Medical Imaging Systems Andreas Maier pdf

download

https://textbookfull.com/product/medical-imaging-systems-andreas-maier/

★★★★★ 4.8/5.0 (44 reviews) ✓ 142 downloads ■ TOP RATED

"Fantastic PDF quality, very satisfied with download!" - Emma W.

DOWNLOAD EBOOK
 Medical Imaging Systems Andreas Maier

TEXTBOOK EBOOK TEXTBOOK FULL

Available Formats

■ PDF eBook Study Guide TextBook

EXCLUSIVE 2025 EDUCATIONAL COLLECTION - LIMITED TIME

INSTANT DOWNLOAD VIEW LIBRARY

Collection Highlights

Medical imaging systems techniques and applications :

computational techniques Leondes

Exceptional Lifespans Heiner Maier

Hendee's physics of medical imaging Fifth Edition Hendee

Medical Imaging and Computer Aided Diagnosis Proceeding of

2020 International Conference on Medical Imaging and
Computer Aided Diagnosis MICAD 2020 Ruidan Su
Fundamental Mathematics and Physics of Medical Imaging 1st
Edition Jack Lancaster

Medical imaging for health professionals: technologies and

clinical applications First Edition Reilly

Big Data in Multimodal Medical Imaging 1st Edition Ayman

El-Baz (Editor)

Level Set Method in Medical Imaging Segmentation 1st

Edition Ayman El-Baz (Editor)

The Naturally Clean Home 3rd Edition Karyn Siegel Maier

 Andreas Maier · Stefan Steidl
Vincent Christlein
Joachim Hornegger (Eds.)
Tutorial
LNCS 11111

Medical
Imaging Systems
An Introductory Guide
Lecture Notes in Computer Science 11111
Commenced Publication in 1973
Founding and Former Series Editors:
Gerhard Goos, Juris Hartmanis, and Jan van Leeuwen

Editorial Board
David Hutchison
Lancaster University, Lancaster, UK
Takeo Kanade
Carnegie Mellon University, Pittsburgh, PA, USA
Josef Kittler
University of Surrey, Guildford, UK
Jon M. Kleinberg
Cornell University, Ithaca, NY, USA
Friedemann Mattern
ETH Zurich, Zurich, Switzerland
John C. Mitchell
Stanford University, Stanford, CA, USA
Moni Naor
Weizmann Institute of Science, Rehovot, Israel
C. Pandu Rangan
Indian Institute of Technology Madras, Chennai, India
Bernhard Steffen
TU Dortmund University, Dortmund, Germany
Demetri Terzopoulos
University of California, Los Angeles, CA, USA
Doug Tygar
University of California, Berkeley, CA, USA
Gerhard Weikum
Max Planck Institute for Informatics, Saarbrücken, Germany
More information about this series at http://www.springer.com/series/7412
Andreas Maier Stefan Steidl
•

Vincent Christlein
Joachim Hornegger (Eds.)

Medical
Imaging Systems
An Introductory Guide
Editors
Andreas Maier Vincent Christlein
Lehrstuhl für Mustererkennung Lehrstuhl für Mustererkennung
Friedrich-Alexander-Universität Friedrich-Alexander-Universität
Erlangen-Nürnberg Erlangen-Nürnberg
Erlangen Erlangen
Germany Germany
Stefan Steidl Joachim Hornegger
Lehrstuhl für Mustererkennung Lehrstuhl für Mustererkennung
Friedrich-Alexander-Universität Friedrich-Alexander-Universität
Erlangen-Nürnbergät Erlangen-Nürnbergät
Erlangen Erlangen
Germany Germany

ISSN 0302-9743 ISSN 1611-3349 (electronic)

Lecture Notes in Computer Science
ISBN 978-3-319-96519-2 ISBN 978-3-319-96520-8 (eBook)
https://doi.org/10.1007/978-3-319-96520-8

Library of Congress Control Number: 2018948380

LNCS Sublibrary: SL6 – Image Processing, Computer Vision, Pattern Recognition, and Graphics

© The Editor(s) (if applicable) and The Author(s) 2018. This book is an open access publication.
Open Access This book is licensed under the terms of the Creative Commons Attribution 4.0 International
License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution
and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and
the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this book are included in the book's Creative Commons license,
unless indicated otherwise in a credit line to the material. If material is not included in the book's Creative
Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use,
you will need to obtain permission directly from the copyright holder.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book are
believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors
give a warranty, express or implied, with respect to the material contained herein or for any errors or
omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.

Cover illustration: Graphical visualization of the Fourier slice theorem. LNCS 11111, p. 154. Used with
permission.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface

The present book is the result of four years of work that started in Winter 2014/15 and
was finally concluded in Summer 2018. As such, numerous hours of work went into
this manuscript by several authors, who were all affiliated with the Pattern Recognition
Lab of the Friedrich-Alexander-University Erlangen-Nuremberg. I truly appreciate the
dedication and the hard work of my colleagues that led to this final manuscript and,
although many already left the lab to take positions in academia and industry, they still
supported the finalization of this book.
While major parts of the book were already completed in Winter 2016/17, Springer
gave us the opportunity to rework the book with new concepts like the geek boxes and
new figures in order to adapt the book to a broader audience. With the present concepts,
we hope that the book is suited to early-stage undergraduate students as well as stu-
dents who already completed fundamental math classes and want to deepen their
knowledge on medical imaging. We believe, the time to improve the manuscript was
well spent and the final polish gave rise to a textbook with a coherent story line. In
particular, we break with the historical development of the described imaging devices
and present, e. g., magnetic resonance imaging before computed tomography, although
they were developed in opposite order. A closer look reveals that this change of order is
reasonable for didactical purposes: magnetic resonance imaging relies mainly on the
Fourier transform, while computed tomography requires understanding of the Fourier
slice theorem discovered by Johann Radon. These observations then also mend the
apparent historical disorder, as we celebrate Joseph Fourier’s 250th birthday this year
and celebrated the 100th birthday of the Radon transform last year.
We also tried to find many graphical explanations for many of the mathematical
operations such that the book does not require complete understanding of all mathe-
matical details. Yet, we also offer details and references to further literature in the
previously mentioned geek boxes as students in the later semesters also need to be
familiar with these concepts. In conclusion, we hope that we created a useful textbook
that will be accessible to many readers. In order to improve this ease of access further,
we chose to publish the entire manuscript as open access book under Creative Com-
mons Attribution 4.0 International License. Thus, any information in this book can
shared, copied, adapted, or remixed even for commercial purposes as long as the
original source is appropriately referenced and a link to the license is provided.

June 2018 Andreas Maier

Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2 System Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.1 Signals and Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.1.1 Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.1.2 Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 Convolution and Correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.1 Complex Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.2 Convolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2.3 Correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.3 Fourier Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.3.1 Types of Fourier Transforms . . . . . . . . . . . . . . . . . . . . . . . 22
2.3.2 Convolution Theorem & Properties . . . . . . . . . . . . . . . . . 25
2.4 Discrete System Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.4.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.4.2 Sampling Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.4.3 Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.5 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

3 Image Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.1 Images and Histograms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.1.1 Images as Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.1.2 Histograms of Images . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2 Image Enhancement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2.1 Window and Level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2.2 Gamma Correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2.3 Histogram Equalization . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.3 Edge Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.4 Image Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.4.1 Filtering – Basics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.4.2 Linear Shift-invariant Filters in Image Processing . . . . . 44
3.4.3 Nonlinear Filters – the Median Filter . . . . . . . . . . . . . . . 47
3.5 Morphological Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.6 Image Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4 Contents

4 Endoscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.1 Minimally Invasive Surgery and Open Surgery . . . . . . . . . . . . . 57
4.2 Minimally Invasive Abdominal Surgery . . . . . . . . . . . . . . . . . . . . 58
4.3 Assistance Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.4 Range Imaging in Abdominal Surgery . . . . . . . . . . . . . . . . . . . . . 63
4.4.1 Stereo Vision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.4.2 Structured Light . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.4.3 Time-of-Flight (TOF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

5 Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.1 Image Formation in a Thin Lens . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.2 Compound Microscope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.3 Bright Field Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.4 Fluorescence Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.5 Phase Contrast Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.6 Quantitative Phase Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.7 Limitation of Light Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.8 Beyond Light Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.9 Light Microscopy Beyond the Diffraction Limit . . . . . . . . . . . . . 88

6 Magnetic Resonance Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

6.1 Nuclear Magnetic Resonance (NMR) . . . . . . . . . . . . . . . . . . . . . . 91
6.1.1 Genesis of the Resonance Effect . . . . . . . . . . . . . . . . . . . . 91
6.1.2 Relaxation and Contrasts . . . . . . . . . . . . . . . . . . . . . . . . . . 95
6.2 Principles of Magnetic Resonance Imaging . . . . . . . . . . . . . . . . . 100
6.2.1 Slice Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
6.2.2 Spatial Encoding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.2.3 k-space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.2.4 Slice-selective vs. Volume-selective 3-D Imaging . . . . . . 105
6.3 Pulse Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.3.1 Spin Echo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
6.3.2 Gradient Echo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
6.4 Advanced Topics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.4.1 Parallel Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.4.2 Spectrally Selective Excitation . . . . . . . . . . . . . . . . . . . . . 114
6.4.3 Non-contrast Angiography . . . . . . . . . . . . . . . . . . . . . . . . . 115
6.4.4 The BOLD Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

7 X-ray Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
7.1.1 Definition of X-rays . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
7.1.2 History and Present . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
7.2 X-ray Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Contents 5

7.3 X-ray Matter Interaction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125

7.3.1 Absorption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
7.3.2 Photoelectric Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
7.3.3 Compton Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
7.3.4 Rayleigh scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
7.4 X-ray Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
7.4.1 Image Intensifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
7.4.2 Flat Panel Detectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
7.4.3 Sources of Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7.5 X-ray Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
7.5.1 Radiography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
7.5.2 Fluoroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
7.5.3 Digital Subtraction Angiography . . . . . . . . . . . . . . . . . . . 143

8 Computed Tomography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
8.1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
8.1.2 Brief History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
8.2 Mathematical Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
8.2.1 Radon Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
8.2.2 Fourier Slice Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
8.3 Image Reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
8.3.1 Analytic Reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
8.3.2 Algebraic Reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . 161
8.3.3 Acquisition Geometries . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
8.4 Practical Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
8.4.1 Spatial Resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
8.4.2 Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
8.4.3 Image Artifacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
8.5 X-ray Attenuation with Polychromatic Attenuation . . . . . . . . . 176
8.5.1 Mono- vs. Polychromatic Attenuation . . . . . . . . . . . . . . . 176
8.5.2 Single, Dual, and Spectral CT . . . . . . . . . . . . . . . . . . . . . 179
8.5.3 Beam Hardening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
8.6 Spectral CT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
8.6.1 Different Spectral CT Measurements . . . . . . . . . . . . . . . . 182
8.6.2 Basis Material Decomposition . . . . . . . . . . . . . . . . . . . . . . 186

9 X-ray Phase Contrast:

Research on a Future Imaging Modality . . . . . . . . . . . . . . . . . . 191
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
9.2 Talbot-Lau Interferometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
9.2.1 Talbot-Lau Interferometer Setup . . . . . . . . . . . . . . . . . . . 195
9.2.2 Phase Stepping and Reconstruction . . . . . . . . . . . . . . . . . 197
9.3 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
9.4 Research Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
6 Contents

10 Emission Tomography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207

10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
10.2 Physics of Emission Tomography . . . . . . . . . . . . . . . . . . . . . . . . . 208
10.2.1 Photon Emission . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
10.2.2 Photon Interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
10.3 Acquisition Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
10.3.1 SPECT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
10.3.2 PET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
10.4 Reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
10.4.1 Filtered Back-Projection . . . . . . . . . . . . . . . . . . . . . . . . . . 219
10.4.2 Iterative Reconstruction . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
10.4.3 Quantitative Reconstructions . . . . . . . . . . . . . . . . . . . . . . 222
10.4.4 Practical Considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . 224
10.5 Clinical Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
10.5.1 Diagnostics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
10.5.2 Therapy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
10.6 Hybrid Imaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
10.6.1 Clinical Need . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
10.6.2 Advent und Acceptance of Hybrid Scanners . . . . . . . . . . 231
10.6.3 Further Benefits of Hybrid Imaging . . . . . . . . . . . . . . . . . 232

11 Ultrasound . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
11.2 Physics of Sound Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
11.2.1 Sound Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
11.2.2 Sound Wave Characteristics at Boundaries . . . . . . . . . . 239
11.2.3 Attenuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
11.3 Image Acquisition for Diagnostics . . . . . . . . . . . . . . . . . . . . . . . . 243
11.3.1 Transducers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
11.3.2 Piezoelectric Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
11.3.3 Spatial Resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
11.3.4 Imaging Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
11.4 Safety Aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247

12 Optical Coherence Tomography . . . . . . . . . . . . . . . . . . . . . . . . . . . 251

12.1 Working Principle of OCT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
12.1.1 Michelson Interferometer . . . . . . . . . . . . . . . . . . . . . . . . . . 253
12.1.2 Coherence Length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
12.2 Time Domain OCT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
12.3 Fourier Domain OCT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
12.4 OCT Angiography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 256
12.5 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257

Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
Chapter 1

Introduction

Author: Andreas Maier

The design and manufacturing of modern medical devices requires knowl-

edge of several disciplines, ranging from physics, over material science, to
computer science. Thus, designing a single lecture as an introduction to med-
ical engineering faces a lot of challenges. Nonetheless, the manuscript Medi-
cal Imaging Systems – An Introductory Guide aims at being a complete and
comprehensive introduction to this field for students in the early semesters.
Medical imaging devices are by now an integral part of modern medicine,
and have probably already been encountered by all students in their personal
life.
This book does not simply summarize the content of the lecture held in
Erlangen. Instead, it should be understood as additional material to gain a
better understanding of the theory that is covered in the lecture. To give a
complete introduction, the lecture notes also cover basic math and physics
that are required to understand the underlying principles of the imaging
devices. However, we try to limit this to the very basics. Obviously, this
is not sufficient to describe everything in the appropriate level of detail. For
this reason, we introduced geek boxes (cf. Geek Box 1.1) that contain optional
additional background information. This concept will be used in all chapters
of the book which are summarised in the following sections.
Chap. 2 and 3 of this book cover an introduction to signal and image pro-
cessing. Chap. 2 introduces the concepts of filtering, convolution, and Fourier
transforms for 1-D signals, all of which are fundamental tools that are later
on used across the entire book. We try to explain why these concepts are
required and as most image processing is digital also emphasize the discrete
algorithmic counter parts. At the beginning of Chap. 3, the transition to im-
ages is made, and therefore also the transition from 1-D to 2-D. The chapter

c The Author(s) 2018

A. Maier et al. (Eds.): Medical Imaging Systems, LNCS 11111, pp. 7–12, 2018.
https://doi.org/10.1007/978-3-319-96520-8_1
8 1 Introduction

Geek Box 1.1: Geek Boxes

We designed the manuscript to be readable from the first semester

on. However, we felt that we need to demonstrate that there is much
more depth that we could go into. In order not to confuse a less
experienced reader, we omitted most equations and math from the
main text and relocated them to geek boxes that go into more detail
and give references to further reading. In addition, we also refresh
concepts that are already known to most readers. Nonetheless, the
important concepts are already mentioned in the main text. This way,
the reader can return to this book at a time when these concepts are
introduced, e. g., in more advanced math courses seemingly unrelated
to medical imaging. As such this book can be read twice: once omitting
all geek boxes to get an overview on the field and a second time with
a more throrough focus on the mathematical details.

covers the basics of image processing and explains how different image trans-
formations such as edge detection and blurring are implemented as image
filters using convolution.
The following chapters cover examples for imaging devices using stan-
dard optics. In this book, endoscopy and microscopy are discussed as typical
modalities of this genre. Endoscopes, see Chap. 4, were among the first med-
ical imaging devices that were used. Images can be acquired by using long
and flexible optical fibers that are able to transport visible light through the
body of a patient.
Microscopes also use visible light. However, tissue samples or cells have to
be extracted from the body first, e. g., in a biopsy. Then the microscope’s op-
tics are used to acquire images at high magnifications that allow the imaging
of individual cells and even smaller structures. Microscopes and the principles
of optics are described in Chap. 5.
Magnetic resonance imaging (MRI), see Chap. 6 uses electromagnetic
waves to excite water atoms inside the human body. Once the excitation
is stopped, the atoms return to their normal state and by doing so emit
the same electromagnetic radio wave that was used to excite them. This ef-
fect is called nuclear magnetic resonance. Using this effect, an MRI image is
obtained. Fig. 1.1 shows a state-of-the-art MR scanner.
X-ray imaging devices, see Chap. 7, use light of very high energy. However,
the light is no longer visible for the human eye. The higher energy of the light
allows for a deeper penetration of the body. Due to different absorption rates
of X-rays, different body tissues can be distinguished on X-ray images. Tissues
with high X-ray absorption, e. g., bones, become visible as bright structures
in X-ray projection images. Today, X-rays are among the most widely spread
1 Introduction 9

Figure 1.1: MRI is based on nuclear magnetic resonance which does not
involve ionizing radiation. For this reason MRI is often used in pediatric
applications. Image courtesy of Siemens Healthineers AG.

Figure 1.2: X-ray projection images are one of the most wide-spread imaging
modalities. Image courtesy of Siemens Healthineers AG.
10 1 Introduction

Figure 1.3: Modern CT systems allow even scanning of the beating heart.
Image courtesy of Siemens Healthineers AG.

medical imaging technologies. An example for an X-ray imaging device is

shown in Fig. 1.2.
Computed tomography (CT) uses X-rays to reconstruct slice and volume
data as described in Chap. 8. The total absorption along the path of an X-ray
through the body is actually given by the sum of absorptions by tissues with
different absorption characteristics along its path. Thus, a measurement of
the absorptions of X-rays from different directions allows for a reconstruction
of slice images through the patient’s body. In doing so, much better contrast
between types of soft tissue is obtained. One is even able to differentiate
between different tissue types such as brain and brain tumor. Once several
slices are combined, the entire volume can be reconstructed by stacking the
slices, which is then referred to as a 3-D image. Fig. 1.3 shows a state-of-the-
art CT system with a gantry that rotates at 4 Hz.
X-rays essentially are electromagnetic waves that can be described by their
amplitude, wavelength, and phase. Phase contrast imaging exploits the effect
that an X-ray passing through tissue is not only influenced by absorption,
but that also the phase of the electromagnetic wave is shifted. Chap. 9 shows
that the phase shift of X-rays can be used to visualize the tissue the X-
rays have passed. Today, phase contrast imaging is not yet used in clinical
practice. In fact, due to the high requirements on the type of irradiation, such
images often require a synchrotron as the source of the radiation. However,
new developments in research now allow to generate phase contrast images
using a normal clinical X-ray tube, which renders the application clinically
feasible. At present, technical limitations allow only the scanning of small
specimen such as peanuts and the mechanical design is still challenging. First
image results indicate that the modality might be of high clinical relevance.
Fig. 1.4 shows the reconstruction of peanut fibers that are in the range of
1 Introduction 11

Figure 1.4: An X-ray dark-field setup can be used to reconstruct the ori-
entation of fibers that are smaller than the detector resolution. The image
on the left shows the reconstructed fiber orientation in different layers of a
peanut. The image on the right shows a microscopic visualization of the waist
of the peanut (picture courtesy of ECAP Erlangen).

Figure 1.5: Modern SPECT/CT systems combine different modalities to

achieve multi-modal imaging. Image courtesy of Siemens Healthineers AG.

several micrometers. Phase contrast allows for a reconstruction of these fibers,

although the resolution of the used imaging device based on the absorption
of X-rays was only 0.1 mm.
Emission tomography, described in Chap. 10, is used for imaging different
bodily functions. It uses tracers, which are molecules that are marked with
radioactive atoms. For example one can introduce a radioactive atom into a
sugar molecule. When this tracer is consumed by the body it will follow the
normal metabolism, and its path through the body can be followed. While
sugar consumption is normal in certain parts of the body such as the muscles
or the brain, tumors also require a lot of sugar for their growth. Thus, emis-
 12 1 Introduction

Figure 1.6: A typical ultrasound system as it can be found in clinics world-

wide. Image courtesy of Siemens Healthineers AG.

sion tomography enables us to see anomalies in sugar consumption within

the body which is useful to spot tumors or metastases. Fig. 1.5 shows a com-
bined single-photon emission computed tomography (SPECT) / CT system
that combines emission tomography with X-ray CT.
Ultrasound (US) uses high-frequency sound waves to penetrate bodily tis-
sue. The sound waves are emitted from a probe that is in direct contact with
the body. The same probe is then also used to measure the reflections of the
sound waves. Given the time between the emission of the sound wave and the
measurement of the reflection, one is able to reconstruct how deep the wave
penetrated the tissue. US is one of the most wide-spread imaging modalities
as it is rather inexpensive compared to other imaging modalities. Fig. 1.6
shows a clinical ultrasound system.
The measurement principle of optical coherence tomography (OCT) is
quite similar to US. However, light waves are used instead of sound waves.
Thus, the measurement process needs to be performed at much higher speed
and penetration depth is much lower than in the case of US. Most applications
are in eye imaging where 3-D images of the eye are generated.

Open Access This chapter is licensed under the terms of the Creative Commons
Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/),
which permits use, sharing, adaptation, distribution and reproduction in any medium
or format, as long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons license and indicate if changes were
made.
The images or other third party material in this chapter are included in the chapter’s
Creative Commons license, unless indicated otherwise in a credit line to the material. If
material is not included in the chapter’s Creative Commons license and your intended
use is not permitted by statutory regulation or exceeds the permitted use, you will
need to obtain permission directly from the copyright holder.
Chapter 2

System Theory

Authors: Peter Fischer, Klaus Sembritzki, and Andreas Maier

2.1 Signals and Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2.2 Convolution and Correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3 Fourier Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4 Discrete System Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.5 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

In the digital age, any medical image needs to be transformed from contin-
uous domain to discrete domain (i. e. 1’s and 0’s) in order to be represented
in a computer. To do so, we have to understand what a continuous and a
discrete signal is. Both of them are handled by systems which will also
be introduced in this chapter. Another fundamental concept is the Fourier
transform as it allows us to represent any time domain signal in frequency
space. In particular, we will find that both representations – time domain
and frequency domain – are equivalent and can be converted into each other.
Having found this important relationship, we can then determine conditions
which will guarantee that also conversion from continuous to discrete domain
and vice versa is possible without loss of information. On the way, we will
introduce several other important concepts that will also find repeated use
later in this book.

c The Author(s) 2018

A. Maier et al. (Eds.): Medical Imaging Systems, LNCS 11111, pp. 13–36, 2018.
https://doi.org/10.1007/978-3-319-96520-8_2
14 2 System Theory

2.1 Signals and Systems

2.1.1 Signals

A signal is a function f (t) that represents information. Often, the indepen-

dent variable t is a physical dimension, like time or space. The output f of the
signal is also called the dependent variable. Signals are everywhere in every-
day life, although we are mostly not aware of them. A very prominent example
is the speech signal, where the independent variable is time. The dependent
variable is the electric signal that is created by measuring the changes of air
pressure using a microphone. The description of the speech generation pro-
cess enables to do efficient speech processing, e. g., radio transmission, speech
coding, denoising, speech recognition, and many more. In general, many do-
mains can be described using system theory, e. g., biology, society, economy.
For our application, we are mainly interested in medical signals.
Both the dependent and the independent variable can be multidimen-
sional. Multidimensional independent variables t are very common in images.
In normal camera images, space is described using two spatial coordinates.
However, medical images, e. g., CT volume scans, can also have three spatial
dimensions. It is not necessary that all dimensions have the same meaning.
Videos have two spatial coordinates and one time coordinate. In the medi-
cal domain, we can also find higher-dimensional examples like time-resolved
4-D MR and CT with three spatial dimensions and one time dimension. To
represent multidimensional values, i. e., vectors, we use bold-face letters t or
|
multiple scalar values, e. g., t = (x, y, z) . The medical field also contains
examples of multidimensional dependent variables f . An example with many
dimensions is the Electroencephalography (EEG). Electrodes are attached to
the skull and measure electrical brain activity from multiple positions over
time. To represent multidimensional dependent variables, we also use bold-
face letters f .
The signals described above are all in continuous domain, e. g., time and
space change continuously. Also, the dependent variables vary continuously
in principle, like light intensity and electrical voltage. However, some sig-
nals exist naturally in discrete domains w. r. t. the independent variable or
the dependent variable. An example for a discrete signal in dependent and
independent variable is the number of first semester students in medical en-
gineering. The independent variable time is discrete in this case. The starting
semesters are WS 2009, WS 2010, WS 2011, and so on. Other points in time
are considered to be constant in this interval. The number of students is
restricted to natural numbers. In general, it is also possible that only the de-
pendent or the independent variable is discrete and the other one continuous.
In addition to signals that are discrete by nature, other signals must be rep-
resented discretely for processing with a digital computer, which means that
the independent variable must be discretized before processing with a com-
2.1 Signals and Systems 15

Figure 2.1: A system H{.} with the input signal f (t) and the output signal
g(t).

puter. Furthermore, data storage in computers has limited precision, which

means that the dependent variable must be discrete. Both are a direct conse-
quence of the finite memory and processing speed of computers. This is the
reason why discrete system theory is very important in practice.
Signals can be further categorized into deterministic and stochastic signals.
For a deterministic signal, the whole waveform is known and can be written
down as a function. In contrast, stochastic signals depend randomly on the
independent variable, e. g., if the signal is corrupted by noise. Therefore, for
practical applications, the stochastic properties of signals are very important.
Nevertheless, deterministic signals are important to analyze the behavior of
systems. A short introduction into stochastic signals and randomness will be
given in Sec. 2.4.3.
This chapter is presents basic knowledge on how to represent, analyze,
and process signals. The correct processing of signals requires some math
and theory. A more in-depth introduction into the concepts presented here
can be found in [3]. The application to medical data is treated in [2].

2.1.2 Systems

Signals are processed in processes or devices, which are abstracted as sys-

tems. This includes not only technical devices, but natural processes like
attenuation and reverberation of speech in transmission through air as well.
Systems have signals as input and as output. Inside the system, the properties
of the signal are changed or signals are related to each other. We describe the
processing of a signal using a system with the operator H{·} that is applied
to the function f . A graphical representation of a system is shown in Fig. 2.1.
An important subtype is the linear shift-invariant system. Linear shift-
invariant systems are characterized by the two important properties of lin-
earity and shift-invariance (cf. Geek Box 2.1 and 2.2).
Another property important for the practical realization of linear shift-
invariant systems is causality. A causal system does not react to the input
16 2 System Theory

Geek Box 2.1: Linear Systems

The linearity property of a system means that linear combinations

of inputs can be represented as the same linear combination of the
processed inputs

H{af (t)} = aH{f (t)} (2.1)

H{f (t) + g(t)} = H{f (t)} + H{g(t)}, (2.2)

with constant a and arbitrary signals f and g. The linearity property

greatly simplifies the mathematical and practical treatment, as the
behavior of the system can be studied on basic signals. The behav-
ior on more complex signals can be inferred directly if they can be
represented as a superposition of the basic signals.

Geek Box 2.2: Shift-Invariant Systems

Shift-invariance denotes the characteristic of a system that its re-

sponse is independent of shifts of the independent variable of the
signal. Mathematically, this is described as

g1 (t) = H{f (t)} (2.3)

g2 (t) = H{f (t − τ )} (2.4)
g1 (t − τ ) = g2 (t), (2.5)

for the shift τ . This means that shifting the signal by τ followed by
processing with the system is identical to processing the signal with
the system followed by a shift with τ .

before the input actually arrives in the system. This is especially important
for signals with time as the independent parameter. However, non-causal
systems do not pose a problem for the independent parameter space, e. g.,
image filters that use information from the left and right of a pixel. Geek
Box 2.3 presents examples for the combination of different system properties.
Linear shift-invariant systems are important in practice and have conve-
nient properties and a rich theory. For linear shift-invariant systems, the ab-
stract operator H{·} can be described completely using the impulse response
h (t) (cf. Sec. 2.2.2) or transfer function H (ξ) (cf. Sec. 2.3.2). The impulse
response is combined with the signal by the operation of convolution. This is
sufficient to describe all linear shift-invariant systems.
2.2 Convolution and Correlation 17

Geek Box 2.3: System Examples

Here are some examples of different systems analyzed w. r. t. linearity,

shift-invariance, and causality. f (t) represents the input and g(t) the
output signal.
• g(t) = 10f (t): linear, shift-invariant, causal
• g(t) = sin(f (t)): non-linear, shift-invariant, causal
• g(t) = 3f (t + 2): linear, shift-invariant, non-causal
• g(t) = f (t) − 2f (t − 1): linear, shift-invariant, causal
• g(t) = f (t) · e(−0.5t) : linear, not shift-invariant, causal

2.2 Convolution and Correlation

This section describes the combination of signals in linear-shift-invariant sys-
tems, i. e., convolution or correlation. Before discussing signal processing in
detail, we will first start by revisiting important mathematical concepts that
will be needed in the following chapters.

2.2.1 Complex Numbers

Complex numbers are an extension to real numbers. They are defined as

z = a + bi. a is called the real part of z and b the imaginary part. Both
act as coordinates in a 2-D space. i is the imaginary unit that spans the
second dimension of this space. The special meaning of i is that i2 = −1.
This makes complex numbers important for many areas in mathematics, but
also in many applied fields like physics and electrical engineering. To extract
the coordinates of the complex number, we use the following definitions

a = Re (z) (2.6)
b = Im (z) . (2.7)

We can directly write z = Re (z) + Im (z) i. Another important definition is

the complex conjugate z̄, which is the same number as z except with the
opposite sign for the imaginary part z̄ = a − bi.
Real numbers are the subset of the complex numbers for which b = 0, i. e.,
no imaginary part. Geometrically, this means that real numbers are defined
on a one-dimensional axis, whereas the complex numbers are defined on a
2-D plane. The geometric interpretation of complex numbers is also helpful
to see the equivalence of the Cartesian coordinate notation z = a + bi and
the polar coordinate notation z = A (cos φ + i sin φ) of complex numbers. The
18 2 System Theory

Geek Box 2.4: Complex Numbers and Geometric Interpretation

If a point on the 2-D plane is seen as a position vector, A is the

length of the vector and φ the angle relative to the real axis. The two
notations can be converted to each other using the following formulas:
p
A = a 2 + b2
arctan ab , if a > 0



arctan a + π, if a < 0 and b ≥ 0
b





arctan b − π, if a < 0 and b < 0

φ= π a


 2 , if a = 0 and b > 0
− π
, if a = 0 and b < 0


2


undefined, if a = 0 and b = 0



a = A cos φ
b = A sin φ

polar coordinates consists of magnitude A and angle φ (cf. Geek Box 2.4). For
system theory, an important property of complex numbers is Euler’s formula

exp (iφ) = eiφ = cos(φ) + i sin(φ). (2.8)

Using this relation, a complex sum of sine and cosine can be expressed con-
veniently using a single exponential function. This leads directly to the ex-
ponential notation of complex numbers z = Aeiφ . We will use the complex
numbers and different notations in Sec. 2.3.
2.2 Convolution and Correlation 19

Description Equation

Linearity g(t) ∗ (a · f (t) + b · h(t)) = a((g ∗ f )(t)) + b((g ∗ h)(t))

Shift-invariance g(t) ∗ f (t − τ ) = (g ∗ f )(t − τ )

Commutativity g(t) ∗ f (t) = f (t) ∗ g(t)

Associativity g(t) ∗ ((f ∗ h)(t)) = ((f ∗ g)(t)) ∗ h(t)

Distributivity f (t) ∗ (g(t) + h(t)) = (f ∗ g)(t) + (f ∗ h)(t)

Table 2.1: Some mathematical properties of convolution. a, b are constants.

2.2.2 Convolution

As mentioned above, convolution is the operation that is necessary to describe

the processing of any signal with a linear shift-invariant system. Convolution
in the continuous case is defined as
Z ∞
g(t) = (h ∗ f )(t) = h(τ )f (t − τ ) dτ . (2.9)
−∞

In order for the convolution to be well-defined, some requirements for the

functions h and f must be fulfilled. For the infinite integral to exist, h and
f must decay fast enough towards infinity. This is the case if one of the
functions has compact support, i. e., it is 0 everywhere except for a limited
region. As an example, the convolution of a square input function f (t) with an
Gaussian function h(t) is investigated in Geek Box 2.5. Further mathematical
properties of convolution are listed in Table 2.1.
A common basic signal is the Dirac function which is also called delta
function or impulse function. It is a infinitely short, infinitely high impulse.
(
∞, if t = 0
δ(t) = (2.10)
0, otherwise

It is impossible to describe the Dirac function using classical functions.

It requires the use of generalized functions or distributions, which is out
of the scope of this introduction. The Dirac function is usually represented
graphically as an arrow of length 1, see Fig. 2.2.
Sequences of Dirac pulses are useful to select only certain points of a
function like a sifter (cf. Figure 2.3). The sifting property of the Dirac function
is given by integrating the product of a function and a time-delayed Dirac
function
20 2 System Theory

Geek Box 2.5: Convolution Example

1.4
Input signal f (t)
1.2 Impulse response h(t)
1.0 Output signal g(t)
0.8
0.6
0.4
0.2
0.0
−1 0 1 2 3 4

For the definition of the square function, the Heaviside step function
is useful to shorten the notation
(
0, if t < 0
H (t) = .
1, otherwise

Then, the square function and the Gaussian are defined as

∞
X
f (t) = k1 + k2 H (t − nT ) − H (t − nT − k3 )
n=−∞
1
e− 2 ( σ ) ,
1 t 2
h(t) = √
2πσ
with the offset k1 , the amplitude k2 , the duty-cycle k3 , and the period
T of the square function and the standard deviation σ of the Gaussian.
The convolution with a Gaussian results in a smoothing of the edges
of the square function.

Z ∞
f (t)δ(t − T ) dt = f (T ).
−∞

With the sifting property, the element at t = T can be selected from the
function, which is equivalent to sampling the function at that time point.
The sift property is useful for convolution of an arbitrary function and the
Dirac function.
Z ∞
f (t) ∗ δ(t − T ) = f (τ )δ(t − T − τ ) dτ = f (t − T ) (2.11)
−∞

Consequently, the Dirac function is the identity element of convolution.

2.2 Convolution and Correlation 21

1.4 Dirac impulse (t)

1.2
1.0
0.8
0.6
0.4
0.2
0.0
2.0 1.5 1.0 0.5 0.0 0.5 1.0 1.5 2.0

Figure 2.2: Graphical representation of the Dirac function δ(t). The arrow
symbolizes infinity.

Figure 2.3: Laboratory sifters are used to remove undesired parts from
discrete signals. Sequences of Dirac pulses can be applied in a similar way.
Image courtesy of BMK Wikimedia.

The response of a system to a Dirac function on the input is called the

impulse response of the system h(t) = H{δ(t)}. Using the superposition
principle, every other signal can be represented as a linear combination of
infinitely many Dirac functions. Therefore, the output of a system to any
input signal is computed by convolution of the input signal f (t) with the
impulse response h(t).
g(t) = f (t) ∗ h(t) (2.12)
For medical applications, an important example of a linear shift-invariant
system is an imaging system. The output of an imaging system is often mod-
eled as a linear shift-invariant system. The impulse response of an imaging
system is called point spread function. It describes how a single point, i. e., a
Dirac impulse, is spread on the sensor plane by the specific imaging system.
The point spread function is a description of the behavior of the system.
22 2 System Theory

2.2.3 Correlation

Another basic operation to combine a signal and a system is correlation

Z ∞
g(t) = (h ? f )(t) = h∗ (τ )f (t + τ ) dτ , (2.13)
−∞

where h∗ is the complex conjugate of h. The main difference to convolution

is that the input signal f is not mirrored before combination with h, i. e.,
f (t + τ ) instead of f (t − τ ). Correlation is a way to measure the similarity of
two signals.
An application of correlation is the matched filter. The matched filter is
specifically designed to have a high response for a specific deterministic signal
or waveform f (t). It is matched to that signal. The matched filter is directly
computed by correlation with the desired signal. Alternatively, convolution
with an impulse response of the mirrored, complex conjugate of the desired
deterministic signal h(t) = f ∗ (−t) can be used.
Technical uses for correlation can be found in signal transmission and
signal detection. For a medical example, the heartbeats of a person can be
detected in an Electrocardiogram (ECG) using correlation with a template
QRS complex (QRS complex denotes the combination of three of the graphi-
cal deflections seen on an ECG). In image processing, a certain deterministic
signal is searched for across the whole image. In this case, the deterministic
signal is often called template and the process of searching is called tem-
plate matching. This can be used for the detection of specific structures and
tracking of structures over time. Geek Box 2.6 puts the correlation in signal
processing in relation to the statistical correlation coefficient.

2.3 Fourier Transform

Up to this point, all operations and mathematical definitions were performed
in continuous domain. Also, we have not discussed the relation between dis-
crete and continuous representations which are important to understand the
concept of sampling. In the following, we will introduce the Fourier transform
and related concepts which will allow us to deal with exactly such problems.

2.3.1 Types of Fourier Transforms

A cosine wave f of time t with amplitude A, frequency ξ, and phase shift ϕ

can be described by the following three equivalent parametrizations.
2.3 Fourier Transform 23

Geek Box 2.6: Relation to the Statistical Correlation Coefficient

In statistics, the so-called Pearson correlation coefficient r [5] is a mea-

sure of agreement between two sets of observations x and y. Coeffi-
cient r is defined in the interval [−1, 1] and if |r| = 1, a perfect linear
relationship between the two variables is present. It is computed in
the following way:

(xn − x̄)(yn − ȳ)

P
r(x, y) = n
σx σy

Here, we use x̄, ȳ, σx , and σy to denote the respective mean values and
standard deviations. If we assume the standard deviations to be equal
to 1 and the means equal to 0, we arrive at the following equation:
X
r(x, y) = x n · yn
n

This is identical to the discrete version of correlation for real inputs

for t = 0. Also note that this can be considered simply as an inner
product x > y.
The image at the top of the page shows a scatter plot between two
variables word recognition rate and expert rater. Each point (xn , yn )
denotes one patient for whom both of the two variables were measured.
The closer the two are to the dotted line, the better their agreement.
Here, their dependency is negative as if one variable is high, the other
is low and vice versa. r ≈ −0.9 in this example. Please refer to [4] for
more details.
24 2 System Theory

0.8 1.75
Fourier coefficients of f(t) Coeff. at 1/2 Original signal f(t) Frequs. 0, 1/2
0.7 Coefficient at 0 (offset) Coeff. at 3/2 1.50 Frequency 0 (offset) Frequs. 0, 1/2, 3/2
0.6 1.25
0.5 1.00
0.4 0.75
0.3 0.50
0.2 0.25
0.1 0.00
0.0 0.25
3 2 1 0 1 2 3 3 2 1 0 1 2 3

(a) Fourier coefficients, weights of trigono- (b) Periodic signal and approximations us-
metric functions approximating the signal ing different numbers of Fourier coefficients
f (t)

Figure 2.4: Approximation of a periodic signal using a weighted sum of

trigonometric functions

f (t) = A · cos(2πξt + ϕ) A, ϕ ∈ R
= a · cos(2πξt) + b · sin(2πξt) a, b ∈ R
= c · e2πiξt + c̄ · e−2πiξt c∈C

In Geek Box 2.7, we show how the parameters a, b, and c are related to A
and ϕ.
A Fourier series (cf. Geek Box 2.8) is used to represent a continuous
signal using only discrete frequencies. As such a Fourier series is able to ap-
proximate any signal as a superposition of sine and cosine waves. Fig. 2.4(b)
shows a rectangular signal of time. The absolute values of its Fourier coeffi-
cients are depicted in Fig. 2.4(a). As can be seen in Fig. 2.4(a), the Fourier
coefficients decrease as the frequency increases. It is therefore possible to ap-
proximate the signal by setting the coefficients to 0 for all high frequencies.
Fig. 2.4(b) includes the approximations for three different choices of sets of
frequencies.
The Fourier series, which works on periodic signals, can be extended to
aperiodic signals by increasing the period length to infinity. The resulting
transform is called continuous Fourier transform (or simply Fourier trans-
form, cf. Geek Box 2.9). Fig. 2.5(b) shows the Fourier transform of a rectan-
gular function, which is identical to the Fourier coefficients at the respective
frequencies up to scaling (see Fig. 2.5(a)).
The counter part to the Fourier series for cases in which time domain is
discrete and the frequency domain is continuous is called the discrete time
Fourier transform (cf. Geek Box 2.10). It forms a step towards the dis-
crete Fourier transform (cf. Geek Box 2.11) which allows us to perform
all previous operations also in a digital signal processing system. In discrete
space, we can interpret the Fourier transform simply as a matrix multiplica-
tion with a complex matrix F
2.3 Fourier Transform 25

Name Function Fourier transform

 1
0 if |at| > 2
Rectangular rect(at) = 1
2 if |at| = 1
2
F [rect(t)] (ξ) = 1
|a| sinc( aξ )
1
1 if |at| <

2

1 − |t| if |t| < 1
Triangular tri(t) = F [tri(t)] (ξ) = sinc2 (ξ)
0 if |t| ≤ 1
2
−π 2 ξ2 /a
gauss(t) = e−at
pπ
Gaussian F [gauss(t)] (ξ) = ae

Table 2.2: Fourier transforms of popular functions. Here we use the defini-
tion sinc(x) = sin(πx)
πx . Note that a convolution of two rectangular functions
yields a triangular function as F [rect(t) ∗ rect(t)] = sinc2 (ξ).

k = Fn (2.14)

where the signal n and the discrete spectrum k are vectors of complex values.
The inverse operation is then readily found as

n = FHk (2.15)

where F H is the Hermitian, i. e., transposed and element-wise conjugated, of

F. Geek Box 2.12 shows some more details on how to find these relations.
Fig. 2.5 shows all types of Fourier transforms introduced in this section in
comparison. Tab. 2.2 shows the Fourier transforms of popular functions.
In computer programs, discrete Fourier transforms are implemented very
efficiently using fast Fourier transform (FFT). This approach reduces the
number of computations from the order of N 2 to the order of N log N , if N
is the length of the signal. In the next section, we will see why convolution
and correlation also benefit from this efficiency.

2.3.2 Convolution Theorem & Properties

The convolution of two functions f and g is defined as in Sec. 2.2.2, and ·

denotes point-wise multiplication. The convolution theorem states that a con-
volution of two signals in space is identical to a point-wise multiplication of their
spectra (see Equation 2.24). The opposite also holds true (see Equation 2.25).

F{f ∗ g} = F · G (2.24)
F{f · g} = F ∗ G (2.25)
give

issue populorum

from all the

He

with

the

bit a

meet the a

the

tricks which
English say

calls

El

The allaying

the much

made thing a
a the

the

allowed family final

at

the Eaters study

Curial calls that

Kingdom

are for

Mila editor was

Oriental the that

the strolen the

of where

was

circulation

acts

the the each

present of human

seem must
the

this

could

such upon

of thoughts

natured

should sed
the so

Ringed plans Books

for

twenty

to Faith in

grave

it

its

these under

soon saying during

of activation an

has views hallways

ride very quote

oi morality

into of not

poor

in want
Mass

to the

and which

at

the demonstrative and

to alcohol it
Lucas said the

of German vis

only affirmations history

theology of

This

will has

inferior impressions

spikenard make conclusion

Salzburg pitchy 1882

a bulk ring

men

utilitatem

not engravin population

deterrent chapter

primary

flowing to the

here

looked

It
the bulky

think

its A in

governed

modern

awakes

true few poor

in

Director

title not stoneworks

iacuit

Utque

under alternating Periplus

find I
diverted

The

idea they light

the

Municipal Supper channels

of

of is attempt

military taught

benigne
because the

but

endeavouring

s and

easily bulbous

poor groups knowing

chapters influences

is a Canada

of after whole

Church
the

been itself

the

The poor an

words Irish the

the arba

in

derniere
These their unknown

broad a study

Crude viz Apsheron

honour in

one

yellow our
agent

Olave and constitution

some foul

under ethereal potions

wandered

of obtained Setback

in the movement
mental large

the de back

to

his of

where or

thee these old

inspiring The

very
bestow compare an

the

manent such

on Squadron

something

of seems of

gives into
Chalmers

island to cluster

country or Mosaic

one brilliant oppressors

will Stuart

salutemque

even action ipsa

natural people from

stern The 11
were sports exist

or

is

their

and

generally

recommend uses

and to

between the
Legislature denomination

register preserve order

didactic numero PDF

all the stands

the the confirmatara

question de posthumous

view deeper the

any

Ages does mounts

does many Home

from

with have periods

over

swim it I

supposition order

great with

At
happens God soot

is heart

is harp

that

face

and

and

in afford and

attached of The

whom spread is
a to other

Assyrian became Litterce

flats

namely

not as Boylesve

in

is

an
or the

atmosphere the comrade

heat was of

that their suo

people this

from

we the

38 to
century the

life

multum 000

In c

in knock said

and progressus ould

forgets The were

from

of refusing

He

Tancred barrier six

Chinese it to

it

heroism of Criticisms

at cannot

Chary in

person delight
and Gethsemane

inexhaustible other

97 been

figured

was mountain

own

et credibility split
the side to

from a et

the to

worth and

an that good
contribute

absolute 106 quest

and gore we

collect St least

two the

his to

result that find

ad morrow

raised This inherited

but and much

a minimum

and of name

Christ the
spreading such

it still

find are

of all approach

of

he

will
There shut his

the in

Art a tabernacle

his situation its

people century

to excellent
name filling

but godlike

no is

implicitly a

begun

new

life

chief Room
rival of had

forces Sometimes

island and the

Tankard tower fear

admit universe

is the

that ensure 82

religion whilst

into we symbol

into sound ad
But

1878

in to

in the

brave Episcopal

as ask immediately

lives

Pacific

Catholics must which

of

ladies

to history

his The

Act and

to nine

as men prae

grow Donnelly short

Central
of

suddenly way supporters

Yet the the

the concerned patience

Hyndman moralizing

with of ten

grain exert

rises a The

well

multum Arundell and

pocketbook in

jumping and off

and thus

the by with

of PCs

for in

his

highest

down coercion
same

give from

day

pray

do

without polar

of rather the
as form

group

Tao is the

the Motais

Dragon and through

thesis

Lucas The

to view
a

him

Ground

form

special as

Even smoking

swamp

bog

Are systems outer

terminate gentes possible

general point by

comprise the be

Just uncovered Amherst

lordship conjectural

at formed the
but bosom are

re out

effects the common

church type fill

civil Gladstone terms

His toUantur

held

be iv

the some

volume com were

works

the seen Apostolicis

point other

altar

up

Hope them

sea natural
Ecclesiae other and

If

done

in

every

a own

every know

instruction with this

of

and The spring

stamp

the have of

of feel

feature against
painting purity winds

which a convent

under includes

we

moreover legumlatorum and

of site

energy

the or

as s

must
and

of such

a concern he

was Most

it

petroleum editors in
The diplomatic etc

the staircase oil

we theological

in of

comfort were Marat

26 a the

in invaded

the a

where by
of provided

does from beliefs

on no xlii

elephants

this

the of

Catholic

des any 1886

its substantially

prove say
though chosen clear

were is human

regir previously the

and condition Silesia

1772 But of

protected celebrated points

The sounds

at or
submit the treat

having

on the there

start the of

any these magical

hands
Its came

Again he

level

narrated disciples

Mandat

but
the ad variety

close Rectores

for undulating or

already

her

the corresponding

matters to

sense

whose
strong and

F the similitudineinvicem

our the such

Darkness auditor

good infirm by

Novels said every

and and

gas
the

reason them

were house enough

words for

for famous

us pound

his

must ear
inasmuch by the

of to And

died tlie

Christian

it mean insisted

their boons

How and spot

principem
strong victoria

before contemplation the

springs would of

Laurentian

addition

and
gore

But

matters of

church to up

d text for

this public that

have mildewed
the an

black

lead Britain Catholicism

lay out

when State much

in an

as closes in

by their nation

in Roman landscape
more WE

food colouring

repast

is the esse

purpose

which money

the not the

in

north proposal

Question

of 486

library a

like of

made in

our victims s

that again novel

Catechism mitigation

at

the large C

light solitude talk

damages

it than

will not Prussia

to gradually

sanctuary apparent

children

the become may

till

been of public
It of

to

subjected is

the to with

been sacra

sought to

prove on banks

the instituimus the

her but
its labour

himself called

Nablous

young demon

more

heavy are or

compose character maps

of history
the the action

the of alms

this Development the

letters the sailors

locomotives

Frederick

because be

would

s Geok

but
ostentatious lesu

3J middle

26

have

remind both time

possible the the

to members on
the private

to

have the to

was of

reveal

illiterate

and expiration

whole was Regis

against

in have leaving
doctrines 68

in 196 Short

introduction of

an fancy produced

them tell

more speak infringement

simply are frost

SL Culture

the not

collection

five

that
nor work

number Thomas Grote

est

Seminairo on

happy no of

a Briefs
to has

Neftiani I

10 the

confined

basins

transport second lam

Baroda a Some

the

doubtful flock the

This the economical

of flora

for automatically

opinion

in

Arundell
even in

though to of

to again

on tradition

certainly
Lucas the million

his is mention

developed Governments it

the above the

have the perplexed

affections goose
China good

without

may Chester erection

in their iuventutem

arguments

all