diastolic phase, the limited spatial resolution of SPECT systems
can produce partial volume artefacts, where the measured signal is less than the true signal. Indeed, this artefact
is actually used for clinical interpretation of myocardial perfusion imaging. The diastolic phases are significantly
affected by the partial volume effect, while in the systolic phases the myocardial wall thickens, which, with the
reduced effect of partial volume, increases the myocardial count level to produce the measure of ‘wall thickening’.

Case

A 54 year old man who was being assessed for coronary artery disease had a stress myocardial imaging
study using 99mTc-MIBI and adenosine pharmacological stress. The resultant images from ungated data show well
defined myocardium (see Fig. 84(a)). In the gated study (see Fig. 84(b)), the end diastole myocardial wall thickness
is similar to that shown in the ungated data, as would be expected given that the heart is mostly in this phase. At end
systole, the wall thickens, reducing the effects of partial volume artefacts and giving a relative increase in uptake
signal which is used to derive the ‘wall thickening’ metric.

Guidance

Non-voluntary motion from the heart can lead to some motion blurring, although in practice, since the heart
is mostly in the end diastolic phase, a time averaged image mostly represents the heart in the ventricle filling phase.

75
FIG. 84. (a) Ungated studies showing a uptake of 99mTc-MIBI in the myocardium and (b) a gated study showing the end diastolic
phase and end systolic phase, together with the concept of ‘wall thickening’.

4.4.3. Postural and positional artefacts

Background

On some dedicated cardiac SPECT scanners, there is the ability to perform scans in an erect or semirecumbent
position instead of the standard supine position. In the absence of AC, using a different position for SPECT scanning
can lead to different attenuation effects.
In some centres, to tackle attenuation artefacts in SPECT only scans of the myocardium (see Case B below),
patients can be positioned prone rather than the more traditional supine position. Although uncomfortable for some,
the prone position can compact (e.g. breast tissue) leading to a less attenuated signal.

Case A

A 59  year old man presenting with breathlessness and showing coronary artery calcifications on a recent
chest CT was referred for a myocardial perfusion scan. Under adenosine pharmacological stress, 1052 MBq of
99m
Tc-MIBI was injected intravenously, with imaging performed in both a semirecumbent and supine position
using a dedicated solid state imaging system. No attenuation correction was performed.
Figure 85 shows differences in image appearance with the patient in a semirecumbent and supine position.
Due to gravity, diaphragmatic attenuation is less of an issue with the patient semirecumbent, which is highlighted
by better perfusion in the inferior wall when compared to supine imaging. However, spill-in of activity from
neighbouring gastrointestinal activity into the myocardium is also more problematic in this patient when
semirecumbent. Perfusion in the septum is reduced in semirecumbent imaging compared to supine imaging, which,
in this instance, is due to the preferable attenuation environment found in the supine position.

Case B

A SPECT only myocardial perfusion study was performed using 99mTc-MIBI on a male patient. In the supine
image shown at the top of Fig. 86, there is a clear inferior wall attenuation artefact. When the patient was imaged
prone (bottom), this artefact disappeared.

76
 99m
FIG. 85. Semirecumbent (upper row of each pair) and supine (lower row) stress Tc myocardial perfusion images. Yellow arrows
show postural differences in uptake in the inferior wall and septum.

Guidance

Posture related artefacts can be reduced by using transmission based AC or by using prone imaging where
possible. When this is not possible, the reporting doctor should be made aware of the patient’s posture during
scanning to take account of attenuation artefacts. Dual scanning in multiple positions may also be helpful in some
instances.

4.4.4. Incomplete injection

Background

During injection of a patient, a drop of injectate may be left in the injection line if not properly flushed. In
that case, image distortions of different types may appear, depending on the method of reconstruction and the
selection of image display parameters. Similar results may appear in the case of a (partial) subcutaneous injection.

Case

A body phantom (volume 9.7 L) was filled with a concentration of 99mTc of 50 MBq/L and ‘arms’ consisting
of 1 L infusion bags with the same activity concentration were placed along each side (see Fig. 87). A droplet of
10 MBq ‘injectate’ (2%) in a test tube was placed at the left arm to simulate the injection site. SPECT/CT was
performed on a Philips Precedence 16. Acquisition parameters of a low energy general purpose (LEGP) collimator,
128 angles, 128 × 128 matrix, 60 s per angle; and reconstruction parameters of (left) FBP (NAC), Hamming filter
with cut-off frequency at 0.5 Nyquist frequency and (right) iterative method (OSEM: PSF, 4 iterations 16 subsets,
CTAC).

77
FIG. 86. 99mTc-MIBI scan of a patient scanned in supine and prone positions. The inferior wall attenuation artefact seen in the supine
image is resolved when the patient is imaged in the prone position.

78
  
FIG. 87. Body phantom with ‘arms’ attached and a test tube placed between the ‘body’ and the ‘left arm’ with a hot droplet simulating
an injection site: (bottom left) FBP reconstruction (NAC); (bottom right) iterative reconstruction with CTAC. The grey scale in both
cases is logarithmic in order to represent the large dynamic range of the image contents.

Guidance

Injection problems may cause small spots of very high activity concentration. FBP reconstruction (NAC)
exhibits large artefacts, while the iterative method is able to handle the situation reasonably correctly. The dynamic
range may require that the images be shown in a logarithmic scale. Accepting an overflow at the ‘point source’ in a
linear scale can bring out the main information, but it can also result in the main part of the image being displayed
with very few levels of grey, resulting in a loss of information.

4.4.5. Radiopharmaceutical cross-contamination

Background

Gamma camera collimators are designed for a certain range of photon energies and are often referred to
as low, medium or high energy. If the upper limit is exceeded, the photons will penetrate the collimator septa to

79
an unacceptable degree, leading to a diffuse image blur or a streak pattern that will impair image resolution and
general quality. In SPECT, reconstruction of inconsistent projections may lead to further errors.

Case

A 50 year old woman was referred for a bone scintigraphy and SPECT and was injected with 545 MBq of
99m
Tc-HDP. An examination was performed on a Siemens Symbia  16 with low energy collimators. The images
showed an unexpected pattern (see Figs 88 and 89), and it was recognized that there was an apparent high activity
in the neck region (front). An investigation revealed that four weeks earlier in another hospital, the patient had been
given a 131I treatment with 400 MBq for a benign thyroid disease. The amount of 131I remaining was not measured
but estimates were about 1–2 MBq.

Guidance

It is important to be informed about a patient’s previous nuclear medicine examinations and radionuclide
treatments. Of particular importance are nuclides used in therapy with higher energy and long half-lives, such as

FIG. 88. Planar bone scintigraphy (99mTc-HDP) of a 50 year old woman who had received 131
I treatment with 400 MBq four weeks
earlier.

80
 FIG. 89. SPECT (99mTc-HDP) of 50 year old woman who had received 131I treatment four weeks earlier.

131
I or 177Lu, but also PET isotopes with an energy of 511 keV can even in small amounts seriously disturb SPECT
imaging at lower energies owing to collimator penetration.

4.4.6. Patient attenuation

Background

When SPECT data are acquired without CTAC, attenuation artefacts can be present in the reconstructed
image. For many SPECT studies, attenuation artefacts do not strongly affect clinical interpretation of the images.
However, for myocardial perfusion SPECT, the semiquantitative nature of the data, and the fact that stress and rest
images may be compared can lead to attenuation related artefacts causing problems in clinical interpretation.

Case A

A common attenuation artefact found in myocardial perfusion imaging is that caused by attenuation of
projection data by breast tissue — particularly in women with large breasts. In Fig. 90, short axis together with
horizontal and vertical long axis data from a myocardial perfusion study are shown highlighting breast attenuation
in the anterior wall (upper row). In the lower row, the breast tissue was strapped and compressed, resulting in a
reduction of the attenuation artefact.

Case B

Typically, when performing a myocardial perfusion study, patients are asked to put their arms above their
heads to avoid unwanted attenuation of the projection signal. For some patients, however, this position can be
difficult to achieve. For these patients, data are to be acquired with arms down. In Fig. 91(a), the sinogram in the
top row shows attenuation artefacts from an arm down by the patient’s side. In the bottom sinogram image, the
patient’s arm is moved above the head and there are no attenuation artefacts. In the reconstructed data in (b), the
upper rows show an apical–lateral artefact related to arm attenuation, which disappears once the arm is moved
above the patient’s head.

81
FIG. 90. Myocardial perfusion study with breast attenuation artefact (upper row), and with breast tissue strapped to reduce the
attenuation artefact (lower row).

Guidance

Attenuation artefacts can cause problems in uncorrected SPECT data when quantification, semiquantitative
or comparative data are required. The simple way to address these issues is to perform CTAC if it is available. If
this is not an option, all steps should be taken to minimize attenuation issues before SPECT data are acquired.

4.5. CT ISSUES

CT image quality may be limited by a number of factors in the scanner design or the choice of imaging
parameters. This section includes some examples that relate to the use of CT with SPECT.

82
FIG. 91. Myocardial perfusion study. (a) Top sinogram shows reduced signal at the lower part of the sinogram because of arm
attenuation. In the bottom row the intensity of the signal is maintained with the patients’ arms raised. (b) An apical–lateral artefact
seen with arms down (upper row) disappears once the patients’ arms are raised (lower row).

4.5.1. CT slice thickness

Background

CT scanners can acquire data in helical or axial mode. In helical mode, there is a high degree of flexibility
in slice thickness because of the interpolation of helical data into transaxial slice bins. However, in axial
mode the transaxial slice thickness is set by the configuration of CT  detectors. In modern multislice scanners,
slice thickness can vary from tenths of a millimetre to several millimetres. In some older systems, particularly
SPECT/CT systems, the transaxial slice thicknesses can be limited to 5–10 mm. As a consequence, although data
within a transaxial slice demonstrate good spatial resolution, in coronal and sagittal planes the spatial resolution is
severely compromised. This can cause problems when measuring feature dimensions across reconstructed slices.

Case

An abdominal CT from an early SPECT/CT system with a 1 cm slice thickness is shown in Fig. 92. While
spatial resolution within the transaxial plane is adequate, the 1 cm slice thickness creates partial volume effects in
coronal and sagittal views, which limits their usefulness.

Guidance

On modern multislice SPECT/CT systems, the issue of large slice thicknesses has virtually disappeared. If
such limits are applicable in a SPECT/CT system, the use of helical mode should be considered.

83
FIG. 92. Transaxial, coronal and sagittal slices through the abdomen using a 1 cm slice thickness. The large slice thickness results in
step like features in sagittal and coronal views.

4.5.2. CT helical pitch

Background

The use of helical CT is very common in SPECT/CT. This is primarily due to dose savings when high levels
of pitch are used. While this may be acceptable in areas where there are no rapid changes in the morphology of
anatomy, in some instances the use of a high pitch CT may break the assumptions required to interpolate missing
data used for reconstruction.

Case

A high pitch (1.375:1) helical CT of a patient’s arm was acquired (see Fig. 93). Due to the oblique position of
the arm, the interpolation used on the helical CT reconstruction produces a wave pattern in the bony structures of
the arm.

Guidance

Wave like artefacts in CT images can be avoided by placing anatomical structures in a straight position. If an
oblique position cannot be avoided, a lower pitch CT should be considered together with a reduced tube current to
maintain a similar patient dose.

4.5.3. Noise

Background

The choice of CT  image quality (noise content) could in principle affect the SPECT image quality and
quantification when the CT image is used for the AC of the SPECT data.

Case

Figure 94 shows CT and corresponding SPECT scans of an anthropomorphic thorax phantom filled with
99m
Tc. CT scans with three different dose levels (CTDIvol: 21, 10, 0.5 mGy) were acquired and then used to correct
the SPECT data for attenuation. The results show that with decreasing CT dose level, the CT images become noisier

84
FIG. 93. Coronal slice of an obliquely positioned arm acquired with high pitch helical CT produces wave like patterns in the structures
of the arm.

FIG. 94. The effect of CT noise on SPECT image quality of an anthropomorphic thorax phantom filled with 99mTc. The top row shows
CT scans of increasing noise content (left to right). The second row shows the corresponding SPECT scan when reconstructed with
these different CTs. The third and fourth rows show CT and SPECT subtraction images, respectively, to highlight the effect of CT
noise on the resultant images. Even with large decreases in CT dose levels, the effect on the CT HU were less than 50 HU and the
corresponding quantitative difference on the SPECT images were less than 3% (red and yellow lines).

85
and the corresponding HU changes. This change, however, has minimal impact on the SPECT image quality and
quantification, as seen in Fig. 94. It is important to note that a large change in CT dose (21 to 0.5 mGy: 2% of the
original dose) resulted in a change of only less than 3% in the SPECT image quantification. The reason for this
small change is primarily due to image smoothing which is applied to the CT images before they are used for AC of
SPECT data.

Guidance

A decrease in CT dose results in noisier CT images but has very little impact on the SPECT image quality and
quantification.

4.5.4. Slow rotation time CT

Background

In some SPECT/CT systems, CT data are acquired with rotation speeds of several seconds. This can cause
problems with patient motion being captured during the tube rotation, which results in inconsistent projection data,
for example the anatomical distribution in the patient changes as the projection data are being acquired. A common
artefact from this effect is seen in the abdomen, where gas in the gastrointestinal tract moves during tube rotation.

Case

Figure 95(A) illustrates an axial abdominal CT acquired on a slow rotation CT system. Involuntary patient
movement during the scan has led to inconsistent projections, some of which originates from moving pockets of
gas within the gastrointestinal tract. As a result, ‘streak’ type artefacts are seen in the transaxial data. Fortunately,
when used for AC, the CT data are smoothed (B) to reduce the level of noise translated into the SPECT data so the
streaks are not seen in the reconstructed emission data.

Guidance

In slow rotation CT systems, these artefacts are difficult to avoid. Faster rotation CT scanners are unlikely to
create this type of artefact.

FIG. 95. (A) Transaxial CT data from a slow rotation CT scan of the abdomen and (B) data are smoothed for AC and loses the streak
artefacts.

86
4.6. SPECT/CT ARTEFACTS

The introduction of SPECT/CT imaging has brought several advances to SPECT imaging, such as anatomical
localization, AC and more recently quantification of SPECT data. However, the advent of this technology has also
introduced several artefacts primarily arising from the use of CT images for AC of SPECT data. Examples of such
artefacts include CT beam hardening due to high density material in the CT X ray beam such as metal objects or
bone, CT contrast media, truncation, and misregistration between the CT and SPECT data.

4.6.1. Beam hardening artefacts

4.6.1.1. External metal objects

Background

As described in Section 2.3.2, metal objects will often create artefacts in CT owing to beam hardening and
excess loss of photons (photon starvation). They will also erroneously reduce counts in the emission projection
data, and these errors can propagate into the final SPECT images depending on the reconstruction algorithm’s use
of CTAC. In this section, such artefacts are demonstrated based on a phantom experiment.

Case

The experiment was performed on a Philips Precedence 16, with a NEMA image quality phantom with
approximately 400 MBq of 99mTc in the background (40 kBq/cm3), spheres and lung insert were left empty for
simplicity. Several high density objects were used and they are listed in Table 3.

TABLE 3. METAL OBJECTS IMAGED

Object Mass (g) Material composition

(a) 2 euro coin 8.5 75% Cu, Zn, Ni

(b) Belt buckle —* —*

(c) Euro cent coins 14.6 Fe, Cu

(d) Gold earring 0.42 59% Au (Ag)

(e) Silver earring 0.28 Ag

(f) Gold ring 3.7 59% Au (Ag)

(g) Silver ring 5.0 Ag

*
—: data not available.

87
The objects were attached to the sides of the phantom at different locations (see Fig. 96). A preview scan was
performed at 120 kV followed by two CT scans, one at the lower and one at the upper bounds for the system X ray
tube energy (90 kV (300 mAs) and 140 kV (200 mAs)). SPECT data were then acquired with 120 projections, 20 s
per projection, in a 128 matrix. Images were reconstructed first with NAC and then using the two CT scans for AC
including scatter correction and resolution recovery (Astonish kernel). No specific corrections were performed for
the presence of metal. High CT values were clipped at 3000 HU.

FIG. 96. Preview scan of the phantom with seven metal objects attached, all easily detectable. The finger rings (f) and (g) were
mounted on water filled syringe ‘fingers’. The remaining objects were taped directly to the surface of the phantom.

Figures 97–103 show a single slice of the NEMA image quality phantom at the location of each of the high
density objects from Fig. 96. In each figure, the top row shows the NAC image. The other two rows display the
AC SPECT images with their respective CTs at two different tube voltages: low (middle row) and high (bottom
row).

88
 FIG. 97. A 2 euro coin on the surface of a 99mTc filled phantom.

In Fig. 97, the NAC image shows a lack of counts at the position beneath the coin. The metal is clearly visible
on the two CT scans with some streak artefacts, more prominent at 90 kV than 140 kV. The AC images on the right
displays a hot spot from overcorrection of the attenuation, more pronounced at 90 kV than at 140 kV (see also
Fig. 13, Section 2.3.1, and Table 1).

89
 FIG. 98. A belt buckle on the surface of a 99mTc filled phantom.

In Fig. 98, the NAC image looks rather unaffected by the presence of the belt, presumable made of rather light
materials. The metal parts, however, are clearly visible on the two CT scans with weak artefacts, more prominent at
90 kV than 140 kV. The AC images on the right do not show clear indications of artefacts.

90
 FIG. 99. Euro cent coins on the surface of a 99mTc filled phantom.

In Fig. 99, the NAC image shows a significant lack of counts at the position beneath the coins. The metal
is clearly visible on the two CT scans with strong streak artefacts, more prominent at 90 kV than 140 kV. The AC
images to the right still displays a count deficit, not entirely corrected by the CTAC.

91
 FIG. 100. Gold earring on the surface of a 99mTc filled phantom.

The tiny object of a gold earring is barely visibly in the NAC image in Fig. 100. The metal creates local streak
artefacts on the two CT scans with little or no difference between 90 kV and 140 kV, which may be explained by
the abrupt change in absorption properties of gold changing at the K edge (80 keV) (see also Fig. 13 and Table 1).
In the AC images on the right, no artefact is visible.

92
 FIG. 101. Silver earring on the surface of a 99mTc filled phantom.

Similarly, with the silver earring in Fig. 101, the tiny object is barely visibly on the NAC image. The metal
creates weak local streak artefacts on the two CT slightly worse with 90 kV than with 140 kV, and an artefact is not
visible in the AC images on the right.

93
 FIG. 102. Gold ring on the surface of a 99mTc filled phantom.

The gold ring creates a surface ‘hole’ in the NAC image in Fig. 102. The metal creates strong streak artefacts
on the two CT scans with little difference between 90 kV and 140 kV which, again, may be understood from the
absorption properties of gold (see Fig. 13). The count loss is partially corrected by the CTAC images to the right.

94
 FIG. 103. Silver ring on the surface of a 99mTc filled phantom.

The silver ring is hardly visible in the NAC image in Fig. 103. Yet, the metal creates strong streak artefacts
on the two CT scans, more evident at 90 kV than at 140 kV. The artefacts do not propagate into the CTAC images
in a visible manner.

95
Guidance

The preview scan should be inspected for any kind of metal objects to be removed if possible, even if quite
small. The effect on both CT as well as the final corrected SPECT images is difficult to predict, since it depends
on the size, material composition and geometry of the object as well as the CT energy chosen and the activity
distribution (not included in this simple simulation).

4.6.1.2. Prostheses

Background

In some patients, it is impossible to avoid CT beam hardening effects when performing SPECT/CT studies.
Indeed, in some orthopaedic referrals, it may be the prosthesis itself that is the reason for the SPECT/CT study. The
presence of prostheses can affect both CT and SPECT images.

Case A

Imaging of bone surrounding prostheses is a common referral for SPECT/CT. In Fig. 104, the upper and
lower parts of a knee prosthesis in the right knee (left side) can be seen. In Fig. 105(A), a transaxial slice through
the lower part of the leg shows radial streak artefacts coming from the very dense centre of the prosthesis, which
extend into the opposite leg — a known metal beam hardening artefact. Figure 105(B) shows the bone SPECT
image from the same patient, together with the count profile from NAC reconstructed data (blue) and the ratio of
CTAC to NAC counts in red.

FIG. 104. Knee prosthesis in situ in an X ray preview scan.

96
FIG. 105. (A) CT and fused bone SPECT/CT images through the lower part of the knee prosthesis and (B) bone SPECT profile
showing the NAC counts (blue) and the ratio of CTAC to NAC counts (red).

The two central red peaks and the smaller, right most peak corresponding to the outside of the leg represent
some misregistration between the emission and CT for AC — possibly due to patient movement. The dip between
these peaks is a consequence of there being no counts between the legs in the uncorrected image, leading to a
void in the plot. The centre of the knee prosthesis displays a clear reduction in counts because of the lack of bone
tissue, but the AC attempts to correct for the high attenuation in this area. Whether this correction is adequate for
a material that is very dense and outside the typical HU range is unknown, but it is unlikely to be under corrected
for attenuation. The smoothness of the profile through the rest of the knee suggests that the streak beam hardening
artefacts have not significantly degraded the AC.

Case B

A patient who had undergone spinal surgery was referred for a bone SPECT/CT scan to better understand
the source of the patient’s pain. A CT preview scan of the patient highlights two thin spinal prosthesis spanning
from the upper thoracic to upper lumbar vertebrae (see Fig. 106). In Fig. 107(a), a metal artefact can be seen with
dark streaks (green arrows) occurring along the lines of greatest attenuation and emanating from the two spinal
prostheses. The reconstructed emission data in Fig. 107(b) together with the profile of NAC counts (blue) and
ratio of CTAC to uncorrected counts show that overcorrection of counts in the small of the back (the lower curved
part of the back) can again be seen because of movement causing misalignment between emission and CT scans.
Furthermore, correction over the spinal prosthesis shows a step reflecting the correction for the dense prosthesis
material.

Guidance

When imaging areas that contain metal prosthesis, there is an effect on CT image quality. As a consequence,
there can also be an effect on SPECT image appearance and quantitative accuracy. However, the use of heavy

97
 FIG. 106. X ray preview scan of the spine with spinal prosthesis in situ.

FIG. 107. (a) CT and fused bone SPECT/CT images through the spine and prosthesis and (b) bone SPECT profile showing the NAC
counts (blue) and the ratio of CTAC to NAC counts (red).

98
smoothing kernels during the CTAC process can remove most small metal artefacts, and the use of extended
HU scales can minimize the under correction for attenuation of SPECT.

4.6.1.3. Arms down versus arms up

Background

Whenever possible, patients undergoing a SPECT examination of the thorax or abdomen should be positioned
with their arms above the head to minimize attenuation. The attenuation of one arm may well constitute a half value
layer at the photon energies applied in both CT and nuclear medicine. Excess attenuation can result in photopenia
and beam hardening artefacts in CT that may influence the resulting SPECT images.

Case

The effect is here exemplified in a phantom study using a body phantom with 1 L infusion bags attached to
simulate the arms (see Fig. 108), and 99mTc was added to the phantom and the infusion bags to a concentration of
approximately 40 MBq/L (total of 500 MBq). The phantom was scanned on a Siemens Symbia SPECT/CT and
a selection of CT  parameters was applied to simulate low dose as well as diagnostic scans (see Figs 109–112).
SPECT was performed over an hour to improve statistics and thereby also improve the visibility of potential effects
implied by the CTAC. Reconstruction was made with Flash3D (8 iterations, 4 subsets), including scatter correction.

Guidance

For SPECT image quality, it is generally better if the patient can keep the arms above the head for the duration
of the examination, avoiding the excess attenuation from the arms. If this is not possible in a SPECT/CT study, the
CT settings should be sufficient to provide the information requested.

FIG. 108. Body phantom with ‘arms’ attached.

99
FIG. 109. Preview of the body phantom without and with arms attached in the form of infusion bags (top row); NAC reconstruction of
phantom without and with attachments (bottom row).

FIG. 110. CT scan of 80 kV and 10 mAs, significant beam hardening artefacts are observed in the CT image when the arms are down
along the body. The influence in the SPECT image is visible.

100
FIG. 111. CT scan of 80 kV and 100 mAs, artefacts are still visible in the CT images with the arms down. Owing to the soft filter
applied to CT before the calculation of CTAC, the streaks do not propagate into the SPECT image, but quantitation is affected.

FIG. 112. CT scan of 130 kV and 170 mAs only a faint shadow is observed in the CT image, even with the arms down, and the
influence on SPECT is negligible.

101
4.6.2. Contrast CT: Negative contrast

Background

In SPECT/CT imaging, negative contrast such as gas pockets can cause image artefacts owing to the transition
of the gas pocket in the gastrointestinal tract between the time of the CT and SPECT data acquisition (see Fig. 95,
Section 4.5.4).

Case

In Fig. 113, a CT scan (A) and corresponding attenuation map (B) illustrate an example of an air pocket
present in the CT scan which is not present in the SPECT acquisition (C) due to the transition of the air pocket

FIG. 113. (A) Transaxial CT data from a slow rotation CT scan of the abdomen showing an air pocket in the gastrointestinal tract and
(B) corresponding CT attenuation map clearly showing the air pocket. (C) The AC SPECT slice clearly shows the air pocket indicating
that the erroneous presence of the air pocket in the image is due to the AC process. (D) The corresponding uncorrected SPECT slice
shows the absence of the air pocket indicating that the air pocket had transitioned between the acquisition of the CT and SPECT data.

102
in the abdomen/gastrointestinal tract between the CT and SPECT data acquisition times. As a consequence, the
AC data (D) show an area of reduced counts (photopenic area) at the position of the gas pocket. This reduction in
counts is not seen in the uncorrected data (C), indicating that the effect was introduced by AC.

Guidance

The transition of air pockets between CT and SPECT acquisitions can lead to artefacts that manifest as
photopenic areas in the reconstructed SPECT images. Unfortunately, such artefacts are unavoidable in SPECT/CT
imaging.

4.6.3. Truncation

Background

On SPECT/CT systems, the CT transaxial FOV is typically limited to a maximum diameter of 50 cm because
of the geometry of the CT tube and detector array used in CT scanners. On some occasions, the CT scan might be
truncated (i.e. part of the patient may be outside the CT FOV). This can occur with patients with large body habitus
with diameters greater than 50 cm, or when patients are mispositioned in the CT scanner so that parts of the body
fall out of the FOV.

Case A

In the bone SPECT/CT study of the neck and shoulder area shown in Fig. 114(A), it is clear that there is
truncation of CT data on both shoulders (red arrows). This truncation is due to the position of the patient being
too low in the CT FOV. A transaxial slice of SPECT data through the same area is also shown (B), with the count
profile from NAC reconstructed data (blue) and the ratio of CTAC to NAC counts in red. In the areas of truncation,
there is clear underestimation of counts — best seen in the AC transaxial image shown and the sharp drop at the

FIG. 114. (A) A CT and fused SPECT/CT image of the neck and shoulder area showing truncation in both shoulders on the CT data,
and (B) bone SPECT profile showing the NAC counts (blue) and the ratio of CTAC to NAC counts (red).

103
very edge of the AC/NAC ratio on the right shoulder (left side). There is also a small amount of motion in the left
shoulder (right side), which is shown as a peak on the other side of the AC/NAC ratio. In the space between the
neck and the left shoulder, the peaks and sharp drop in the AC/NAC ratio are a consequence of the lack of counts in
this area with NAC (zero), which if counts were present would match the profile seen on the opposite side because
of the bone present in this area.

Case B

A coronal view of an 111In labelled monoclonal antibody radiotracer SPECT/CT scan of a patient showing
truncation along both arms of the patient is shown in Fig. 115. The CT truncation results in large CT numbers
adjacent to the truncated anatomy (red arrows), which in turn also results in an artificial boosting of the radioactivity
uptake in these areas (white arrows).

Guidance

CT truncation has an effect on AC emission data and should be avoided wherever possible by positioning
patients centrally in the FOV. Some systems offer the ability of extended CT FOV reconstruction, which uses
non-truncated projections to estimate the signal outside the standard FOV [51]. Iterative CT reconstruction may
also offer better solutions with truncated CT data than FBP. Nevertheless, these CT reconstructions are estimates
and truncation should be avoided wherever possible.

FIG. 115. A coronal view of an 111In labelled monoclonal antibody radiotracer SPECT/CT scan of a patient showing truncation along
both arms of the patient.

104
4.6.4. Misregistration

4.6.4.1. Cardiac

Background

In many centres, when performing myocardial perfusion SPECT imaging, it is common to acquire a CT in the
same examination to correct for photon attenuation. This can be helpful to deal with attenuation from breast tissue
in women or diaphragmatic attenuation in men, both of which can lead to the false positive reporting of myocardial
ischemia or infarction. However, since SPECT and CT are performed sequentially on SPECT/CT systems, there
is the possibility for mismatch in the emission and AC data. This is particularly problematic in the thoracic region
where the heart is located. Breathing can easily lead to the misregistration of data in this region because the SPECT
images are an average representation of uptake through the respiratory cycle, while the CT is a snapshot of anatomy
at a particular respiratory phase. Misregistration of SPECT and CT data can induce artefactual changes in the
displayed myocardial blood flow polar map.

Case A

A 66 year old women with previous myocardial infarctions and stents was referred for a myocardial perfusion
study after complaining of dizziness and chest pain. A pharmacological stress myocardial perfusion SPECT/CT
imaging study was performed after the injection of 866 MBq of 99mTc-MIBI.
Figure 116(A) shows the position of the CT in relation to the SPECT data, with clear misregistration in the
apex of the heart. This leads to a mild apical defect (B). When the misregistration is corrected (C), the appropriate
AC is applied, and the severity and extent of this apical defect changes slightly.

Guidance

In cardiac SPECT/CT imaging, misregistration of emission and AC data from patient breathing is possible.
There is no easy method of reducing the likelihood of this artefact. Software postprocessing of the data to correct
for misregistration should be performed before reporting to ensure that the images can be interpreted correctly,
although any significant difference in the shape of the liver or spleen cannot be corrected for.

FIG. 116. Stress myocardial perfusion SPECT/CT study showing (A) misregistration of SPECT and CT data that results in (B) a small
apical defect. (C) Once the misregistration is corrected, the severity and extent of the apical defect changes.

105
FIG. 117. (A) Myocardial perfusion SPECT/CT study showing misregistration between the emission and CT for attenuation scan and
(B) the resulting AC images and polar maps from this misregistration (IRACRR) together with NAC (IRNCRR).

Case B

In Fig. 117(A), the misregistration of emission data with CT data can be seen, showing the apex of the
heart on the emission data encroaching beyond the chest wall, and the lateral wall of the myocardium on SPECT
positioned in the lung on the CT. The consequences of this misregistration are shown in Fig. 117(B), where slice
and polar plot AC (IRACRR) data, and NAC (IRNCRR) data are shown.

106
FIG. 118. (A) Myocardial perfusion SPECT/CT study showing good registration between the emission and CT for attenuation scan
and (B) the resulting AC (IRACRR) and NAC (IRNCRR) reconstructed images and polar maps from this registration.

Case C

In Fig. 118, the same patient study as Case B is shown with good registration between emission and CT for
AC scans. Comparing the two data (good and bad registration), the difference in counts in the polar plots with and
without AC shows that misregistration in the apex leads to a relatively small change in the accuracy of the AC
because of the small difference in the attenuation of myocardium and soft tissue beyond the chest wall. However,

107
the lateral wall misregistration shows a bigger difference (most clearly seen in the polar plots) owing to the large
difference in attenuation between soft tissue and lung.

Guidance

Misregistration of CT and SPECT has a detrimental effect on the accuracy of AC used for myocardial
perfusion SPECT/CT studies. Most system manufacturers provide software to correct for any misregistration. Use
of this software avoids any potential misinterpretation of clinical myocardial perfusion studies.

4.6.4.2. Respiratory motion

Background

Mismatch between transmission and emission data is relatively common in SPECT/CT imaging of the
thoracic region. With a breathing cycle typically taking around 6 s, a 15–20 s SPECT projection will capture an
average image over the time period. However, a CT over the same area can take just over a second, and therefore
captures the respiratory cycle at a particular phase. During a typical breathing cycle, most time is spent in the
end expiration phase, which means that misregistration between SPECT and CT is often avoided or is small in
magnitude. However, if CT captures an inspiration phase misregistration will be seen — typically in the base of the
lung or in the dome of the liver where motion is largest.
In addition to misregistration, there can also be additional artefacts when CTAC is performed. These present
as an underestimation of the signal in the liver, and can also lead to an underestimation in the uptake in the inferior
wall of the heart in myocardial perfusion studies.

Case

A 65 year old man with metastatic neuroendocrine cancer was sent for post-therapy scanning following a
therapeutic 7.7 GBq infusion of 177Lu-DOTATATE. In the SPECT/CT scan acquired 24 hours post-infusion (see
Fig. 119), there is mismatch between the CT scan used for localization and AC, and the SPECT data. The likely
cause of this misregistration is the acquisition of CT data in an inspiration phase. The dome of the liver from
SPECT  (D) extends into what is seen as lung tissue on the CT  (A). As a consequence, the area of mismatch is
undercorrected (B and C) because the reconstruction algorithm is correcting for lung attenuation when it should
be correcting for liver tissue attenuation. Figure 119(D), which does not use AC, provides a more accurate
representation of the activity distribution.

Guidance

Mismatch of SPECT and CT data in the thoracic region is difficult to avoid, given that it is difficult to acquire
the CT in the same phase as the time averaged SPECT. Although some have advocated the use of breath hold end
expiration CT to avoid this issue, the best results have been found when allowing free breathing. Steps to minimize
misregistration include warning the patient to avoid deep breathing during the scan, and allowing the patient to
rest for a short period on the bed before starting the CT scan. This final technique allows the patient to settle into a
regular resting breathing cycle before the CT starts.

4.6.4.3. Bladder

Background

Mismatch between CT and SPECT can occur in the pelvic region owing to the time difference between the
two acquisitions. During this time, the bladder may undergo significant filling, which will displace surrounding
tissue and potentially lesions. The probability of large changes increases with time, and therefore multiple position
SPECT images are more likely to show such an effect if performed with the head first (assuming that CT is
performed before SPECT).

108
FIG. 119. 177Lu-DOTATATE SPECT/CT study showing (A) the CT used for localization and AC; (B) AC SPECT; (C) fused CT and AC
SPECT; and (D) uncorrected SPECT data.

Case

A 73 year old man underwent a whole body planar bone scintigraphy with 99mTc followed by a SPECT/
CT extending over three camera FOV, each lasting 8 min. The SPECT scan was started from the head position.
The time difference between the CT and the SPECT of the bladder (midpoint time) was approximately 25 min.
Figure 120 shows a considerable change in bladder size between the two modalities. In this case, there were no
lesions in the vicinity of the bladder, and the displacement is therefore without clinical significance. If a lesion is
displaced, it is important to remember that its anatomical correlate on the CT is not correctly placed.

Guidance

Looking for lesions in the pelvis, it is important to notice potential differences in bladder filling that may
displace lesions between CT and SPECT. In order to shorten the time between the scans, a composite SPECT scan
should be started from the pelvis upwards (assuming that CT is performed before SPECT).

4.6.4.4. Patient motion

Background

When using CT for AC, it is important that the SPECT and CT scans be in alignment so that appropriate
correction factors are applied. While misregistration can occur anywhere in the body, movement of limbs during
the scan can be particularly problematic.

109
FIG. 120. The time difference between CT and SPECT may cause the bladder to present with different degree of filling (arrows). This
may potentially displace lesions, causing a mismatch.

Case

Figure 121 shows a bone SPECT/CT study of the pelvis with motion of the pelvic area. In addition, movement
of the right hand (left side) can be seen in the emission data but not the CT data. There is also CT truncation of the
left hand (right side) and right pelvis.

110
 In Fig. 122(A), a transaxial slice of SPECT data and profile through the hand are shown, with the count
profile from NAC reconstructed data (blue) and the ratio of CTAC to NAC counts in red. The ratio of counts in
the right hand (left side) shows an AC/NAC ratio close to 1.0, which confirms that counts have not been corrected
for attenuation. In the left hand (right side), there has been a very small correction for attenuation with the hand
partially in the CT FOV. There are also several large spikes seen in the AC/NAC profile, which is indicative of
motion or misregistration of the emission data and CT used for AC.
In Fig. 122(B), a similar profile through the anterior aspect of the pelvis is dominated by patient motion
leading to misregistration between the emission and CT for AC data. This can be seen by the red peaks at either
side of the body. The misregistration of the most anterior aspects of the pelvic bones with the corresponding areas
on the CT have led to an under correction of counts in this area, which can be identified by the similar correction of
pixels in soft tissue areas neighbouring the bone.

FIG. 121. CT and fused bone SPECT/CT image of the pelvis area showing motion of the pelvis and hand, and truncation of the
opposite hand and slight truncation of the right pelvis (left side).

FIG. 122. Bone SPECT profiles showing the NAC counts (blue) and the ratio of CTAC to NAC counts (red) (A) through the hands and
(B) through the pelvis.

111
Guidance

Misregistration of CT and SPECT clearly has an effect on CTAC of counts. Good patient instructions and
strapping and light immobilization can help to avoid these problems in many instances.

4.7. QUANTITATIVE ISSUES

The addition of CTAC to SPECT imaging along with a host of other corrections, such as scatter, dead time and
efficiency for various radionuclides and collimators, has introduced for the first time a reliable approach to generate
quantitative SPECT data. Today, several commercial SPECT/CT systems have optional quantitative modules that
allow the user to quantify SPECT images in terms of standardized uptake value (SUV) similar to what is used in
PET imaging. The accuracy of quantification, however, depends on several factors, such as the reconstruction
algorithm parameters (number of iterations and subsets), the accuracy of data corrections (i.e. attenuation and
scatter), the presence of contrast or metal, and patient motion (voluntary or involuntary) that leads to mismatch
between CT and SPECT data.

4.7.1. Iterations and subsets

Background

SPECT iterative reconstruction creates a tomographic image by making many attempts (iterations) at
matching an estimated image to the measured projections updating estimates typically using MLEM. One method
of accelerating this process is to use OSEM, in which each update to the image is based only on angularly spaced
subset of the projections rather than comparing all projections. The product of subsets and iterations in OSEM can
be considered to be equivalent to the number of full expectation maximization realizations.
This process provides at each iteration an improved estimate of the image, and is typically stopped when an
acceptable image has been formed. Quantitatively, the uptake in this image may not fully have reached convergence.
Furthermore, convergence for different areas in the image may be reached after different numbers of iterations.

Case

A 79 year old man was referred for a 123I-ioflupane study to assess whether he had idiopathic Parkinson’s
disease or essential tremor. Following the injection of 178 MBq of 123I-ioflupane, the patient returned for a
SPECT/CT scan 3.5 hours later. Figure 123 shows a transaxial slice through the basal ganglia following 1, 2, 5, 10,
15 and 20 iterations of 10 subsets effectively equivalent to 10, 20, 50, 100, 150 and 200 full MLEM iterations. It can
be clearly seen how the image develops visually following multiple iterations. Figure 124(A) shows quantitatively
how the uptake in the left striatum changes with iteration number as a percentage of the value measured using
FBP — a method that does not have convergence issues. Not until five iterations does convergence occur in this
region. However, in the background reference region, which is used to define the commonly used specific binding
ratio (SBR), convergence has already occurred after two iterations.
On closer inspection, the values of uptake in the background region using iterative reconstruction are actually
higher than that defined by FBP. This is a known problem with MLEM and OSEM called the non-negativity
constraint bias and occurs in areas of low uptake. During the MLEM and OSEM reconstruction process estimates
are not allowed to be negative, which if uptake is close to zero has the consequence of pushing the distribution on
which the estimate is made upwards producing an overly high final value.

Guidance

When performing quantitative SPECT, it is essential that sufficient iterations and subsets be used to ensure
convergence of uptake values. In areas of low uptake, the non-negativity constraint of MLEM and OSEM
algorithms can lead to an overestimation of the true value.

112
FIG. 123. 123I-ioflupane image of the basal ganglia reconstructed using 1, 2, 5, 10, 15 and 20 iterations of 10 subsets and the
FBP image.

FIG. 124. (A) Striatal uptake and (B) background uptake represented as a fraction of the uptake calculated using FBP.

4.7.2. Corrections

Background

Unlike in PET, where many corrections are applied by default, in SPECT, corrections for physical effects
might not be applied, owing to either complexity or unavailability of required data or software licences. However,
when performing quantification, all corrections should be applied to achieve more accurate values of uptake.
Modern systems have quantification modules that apply all these corrections automatically.

Case

As in Section 4.7.1, a 79 year old man was referred for a 123I-ioflupane study to assess whether he had
idiopathic Parkinson’s disease or essential tremor. Following the injection of 178 MBq of 123I-ioflupane the patient

113
returned for a SPECT/CT scan 3.5 hours later. Figure 125 shows a transaxial slice through the basal ganglia
following OSEM reconstruction with 10 iterations and 10 subsets using:

(a) No corrections;
(b) Correction for scatter using a triple energy window scatter correction;
(c) Correction for attenuation;
(d) Correction for attenuation and scatter using a triple energy window scatter correction;
(e) Correction for attenuation, scatter and resolution losses.

Figure 126 shows the caudate count density and caudate specific binding ratio to an occipital lobe reference
region following each correction:

striata count density − reference count density

SBR =
reference count density

What is clear from both A and B is that the scatter correction makes a significant difference to the contrast
between the striata and the background. Although the counts in the striata are reduced using this triple energy
window method, the counts in the background are also reduced, leading to an increased caudate specific binding
ratio. This benefit arises because of removal of the down scatter and septal penetration of higher energy emissions
from 123I. A similar measurement with 99mTc would not have shown as much of a benefit, but many other
radionuclides used in nuclear medicine have several gamma emissions and therefore do significantly benefit from

FIG. 125. Transaxial slice through the basal ganglia when reconstructed (a) without corrections, (b) with a correction for scatter,
(c) with AC, (d) with correction for attenuation and scatter and (e) with correction for attenuation, scatter and resolution losses.

FIG. 126. Charts showing caudate count density and caudate specific binding ratio (to an occipital lobe reference region) with
different reconstruction strategies.

114
this correction. AC makes an additional increase to counts and uptake ratios (dependent on the position of the
reference region). Owing to the size of the organ, resolution modelling makes a large difference to the caudate
count density and also increases specific binding ratios.

Guidance

The values derived from quantitative SPECT will depend on the corrections applied. To yield the most
accurate values, corrections for all physical effects should generally be used, although in some instances this might
lead to greater uncertainty in the measure. Where absolute quantification is required, a sensitivity calibration is
necessary to relate count densities in the image to activities per unit volume.

4.7.3. Iodine contrast in 99mTc- and 111In-SPECT

Background

When iodine contrast is applied in a patient, the number of projection counts slightly decreases owing to
the excess attenuation of the iodine. This effect depends on the gamma energy and is larger for 99mTc than for
111
In (see Fig. 127). However, CTAC generally increases the quantitative values of the final reconstructed SPECT
image, as explained in Section 2.3.2. Two situations of interest should be considered. A high and visible local
concentration of contrast in the blood vessels at the time of CT (which is one reason for applying contrast media)
may, at least in principle, lead to local SPECT artefacts by reconstruction, although the low SPECT resolution and
the filtering of attenuation maps can tend to hide them. A more distributed concentration at later times (e.g. repeated
CT scans) will slightly affect overall quantification. The latter situation is shown here.

Case

Phantom studies were performed to illustrate the effects (exaggerated). A 20 cm cylinder phantom (5.15 L)
was filled with water and activity was added (two independent studies with 50  MBq of 111In and 300  MBq of
99m
Tc). Measurements were performed with pure water, and with iodine contrast added (100 mL and 200 mL of
ioversol, 350 mg/mL), leading to iodine concentrations in the phantom of 6.8 g/L and 13.5 g/L, respectively. The
total amount of contrast agent given to a patient would normally be in the order of 100  mL (i.e. <50  g), and

 
FIG. 127. Relative detected raw counts (in %) from 99mTc and 111In as a function of increasing iodine concentration in a 20 cm cylinder
phantom. Likely patient concentrations are in the shaded area to the left.

115
therefore the average concentration over extended volumes during SPECT would not exceed a few g/L (shaded
area in Fig. 127). SPECT acquisitions were performed on a Philips Precedence 16.
Images were reconstructed without AC (NAC) and with attenuation maps derived from CT scans at
peak voltages of 90, 120 and 140 kV. When an attenuation map for SPECT is created from the CT, the HU are
typically converted using a bilinear mapping (see Section 2.3.2). The presence of CT contrast will tend to create
an overestimation of the SPECT attenuation, and the size of the effect will depend on CT energy, actual SPECT
energy as well as the local concentration of iodine (see Figs 128 and 129). The effect on 99mTc (140 keV) is smaller
than that for 111In (average of 171 keV and 245 keV). Numerical values will depend on the actual implementation
that may differ between vendors.

Guidance

The quantitative effect of distributed iodine contrast during CT scans used for AC of SPECT is rather limited
and does not exclude the use of contrast scans for this purpose.

FIG. 128. Relative change of reconstructed 99mTc-SPECT values as a function of the concentration of iodine for NAC and for three
different CT energies. Likely patient concentrations are in the shaded area to the left.

FIG. 129. Relative change of reconstructed 111 In-SPECT values as a function of the concentration of iodine for NAC and for three
different CT energies. Likely patient concentrations are in the shaded area to the left.

116
4.7.4. Local effects of iodine contrast in 99mTc-SPECT

Background

Iodine contrast is applied in CT examinations with the purpose of enhancing structures like blood vessels.
The timing of the CT scan relative to injection determines the distribution (e.g. arterial or venous). Within a few
minutes, the contrast will be distributed in a large volume of body fluids, and therefore have a relatively low
concentration. If a CT scan with applied contrast is followed by a SPECT scan, the influence of contrast on the
SPECT scan itself is very limited. However, if the contrast enhanced CT scan is used to create an attenuation map
for SPECT, the AC may, in principle, create artificially high activity concentrations in the area where contrast is
present. The size of the effect depends on the contrast concentration, the CT energy applied and the SPECT energy
(see Section 2.3.2). The problem is much smaller for SPECT than for PET due to: (i) the lower energy of SPECT
(closer to CT); and (ii) the limited resolution where the CT based attenuation maps are filtered by a smooth filter to
match the resolution of SPECT.

Case A

A phantom study was performed to illustrate this effect (slightly exaggerated). The phantom (see Fig. 130)
is a simplified thorax with fillable heart. The lungs are made from a light material that displaces water but cannot
take up activity. The lid has a number of valves that allow independent filling of internal structures. A tube was
connected in an internal loop between two valves, allowing it to be filled with contrast (ioversol, 350  mg/mL)
without removing it from the scanner. Figure 130 shows a three dimensional rendering of the phantom with contrast.
The phantom background (6.7  L) and the ‘myocardium’ (0.30  L) were injected with 300  MBq and 40  MBq of
99m
Tc, respectively, leading to a ratio of 3:1. The experiments were performed on a Philips Precedence  16. The
CT scans were performed at peak voltage 90 kV and 140 kV followed by SPECT for one hour (see Fig. 131). The
internal tube ‘vessel’ was then injected with undiluted contrast and the scans repeated.
Images were reconstructed NAC (not shown) and with attenuation maps derived from the CT scans with and
without contrast present (see Fig. 132). Only SPECT data without contrast are shown in Fig. 133, since this is the
most relevant condition, simulating that the injected contrast has been distributed in the body. Only if a difference
image is calculated is it clear that the contrast material induces an increase in reconstructed SPECT counts through
the attenuation map (see Fig. 134).

FIG. 130. Thorax phantom and a three dimensional rendering of SPECT/CT with contrast.

117
FIG. 131. Reconstructed transaxial CT scans at two energies and with and without contrast present. The contrast induces streak
artefacts at tube sites, slightly worse at 90 kV than at 140 kV.

FIG. 132. The generated attenuation maps from two different CT kV values each with and without contrast. The extent of the area of
increased attenuation is larger than the tube due to the applied smoothing. No streak artefacts are visible.

118
FIG. 133. Reconstructed SPECT images based on attenuation maps from two different CT kV values, with and without contrast. It is
difficult to detect differences by pure visual inspection.

FIG. 134. Difference image between reconstructions of acquisitions performed with and without contrast. Only if a difference image
is calculated is it clear that the contrast material introduces an increase in reconstructed SPECT counts through the attenuation
correction. In this case about 5–10% of maximum. Images reconstructed with attenuation maps based on 140 kV CT.

119
Case B

A 22 year old woman had a renal arteriography with iodine contrast in the morning, followed by a SPECT/
CT of the abdomen later the same day after injection with 80  MBq of 99mTc-DMSA. Owing to reduced kidney
function, local, high iodine concentrations were seen in the CT images used for AC of SPECT, for example in the
right kidney (see Fig. 135). A visual comparison of the SPECT images without and with CTAC does not show any
significant difference (artefact) from the iodine (see Figs 136 and 137).

FIG. 135. Iodine contrast is present in the right kidney (maximum value on the CT scan is about 400 HU).

FIG. 136. SPECT image without CTAC.

120
 FIG. 137. SPECT image with CTAC.

Guidance

The local effect of the presence of iodine contrast during CT scans used for AC of SPECT is generally rather
limited and does not exclude the use of contrast scans for this purpose. The magnitude of the effect depends on
several parameters, and it will also depend on the vendor’s actual implementation of the algorithm.

4.7.5. ECG gating: Arrhythmia

Background

Many myocardial SPECT studies require ECG gating to assess cardiac kinetics. If the regularity of the cardiac
cycle is compromised (e.g. by atypical cardiac arrhythmias), the resulting analysis of gated data can be affected.

Case

In the myocardial perfusion study shown in Fig. 138(A), an arrhythmia has caused problems with the analysis
of the cardiac study resulting in an inversion of the cardiac volume versus time curve and an inaccurate measured
left ventricular ejection fraction (LVEF). In Fig. 138(B), once the arrhythmia has been resolved in the patient, and
with the same injection, the data are processed correctly and the cardiac cycle is now accurately represented.

Guidance

When performing ECG gated SPECT, the ECG trace should be assessed to ensure that the cardiac cycle is
regular before acquiring data for clinical interpretation.

121
FIG. 138. Myocardial perfusion SPECT study. (A) Patient with arrhythmia demonstrating a poor representation of the myocardial
cycle. (B) Once the arrhythmia has been resolved, the gated data are better represented.

122
 REFERENCES

[1] KUHL, D.E., EDWARDS, R.Q., Image separation radioisotope scanning, Radiol. 84 (1963) 653–662.
[2] MUEHLLEHNER, G., A tomographic scintillation camera, Phys. Med. Biol. 16 (1971) 87–96.
[3] BUDINGER, T.F., GULLBERG, G.T., Three-dimensional reconstruction in nuclear medicine emission imaging, IEEE Trans.
Nucl. Sci. 21 (1974) 2–20.
[4] LARSSON, S.A., Gamma camera emission tomography: Development and properties of a multi-sectional emission computed
tomography system, Acta Radiol. Suppl. 363 (1980) 1–75.
[5] HASEGAWA, B.H., GINGOLD, E.L., REILLY, S.M., LIEW, S.-C., CANN, C.E., “Description of a simultaneous emission-
transmission CT system”, Medical Imaging IV: Image Formation (Proc. SPIE Conf. 1231, Newport Beach, CA, 1990),
(SCHNEIDER, R.H., Ed.), International Society for Optics and Photonics (1990) 50–60.
[6] INTERNATIONAL ATOMIC ENERGY AGENCY, IAEA Quality Control Atlas for Scintillation Camera Systems, IAEA,
Vienna (2003).
[7] INTERNATIONAL ATOMIC ENERGY AGENCY, PET/CT Atlas on Quality Control and Image Artefacts, IAEA Human
Health Series No. 27, IAEA, Vienna (2014).
[8] ERLANDSSON, K., KACPERSKI, K., VAN GRAMBERG, D., HUTTON, B.F., Performance evaluation of D-SPECT: A novel
SPECT system for nuclear cardiology, Phys. Med. Biol. 54 (2009) 2635–2649.
[9] BOCHER, M., et al., A fast cardiac gamma camera with dynamic SPECT capabilities: Design, system validation and future
potential, Eur. J. Nucl. Med. Mol. Imaging 37 (2010) 1887–1902.
[10] DICKSON, J.C., et al., The impact of reconstruction method on the quantification of DaTSCAN images, Eur. J. Nucl. Med.
Mol. Imaging 37 (2009) 23–35.
[11] JASZCZAK, R.J., LI, J., WANG, H., ZALUTSKY, M.R., COLEMAN, R.E., Pinhole collimation for ultra-high-resolution,
small-field-of-view SPECT, Phys. Med. Biol. 39 3 (1994) 425–437.
[12] MAHMOOD, S.T., ERLANDSSON, K., CULLUM, I., HUTTON, B.F., Design of a novel slit-slat collimator system for
SPECT imaging of the human brain, Phys. Med. Biol. 54 (2009) 3433–3449.
[13] GOORDEN, M.C., RENTMEESTER, M.C., BEEKMAN, F.J., Theoretical analysis of full-ring multi-pinhole brain SPECT,
Phys. Med. Biol. 54 (2009) 6593–6610.
[14] VIDAL, R., et al., Impact of attenuation correction by simultaneous emission/transmission tomography on visual assessment of
201
Tl myocardial perfusion images, J. Nucl. Med. 40 (1999) 1301–1309.
[15] HAWMAN, P.C., HAINES, E.J., The cardiofocal collimator: A variable-focus collimator for cardiac SPECT, Phys. Med.
Biol. 39 (1994) 439–450.
[16] BORGES-NETO, S., et al., Clinical results of a novel wide beam reconstruction method for shortening scan time of Tc-99m
cardiac SPECT perfusion studies, J. Nucl. Cardiol. 14 (2007) 555–565.
[17] GAMBHIR, S.S., et al., A novel high-sensitivity rapid-acquisition single-photon cardiac imaging camera, J. Nucl. Med. 50
(2009) 635–643.
[18] DEPUEY, E.G., BOMMIREDDIPALLI, S., CLARK, J., THOMPSON, L., SROUR, Y., Wide beam reconstruction ‘quarter-
time’ gated myocardial perfusion SPECT functional imaging: A comparison to ‘full-time’ ordered subset expectation maximum,
J. Nucl. Cardiol. 16 (2009) 736–752.
[19] STONE, J.M., BRESSAN, R.A., ERLANDSSON, K., ELL, P.J., PILOWSKY, L.S., Non-uniform blockade of intrastriatal
D2/D3 receptors by risperidone and amisulpride, Psychopharmacol. 180 (2005) 664–669.
[20] STONE, J.M., et al., Ketamine displaces the novel NMDA receptor SPET probe [123I]CNS-1261 in humans in vivo, Nucl. Med.
Biol. 33 (2006) 239–243.
[21] BEN-HAIM, S., et al., Quantification of myocardial perfusion reserve using dynamic SPECT imaging in humans: A feasibility
study, J. Nucl. Med. 54 (2013) 873–879.
[22] CHERRY, S.R., SORENSON, J.A., PHELPS, M.E., Physics in Nuclear Medicine, Elsevier Health Sciences, Philadelphia,
PA (2012).
[23] BRUYANT, P.P., Analytic and iterative reconstruction algorithms in SPECT, J Nucl Med 43 10 (2002) 1343–1358.
[24] SHEPP, L.A., VARDI, Y., Maximum likelihood reconstruction for emission tomography, IEEE Trans. Med. Imaging 1
(1982) 113–122.
[25] LANGE, K., CARSON, R., EM reconstruction algorithms for emission and transmission tomography, J. Comput. Assist.
Tomogr. 8 (1984) 306–316.
[26] NUYTS, J., et al., Simultaneous maximum a posteriori reconstruction of attenuation and activity distributions from emission
sinograms, IEEE Trans. Med. Imaging 18 (1999) 393–403.
[27] LEHOVICH, A., et al., Human-observer LROC study of lesion detection in Ga-67 SPECT images reconstructed using map
with anatomical priors, IEEE Nucl. Sci. Symp. Conf. Rec. 3 (2006) 1699–1702.

123
[28] VIJA, H., Introduction to xSPECT* Technology: Evolving Multi-modal SPECT to Become Context-based and Quantitative,
White Paper, Siemens Medical Solutions USA, Hoffman Estates, IL (2015).
[29] HUDSON, H.M., LARKIN, R.S., Accelerated image reconstruction using ordered subsets of projection data, IEEE Trans. Med.
Imaging 13 (1994) 601–609.
[30] CHANG, L.T., A method for attenuation correction in radionuclide computed tomography, IEEE Trans. Nucl. Sci. 25
(1978) 638–643.
[31] SOHLBERG, A., WATABE, H., IIDA, H., Three-dimensional SPECT reconstruction with transmission-dependent scatter
correction, Ann. Nucl. Med. 22 (2008) 549–556.
[32] ICHIHARA, T., OGAWA, K., MOTOMURA, N., KUBO, A., HASHIMOTO, S., Compton scatter compensation using the
triple-energy window method for single- and dual-isotope SPECT, J. Nucl. Med. 34 (1993) 2216–2221.
[33] INTERNATIONAL ATOMIC ENERGY AGENCY, Nuclear Medicine Physics: A Handbook for Teachers and Students, IAEA,
Vienna (2014).
[34] WILLEMINK, M.J., et al., Iterative reconstruction techniques for computed tomography — Part 1: Technical principles, Eur.
Radiol. 23 (2013) 1623–1631.
[35] INTERNATIONAL ATOMIC ENERGY AGENCY, Quality Assurance Programme for Computed Tomography: Diagnostic and
Therapy Applications, IAEA Human Health Series No. 19, IAEA, Vienna (2012).
[36] BEISTER, M., KOLDITZ, D., KALENDER, W.A., Iterative reconstruction methods in X-ray CT, Phys. Med. 28 (2012) 94–108.
[37] KALENDER, W.A., Computed Tomography, 2nd edn, Publicis, Erlangen (2005).
[38] BERGER, M.J., et al., XCOM: Photon Cross Sections Database, NIST Standard Reference Database 8 (XGAM), National
Institute of Standards and Technology,
www.nist.gov/pml/xcom-photon-cross-sections-database
[39] HUBBELL, J.H., SELTZER, S.M., X-ray Mass Attenuation Coefficients, NIST Standard Reference Database 126, National
Institute of Standards and Technology,
www.nist.gov/pml/x-ray-mass-attenuation-coefficients
[40] CHUANYONG, B., LING, S., DA SILVA, A.J., ZUO, Z., A generalized model for the conversion from CT numbers to linear
attenuation coefficients, IEEE Trans. Nucl. Sci. 50 (2003) 1510–1515.
[41] INTERNATIONAL COMMISSION ON RADIATION UNITS AND MEASUREMENTS, Tissue Substitutes in Radiation
Dosimetry and Measurement, ICRU Report 44, ICRU, Bethesda, MD (1989).
[42] INTERNATIONAL ATOMIC ENERGY AGENCY, Quantitative Nuclear Medicine Imaging: Concepts, Requirements and
Methods, IAEA Human Health Reports No. 9, IAEA, Vienna (2014).
[43] BLANKESPOOR, S.C., et al., Development of an emission-transmission CT system combining X-ray CT and SPECT, IEEE
Nucl. Sci. Symp. Conf. Rec. 4 (1994) 1758–1761.
[44] SERET, A., NGUYEN, D., BERNARD, C., Quantitative capabilities of four state-of-the-art SPECT–CT cameras, EJNMMI
Res. 2 (2012) 45.
[45] INTERNATIONAL ATOMIC ENERGY AGENCY, Quality Assurance for SPECT Systems, IAEA Human Health Series No. 6,
IAEA, Vienna (2009).
[46] AMERICAN ASSOCIATION OF PHYSICISTS IN MEDICINE, Acceptance Testing and Annual Physics Survey
Recommendations for Gamma Camera, SPECT, and SPECT/CT Systems, AAPM Report No. 177, AAPM, Alexandria,
VA (2019).
[47] NATIONAL ELECTRICAL MANUFACTURERS ASSOCIATION, Performance Measurements of Gamma Cameras, NEMA
Standard NU 1-2018, Rosslyn, VA (2019).
[48] BUSEMANN SOKOLE, E., PŁACHCINSKA, A., BRITTEN, A., Acceptance testing for nuclear medicine instrumentation,
Eur. J. Nucl. Med. Mol. Imaging 37 (2010) 672–681.
[49] BUSEMANN SOKOLE, E., et al., Routine quality control recommendations for nuclear medicine instrumentation, Eur. J.
Nucl. Med. Mol. Imaging 37 (2010) 662–671.
[50] ALDRIDGE, M.D., et al., Clinical evaluation of reducing acquisition time on single-photon emission computed tomography
image quality using proprietary resolution recovery software, Nucl. Med. Commun. 34 (2013) 1116–1123.
[51] MAWLAWI, O., et al., Truncation artifact on PET/CT: Impact on measurements of activity concentration and assessment of a
correction algorithm, AJR Am. J. Roentgenol. 186 (2006) 1458–1467.

124
 ABBREVIATIONS

AAPM American Association of Physicists in Medicine

AC attenuation correction
AOR axis of rotation
CFOV central field of view
COR centre of rotation
CT computed tomography
CTAC CT based attenuation correction
CTDI computed tomography dose index
CZT cadmium zinc telluride
EANM European Association of Nuclear Medicine
ECG electrocardiogram
FBP filtered back projection
FOV field of view
FWHM full width at half maximum
HDP hydroxyethylene diphosphonate
HU Hounsfield unit
LEHR low energy high resolution
LEUHR low energy ultra high resolution
LPO left posterior oblique
MEGP medium energy general purpose
MHR multiple head registration
MIBI methoxyisobutylisonitrile
MLEM maximum likelihood expectation maximization
NAC no attenuation correction
NEMA National Electrical Manufacturers Association
OSEM ordered subset expectation maximization
PET positron emission tomography
PSF point spread function
RAO right anterior oblique
ROI region of interest
SBR specific binding ratio
SPECT single photon emission computed tomography
UFOV useful field of view

125
 CONTRIBUTORS TO DRAFTING AND REVIEW

Arreola, M. University of Florida, United States of America

Bianchi, C. ASST Papa Giovanni XXIII, Italy

Dickson, J.C. University College London Hospitals, United Kingdom

Fuglsang, S. Copenhagen University Hospital, Denmark

Holm, S. Rigshospitalet, Copenhagen University Hospital, Denmark

Izaki, M. Instituto do Coração, Hospital das Clínicas da Faculdade de Medicina da Universidade de São
Paulo, Brazil

Kappadath, C.S. The University of Texas MD Anderson Cancer Center, United States of America

Mawlawi, O. The University of Texas MD Anderson Cancer Center, United States of America

de Nijs, R. Rigshospitalet, Copenhagen University Hospital, Denmark

Oliveira, M.A. Instituto do Coração, Hospital das Clínicas da Faculdade de Medicina da Universidade de São
Paulo, Brazil

Poli, G.L. International Atomic Energy Agency

Reichkendler, M.H. Rigshospitalet, Copenhagen University Hospital, Denmark

Robilotta, C.C. Institute of Physics, University of São Paulo, Brazil

Søndergaard, L.R. Copenhagen University Hospital, Denmark

Teles Garcez, A. Instituto de Radiologia, Hospital das Clínicas da Faculdade de Medicina da Universidade de
São Paulo, Brazil

Consultants Meetings

Vienna, Austria: 29–31 October 2014, 6–8 May 2015, 22–26 February 2016, 12–16 December 2016

127
@ No. 26

ORDERING LOCALLY
IAEA priced publications may be purchased from the sources listed below or from major local booksellers.
Orders for unpriced publications should be made directly to the IAEA. The contact details are given at
the end of this list.

NORTH AMERICA
Bernan / Rowman & Littlefield
15250 NBN Way, Blue Ridge Summit, PA 17214, USA
Telephone: +1 800 462 6420 • Fax: +1 800 338 4550
Email: [email protected] • Web site: www.rowman.com/bernan

Renouf Publishing Co. Ltd

22-1010 Polytek Street, Ottawa, ON K1J 9J1, CANADA
Telephone: +1 613 745 2665 • Fax: +1 613 745 7660
Email: [email protected] • Web site: www.renoufbooks.com

REST OF WORLD
Please contact your preferred local supplier, or our lead distributor:

Eurospan Group
Gray’s Inn House
127 Clerkenwell Road
London EC1R 5DB
United Kingdom

Trade orders and enquiries:

Telephone: +44 (0)176 760 4972 • Fax: +44 (0)176 760 1640
Email: [email protected]

Individual orders:
www.eurospanbookstore.com/iaea

For further information:

Telephone: +44 (0)207 240 0856 • Fax: +44 (0)207 379 0609
Email: [email protected] • Web site: www.eurospangroup.com

Orders for both priced and unpriced publications may be addressed directly to:
Marketing and Sales Unit
International Atomic Energy Agency
Vienna International Centre, PO Box 100, 1400 Vienna, Austria
Telephone: +43 1 2600 22529 or 22530 • Fax: +43 1 26007 22529
Email: [email protected] • Web site: www.iaea.org/publications
19-00661
 RELAT ED P U B L I C ATI ON S

IAEA HUMAN HEALTH SERIES PUBLICATIONS

QUALITY ASSURANCE FOR PET AND PET/CT SYSTEMS
The mandate of the IAEA human health programme originates from Article II of
IAEA Human Health Series No. 1
its Statute, which states that the “Agency shall seek to accelerate and enlarge the STI/PUB/1393 (145 pp.; 2009)
contribution of atomic energy to peace, health and prosperity throughout the world”. ISBN 978–92–0–103609–4 Price: €32.00
The main objective of the human health programme is to enhance the capabilities of
IAEA Member States in addressing issues related to the prevention, diagnosis and
QUALITY ASSURANCE FOR SPECT SYSTEMS
treatment of health problems through the development and application of nuclear
IAEA Human Health Series No. 6
techniques, within a framework of quality assurance. STI/PUB/1394 (249 pp.; 2009)
Publications in the IAEA Human Health Series provide information in the areas ISBN 978–92–0–103709–1 Price: €52.00
of: radiation medicine, including diagnostic radiology, diagnostic and therapeutic nuclear
medicine, and radiation therapy; dosimetry and medical radiation physics; and stable
QUALITY ASSURANCE PROGRAMME FOR COMPUTED TOMOGRAPHY:
isotope techniques and other nuclear applications in nutrition. The publications have a
DIAGNOSTIC AND THERAPY APPLICATIONS
broad readership and are aimed at medical practitioners, researchers and other
IAEA Human Health Series No. 19
professionals. International experts assist the IAEA Secretariat in drafting and reviewing STI/PUB/1557 (171 pp.; 2012)
these publications. Some of the publications in this series may also be endorsed or co- ISBN 978–92–0–128910–0 Price: €48.00
sponsored by international organizations and professional societies active in the relevant
fields.
PET/CT ATLAS ON QUALITY CONTROL AND IMAGE ARTEFACTS
There are two categories of publications in this series: IAEA Human Health Series No. 27
STI/PUB/1642 (101 pp.; 2014)
IAEA HUMAN HEALTH SERIES ISBN 978–92–0–101014–8 Price: €48.00
Publications in this category present analyses or provide information of an
advisory nature, for example guidelines, codes and standards of practice, and quality
assurance manuals. Monographs and high level educational material, such as graduate
texts, are also published in this series.

IAEA HUMAN HEALTH REPORTS

Human Health Reports complement information published in the IAEA Human
Health Series in areas of radiation medicine, dosimetry and medical radiation physics,
and nutrition. These publications include reports of technical meetings, the results of
IAEA coordinated research projects, interim reports on IAEA projects, and educational
material compiled for IAEA training courses dealing with human health related subjects.
In some cases, these reports may provide supporting material relating to publications
issued in the IAEA Human Health Series.

All of these publications can be downloaded cost free from the IAEA web site:
http://www.iaea.org/Publications/index.html
Further information is available from:
Marketing and Sales Unit
International Atomic Energy Agency
Vienna International Centre
PO Box 100
1400 Vienna, Austria

Readers are invited to provide their impressions on these publications.

Information may be provided via the IAEA web site, by mail at the address given above,
or by email to:
[email protected].

www.iaea.org/books
 IAEA HUMAN HEALTH SERIES No.36
The complexity of a combined SPECT/CT system requires
rigorous quality control procedures. However, the information
provided by clinical examinations needs to be complemented
with an understanding of potential problems arising from the
combined imaging procedure. This publication presents an
overview of quality control procedures in SPECT and SPECT/CT
and describes the pitfalls and image artefacts that can occur.

IAEA HUMAN HEALTH SERIES IAEA HUMAN HEALTH SERIES

No. 36

SPECT/CT Atlas of Quality Control and Image Artefacts

SPECT/CT Atlas of
Quality Control and
Image Artefacts

INTERNATIONAL ATOMIC ENERGY AGENCY

VIENNA
ISBN 978–92–0–103919–4
ISSN 2075–3772 1