QUALITY CONTROL AND

DOSIMETRY IN THE
NUCLEAR MEDICINE CLINIC
Matthew Palmer, PhD, DABR
Nuclear Medicine Physicist
Beth Israel Deaconess Medical Center
Assistant Professor, Radiology
Harvard Medical School
Conflict of Interest Statement
• No conflicts to disclose.
• No business or professional relationships that could be
construed as influencing what I am about to present.
Outline
• Quality Control in Nuclear Medicine – an
overview
• QA Modernization efforts at BIDMC
• A Phantom for Annual PET/CT testing
• Dosimetry for Clinical Nuclear Medicine
• PET/CT Dosimetry
• PET/CT Dose Reduction
Part I – Quality Control
• Quality Control in Nuclear Medicine – an
overview
• QA Modernization efforts at BIDMC
• A Phantom for Annual PET/CT testing
• Dosimetry for Clinical Nuclear Medicine
• PET/CT Dosimetry
• PET/CT Dose Reduction
Quality Control
• Instrument QC in the clinic consists of:
• Measurement Devices: daily checks for all devices, quarterly and
annual tests for dose calibrators, well-counters, and survey
instruments
• Diagnostic Imaging Equipment:
• Gamma Cameras: Daily checks for all cameras then weekly, monthly,
quarterly and annual tests
• PET Scanner: Daily checks, quarterly tests and annual tests.

• Non Instrument QC
• Radiopharmacy (off site) performs quality control for
radiopharmaceuticals. We only verify activity prior to injection.
• Physicians are QC’d – they perform randomized double-reads and
do followup of cases that are referred to cath-lab or surgery.
Quality Control
• Instrument QC in the clinic consists of:
• Measurement Devices: daily checks for all devices, quarterly and
annual tests for dose calibrators, well-counters, and survey
instruments
• Diagnostic Imaging Equipment:
• Gamma Cameras: Daily checks for all cameras then weekly, monthly,
quarterly and annual tests
• PET Scanner: Daily checks, quarterly tests and annual tests.

• Non Instrument QC
• Radiopharmacy (off site) performs quality control for
radiopharmaceuticals. We only verify activity prior to injection.
• Physicians are QC’d – they perform randomized double-reads and
do followup of cases that are referred to cath-lab or surgery.
Camera Quality Control – Why?
• Quality control (QC) for gamma cameras is mandated
by the State (DPH).

we have to do it.

• Our QC program complies with ACR

guidelines since we are ACR accredited for
Nuclear Medicine and PET.

we have to do it well.

• Scope: 9 gamma cameras (one SPECT/CT),

PET/CT, distributed over 3 sites, and we
perform daily, weekly, monthly, quarterly and
annual testing.
Nukes Imaging Equipment at
BIDMC Philips Precedence
ADAC Fortes Siemens E.CAM SPECT/CT

Digirad Cardius X-ACT GE Discovery LS – PET/CT

ADAC Argus
Daily QC – “Machine Checks”
• Daily machine checks are rapid and sensitive tests that
certify basic operation.
• They test:
• uniformity of response over the field of view
• constancy of detector sensitivity
• accuracy of energy gain calibration
• (cardiac cameras) uniformity of transmission scan reference map
The Daily Flood
• On most cameras, a sheet source
containing Co-57 is placed between
the heads. 10 million counts are
acquired.
• Transmission scan reference maps
are acquired with the built-in Gd-157
rod sources.
• Floods are analyzed with software.

UFOV
CFOV
Uniformity Calculations

UFOV
CFOV

• Defined by NEMA
• Flood image rebinned as 64x64 pixels
• Useful field of view (UFOV) is “area of detector used for imaging…”
• Central field of view (CFOV) is 70% of the UFOV.
• In each FOV, we calculate:
Integral Uniformity (%) =100 * (max-min) / (max+min)
Differential Uniformity = same calc in 5-column/row subfields
Camera QC: Before Modernization

Physicist
QC Log

• Reduce paperwork, reduce errors.

• Central, consistent, standard procedure
independent of manufacturer.
• Automatic tagging/flagging of problems.
• Enable trend analyses and generate periodic
reports.
Web-based Analysis and Log
• We were already sending daily studies to our PACS – we
just had to modify the protocol to maintain data integrity.
• Implementation (phase I – for daily floods):
• Retrieve daily flood images from PACS
• Extract pertinent information (energy, sensitivity) and perform
uniformity analysis.
• Interact with technologist to present results and then maintain them
in a database.
 IT Topology
NM PACS

Image Archive
ADAC n
Needham
ADAC 9
ADAC 6 ADAC 10 QC Images
East Campus

ADAC 2
ADAC 1 ADAC 5
Shapiro ADAC 8
ADAC 11
Siemens

Server Database
Web Analysis
ADAC 7
West Campus
e
QC Imag
Web
Browser
QC Form
Technologist

Engine
Computation
QC Data
QC Server
Monitoring & Auditing NM PACS

Image Archive
QC Image

Web

Server
QC Analysis

Web
Physicist Browser
Trends Analyses

Database
Analysis
QC Server
• Web server uses
ASP
• Coded in VBScript
• Launches
Compiled MATLAB
executable to
perform
calculations
Flagging Out of Bounds Values
 Trend Analysis
adac6 Head 1 emission adac6 Head 2 emission
7 7

6 6

5 5

4 4

3 3

2 2

1 1

0 0
01/01/08 02/01/08 03/01/08 01/01/08 02/01/08 03/01/08
Future Additions
• Weekly QC tests
• Bar Phantoms for resolution and linearity.
• 1/4”, 3/16”, 5/32”, 1/8”
• 1/6”, 1/8”, 1/10”, 1/12”
• 3.5mm, 3.0mm, 2.5mm, 2.0mm

Sheet source
bars
Collimator

head
Future Additions
• Monthly QC Tests
• Center of Rotation (COR) check performs a SPECT study of a
point source and tracks it’s motion in the field of view.
Part II – PET/CT Annual Testing
• Quality Control in Nuclear Medicine – an
overview
• QA Modernization efforts at BIDMC
• A Phantom for Annual PET/CT testing
• Dosimetry for Clinical Nuclear Medicine
• PET/CT Dosimetry
• PET/CT Dose Reduction
PET/CT Annual Testing
• For PET scanners, ACR gives guidance on annual testing
but does not prescribe specific tests as they do for
gamma cameras
• I do:
• Basic physical, mechanical, laser-alignment, …
• SUV calibration check (actually quarterly)
• In-plane spatial resolution and z-axis, slice profile
• Check of PET/CT alignment
• Subjective image quality (ACR phantom)
• Count-rate linearity
PET/CT Annual Testing
• For PET scanners, ACR gives guidance on annual testing
but does not prescribe specific tests as they do for
gamma cameras
• I do:
• Basic physical, mechanical, laser-alignment, …
• SUV calibration check (actually quarterly)
• In-plane spatial resolution and z-axis, slice profile
• Check of PET/CT alignment Using a custom phantom
• Subjective image quality (ACR phantom)
• Count-rate linearity
PET/CT Alignment/Resolution Phantom
• Styrofoam block with three pairs
of glass capillary tubes
• Tubes are <1mm ID, about 8
cm long. Filled with about 0.1
mCi per ml of F-18.
• Scanned with clinical protocols
in both 2D (15 minutes) and 3D
mode. 45 °
• Reconstructed with FBP
(minimal filtering) and with
iterative reconstruction using
clinical defaults.
Identify lines in PET&CT – Semi automated
Reference Points
• The 6 lines are located in 3D in each
of the PET and CT coordinates from
the best-fit straight line through
points on about 10 axial slices
(~4 mm intervals).
• From each pair of lines, for each
modality, we find the midpoint along
the minimum distance orthogonals.
• At the 3 midpoint/reference points
determined by each modality, the
distances are calculated.
• It’s also possible to calculate the
rigid-body translation/rotation
relating PET and CT.
Sample Results
• GE DVCT (700 mm FOV)
Point PET (x,y,z) (mm) CT (x,y,z) (mm) Displacement (mm)
1 (-89.7, 66.0, 53.0) (-90.5, 66.0, 52.4) 1.0
2 (85.8, 90.1, 50.4) (85.7, 89.6, 50.6) 0.6
3 (1.8, -69.8, 105.1) (1.4, -70.7, 105.4) 1.1

• GE DLS (55 mm FOV)

Point PET (x,y,z) CT (x,y,z) (mm) Displacement (mm)

1 (mm) 71.8, 11.8)
(-90.0, (-89.2, 73.0, 11.5) 1.5
2 (85.0, 96.5, 11.8) (86.7, 97.6, 11.7) 1.9
3 (1.3, -63.8, 23.7) (2.6, -63.0, 23.6) 1.4

Optimal translation/rotation: (1.3, 1.0, -0.1), 0.1°

Spatial Resolution for PET
• From the z-aligned tubes, we determine in-plane
resolution.
• The PET Line-spread function (LSF) is well-characterized
by Gaussian distribution so we fit a Gaussian and specify
the FWHM.
-100 -90 -80 -70 -60 -50 -40
Position (mm)
Slice 23 Vertical: FWHM=7.33
1

0.8

0.6

0.4

0.2

0
55 60 65 70 75 80 85 90 95 100 105
Position (mm)
Spatial Resolution: Z-Axis (slice profile)
• From the oblique tubes we fit a Gaussian to the LSF both
horizontally and vertically Slice 23 Horizontal: FWHM=9.22
1
• If the PSF in 3D is 0.8

separable (which is 0.6

0.4
certainly true for a 0.2

Gaussian) and assuming -100 0

-90 -80 -70 -60 -50 -40
that the in-plane resolution Position (mm)
Slice 23 Vertical: FWHM=7.33

in x and y are the same 0.81

(true near the center), then 0.6
the image of a line at 45°to 0.4
0.2
the z-axis is just the 0
55 60 65 70 75 80 85 90 95 100 105
convolution Position (mm)
Part III – Dosimetry
• Quality Control in Nuclear Medicine – an
overview
• QA Modernization efforts at BIDMC
• A Phantom for Annual PET/CT testing
• Dosimetry for Clinical Nuclear Medicine
• Effective Dose from PET/CT
• PET/CT Dose Reduction
 Evolving picture of dose to the public
Cosmic
Terrestrial

Internal

Radon Medical 14%

Radiology
Nuclear Medicine
Background
Consumer Products
13% Occupational/Ind
Other
1985 Radon
ustrial
3%

36%

CT
23%

2006 Nuclear
Medicine
12%
Intervention

Plain
al Medical
Film/Fluoro 7%
47%
5%
Radiological Exams – Relative Dose
Effective dose as a multiple of one AP Chest radiograph (0.02 mSv)
Dosimetry in Nuclear Medicine
• The approach to dosimetry with
internal emitters, like other
modalities, is based on a simple,
idealized phantom model of a 70
kg human.
• Effective dose is calculated as a
weighted sum of absorbed doses
to organs.

• Weights are defined in

ICRP30(1979), ICRP60(1991),
and now ICRP103(2008)
Dosimetry in Nuclear Medicine
• The added complication here is that
activity distributes spatially and
temporally throughout the body.
• Each organ is both a source and a
target. Activity washes in, washes
out (biological decay) and decays
(physical decay).
• So for dosimetry, we have to make
our standard geometrical human
become a standard physiological
human.
The MIRD Method
• For each radiopharmaceutical used in
nuclear medicine the MIRD
committee (SNM) has figured out the
cumulative activity. This is the total
number of disintegrations in an organ
per unit of administered dose.
• They have also figured out by Monte
Carlo simulation, for each
radionuclide, “S-factors” which relate
the absorbed dose in each target
organ per disintegration in each
source organ.
PET(/CT) – The future of Nuclear Medicine?
• First PET scanner was built in
early 1950s by Gordon
Brownell at MGH. It had two
detectors.
• Modern PET scanners were
used through the 70s and 80s
and then clinical PET really
took off in late 1990s, early
200s
• Two key developments:
• 18F-DG as an almost universal
tracer
• Hybrid PET/CT scanners
 PET(/CT) – The future of Nuclear Medicine?
• PET/CT is the
current bright spot
in nuclear imaging.
Volume is steadily
increasing.
• New agents for
Alzheimer’s
screening and 18F-
based agents for
cardiac imaging will
likely accelerate
that trend.

2005 2011
Effective Dose From PET/CT
• Effective dose is dose due to 18F

plus dose from CT scan.

• Even though the CT techniques are
usually “low-dose”, the near whole-
body coverage means that effective
doses are large.
Year F-18 CT Total Dose
2003 20 mCi 140kV/120 mAs
14 mSv 18 mSv 32 mSv
2008 15 mCi 140kV/40 mAs
10 mSv 6 mSv 16 mSv
2013 <10 mCi ?
<7 mSv <6 mSv <8 mSv? ImPACT Calculator
Low-Dose PET/CT of the Future
Effective Dose Equation
• Dose scales linearly with mAs.
• Dose scales quadratically with kVp.
• Effective dose from eyes-to-thighs (torso) CT for 40 mAs calculated
from the ImPACT calculator for GE LS(4) scanner:
Attenuation Correction and Anatomical
Localization
• CT devices on modern PET/CT scanners are full multi-
slice (16/64) diagnostic-quality CT units.
• BUT, they are rarely used as diagnostic CTs.
• They are often read by non-radiologists
• They are usually done as free-breathing, non-contrast studies.

2005: Female, 5’4” 135 lb

colorectal CA, arms-down,
32 mSv

2012: Female, 5’3” 134 lb

breast CA, arms-down,
16 mSv
Attenuation Correction and Anatomical
Localization
• Predicting the CT image quality needed for anatomical
localization is difficult. We have thought about this
problem but haven’t done any work on this.
• Understanding the quality needed for adequate
attenuation correction is more tractable, or at least more
in the realm of physics and engineering, and we have
done some work on that.

Female, BMI=44, arms up Male, BMI=27, arms down

140 kVp, 96 mAs 140 kVp, 60 mAs
Attenuation Correction in PET/CT
• Attenuation correction in PET is the most important
correction to ensure accuracy. Attenuation factors for 511
keV photons at the center of the body are on the order of
20, and at the center of the head are on the order of 7.

Attenuation maps are

prepared from the CT
cross-sections by
mapping Houndsfield
units to 511 keV
equivalent coefficients.
Typically this is a multi-
linear transform:
 Two problems with low-dose CT for AC
• A bias is introduced to the AC map because of non-linear
mapping of noisy samples. This propagates to the PET
reconstruction.
CT

Mu-map

PET

Fahey, Palmer et al., Radiology, 2007

Photon Starvation
• Streak artifacts due to photon starvation
appear along rays that traverse a lot of
dense bone.
• The are particularly severe in low-dose CT
studies, especially at lower kVp.
• If we are going to lower the CT dose even
further, photon starvation is probably the
limiting effect.
• This is a non-linear phenomenon that isn’t
handled well by filtered backprojection
algorithms. It is our hope that emerging
iterative CT reconstruction algorithms will
help.
Data Spectrum Torso Phantom
• As a first step, we’ve begun to
characterize the phenomenon and
test newly-available iterative CT
reconstructions.
• Slabs of Al+Cu (Z=19 versus ~20 for
dense bone) varying thickness
(x3.5cm) attached to phantom.
Starvation Artifact Phantom Experiment
• Scanned in the 4-slice GE LS
scanner (PET/CT) with 4x2.5 mm 1
collimation and techniques from 80
kVp/5 mAs to 140 kVp/160 mAs.
2
• Circular ROIs, ~2 cm diam (1300
pixels), calculate mean and SD.
• Plot versus “inverse dose index” =
106/(kVp2mAs).
• Graph for ROI#2, 3.5cm bone

Photon Starvation vs Bone Thickness
• The more bone, the greater the bias due to starvation.
• Composite data for 80 mAs:

• Relationship, especially with kVp, appears complex and

non-linear. May be difficult to characterize.
Iterative (?) Reconstruction
• Similar experiment with metal slabs simulating bone and
Data Spectrum torso phantom scanned in a GE CT750
HD scanner and the images reconstructed with ASiR
(100%).
• ROI Statistics:
Summary (so-far)
• Both noise and photon starvation in low-dose CTAC
contribute to an underestimate of 511 keV attenuation.
• ASiR (GE’s early “iterative recon”) may reduce noise but
has no effect on photon starvation.
• Effect on PET quantitation is complex function of mAs,
kVp and bone thickness.
Low BMI (<25)
• Possible next-step is to fill mAs
phantom and perform emission Arms Up
scans, correct them and define
Arms Down
an error threshold (say 10%),
then produce charts like

kVp
Acknowledgements
• Frederic Fahey, Boston Children’s Hospital
• Robert Zimmerman
• Larry Barbaras
• Many other colleagues and mentors