-
Biosignal and Biomedical Image Processing MATLA B-Based
Applications
JOHN L. SEMMLOW Robert Wood Johnson Medical School New
Brunswick, New Jersey, U.S.A.
Rutgers University Piscataway, New Jersey, U.S.A.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
-
Although great care has been taken to provide accurate and
current information, neitherthe author(s) nor the publisher, nor
anyone else associated with this publication, shall beliable for
any loss, damage, or liability directly or indirectly caused or
alleged to becaused by this book. The material contained herein is
not intended to provide specificadvice or recommendations for any
specific situation.
Trademark notice: Product or corporate names may be trademarks
or registered trade-marks and are used only for identification and
explanation without intent to infringe.
Library of Congress Cataloging-in-Publication DataA catalog
record for this book is available from the Library of Congress.
ISBN: 08247-48034
This book is printed on acid-free paper.
HeadquartersMarcel Dekker, Inc., 270 Madison Avenue, New York,
NY 10016, U.S.A.tel: 212-696-9000; fax: 212-685-4540
Distribution and Customer ServiceMarcel Dekker, Inc., Cimarron
Road, Monticello, New York 12701, U.S.A.tel: 800-228-1160; fax:
845-796-1772
Eastern Hemisphere DistributionMarcel Dekker AG, Hutgasse 4,
Postfach 812, CH-4001 Basel, Switzerlandtel: 41-61-260-6300; fax:
41-61-260-6333
World Wide Webhttp://www.dekker.com
The publisher offers discounts on this book when ordered in bulk
quantities. For moreinformation, write to Special
Sales/Professional Marketing at the headquarters addressabove.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
Neither this book nor any part may be reproduced or transmitted
in any form or by anymeans, electronic or mechanical, including
photocopying, microfilming, and recording,or by any information
storage and retrieval system, without permission in writing fromthe
publisher.
Current printing (last digit):
10 9 8 7 6 5 4 3 2 1
PRINTED IN THE UNITED STATES OF AMERICA
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
http:/ /www.dekker.com
-
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
-
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
-
To Lawrence Stark, M.D., who has shown me the many possibilities
. . .
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
-
Series Introduction
Over the past 50 years, digital signal processing has evolved as
a major engi-neering discipline. The fields of signal processing
have grown from the originof fast Fourier transform and digital
filter design to statistical spectral analysisand array processing,
image, audio, and multimedia processing, and shaped de-velopments
in high-performance VLSI signal processor design. Indeed, thereare
few fields that enjoy so many applicationssignal processing is
everywherein our lives.
When one uses a cellular phone, the voice is compressed, coded,
andmodulated using signal processing techniques. As a cruise
missile winds alonghillsides searching for the target, the signal
processor is busy processing theimages taken along the way. When we
are watching a movie in HDTV, millionsof audio and video data are
being sent to our homes and received with unbeliev-able fidelity.
When scientists compare DNA samples, fast pattern
recognitiontechniques are being used. On and on, one can see the
impact of signal process-ing in almost every engineering and
scientific discipline.
Because of the immense importance of signal processing and the
fast-growing demands of business and industry, this series on
signal processingserves to report up-to-date developments and
advances in the field. The topicsof interest include but are not
limited to the following:
Signal theory and analysis Statistical signal processing Speech
and audio processing
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
-
Image and video processing Multimedia signal processing and
technology Signal processing for communications Signal processing
architectures and VLSI design
We hope this series will provide the interested audience with
high-quality,state-of-the-art signal processing literature through
research monographs, editedbooks, and rigorously written textbooks
by experts in their fields.
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
-
Preface
Signal processing can be broadly defined as the application of
analog or digitaltechniques to improve the utility of a data
stream. In biomedical engineeringapplications, improved utility
usually means the data provide better diagnosticinformation. Analog
techniques are applied to a data stream embodied as a time-varying
electrical signal while in the digital domain the data are
represented asan array of numbers. This array could be the digital
representation of a time-varying signal, or an image. This text
deals exclusively with signal processingof digital data, although
Chapter 1 briefly describes analog processes commonlyfound in
medical devices.
This text should be of interest to a broad spectrum of
engineers, but itis written specifically for biomedical engineers
(also known as bioengineers).Although the applications are
different, the signal processing methodology usedby biomedical
engineers is identical to that used by other engineers such
electri-cal and communications engineers. The major difference for
biomedical engi-neers is in the level of understanding required for
appropriate use of this technol-ogy. An electrical engineer may be
required to expand or modify signalprocessing tools, while for
biomedical engineers, signal processing techniquesare tools to be
used. For the biomedical engineer, a detailed understanding ofthe
underlying theory, while always of value, may not be essential.
Moreover,considering the broad range of knowledge required to be
effective in this field,encompassing both medical and engineering
domains, an in-depth understandingof all of the useful technology
is not realistic. It is important is to know what
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
-
tools are available, have a good understanding of what they do
(if not how theydo it), be aware of the most likely pitfalls and
misapplications, and know howto implement these tools given
available software packages. The basic conceptof this text is that,
just as the cardiologist can benefit from an
oscilloscope-typedisplay of the ECG without a deep understanding of
electronics, so a biomedicalengineer can benefit from advanced
signal processing tools without always un-derstanding the details
of the underlying mathematics.
As a reflection of this philosophy, most of the concepts covered
in thistext are presented in two sections. The first part provides
a broad, general under-standing of the approach sufficient to allow
intelligent application of the con-cepts. The second part describes
how these tools can be implemented and reliesprimarily on the
MATLAB software package and several of its toolboxes.
This text is written for a single-semester course combining
signal andimage processing. Classroom experience using notes from
this text indicatesthat this ambitious objective is possible for
most graduate formats, althougheliminating a few topics may be
desirable. For example, some of the introduc-tory or basic material
covered in Chapters 1 and 2 could be skipped or treatedlightly for
students with the appropriate prerequisites. In addition, topics
suchas advanced spectral methods (Chapter 5), time-frequency
analysis (Chapter 6),wavelets (Chapter 7), advanced filters
(Chapter 8), and multivariate analysis(Chapter 9) are pedagogically
independent and can be covered as desired with-out affecting the
other material.
Although much of the material covered here will be new to most
students,the book is not intended as an introductory text since the
goal is to provide aworking knowledge of the topics presented
without the need for additionalcourse work. The challenge of
covering a broad range of topics at a useful,working depth is
motivated by current trends in biomedical engineering educa-tion,
particularly at the graduate level where a comprehensive education
mustbe attained with a minimum number of courses. This has led to
the developmentof core courses to be taken by all students. This
text was written for just sucha core course in the Graduate Program
of Biomedical Engineering at RutgersUniversity. It is also quite
suitable for an upper-level undergraduate course andwould be of
value for students in other disciplines who would benefit from
aworking knowledge of signal and image processing.
It would not be possible to cover such a broad spectrum of
material to adepth that enables productive application without
heavy reliance on MATLAB-based examples and problems. In this
regard, the text assumes the studenthas some knowledge of MATLAB
programming and has available the basicMATLAB software package
including the Signal Processing and Image Process-ing Toolboxes.
(MATLAB also produces a Wavelet Toolbox, but the section onwavelets
is written so as not to require this toolbox, primarily to keep the
num-ber of required toolboxes to a minimum.) The problems are an
essential part of
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
-
this text and often provide a discovery-like experience
regarding the associatedtopic. A few peripheral topics are
introduced only though the problems. Thecode used for all examples
is provided in the CD accompanying this text. Sincemany of the
problems are extensions or modifications of examples given in
thechapter, some of the coding time can be reduced by starting with
the code of arelated example. The CD also includes support routines
and data files used inthe examples and problems. Finally, the CD
contains the code used to generatemany of the figures. For
instructors, there is a CD available that contains theproblem
solutions and Powerpoint presentations from each of the
chapters.These presentations include figures, equations, and text
slides related to chapter.Presentations can be modified by the
instructor as desired.
In addition to heavy reliance on MATLAB problems and examples,
thistext makes extensive use of simulated data. Except for the
section on imageprocessing, examples involving biological signals
are rarely used. In my view,examples using biological signals
provide motivation, but they are not generallyvery instructive.
Given the wide range of material to be presented at a workingdepth,
emphasis is placed on learning the tools of signal processing;
motivationis left to the reader (or the instructor).
Organization of the text is straightforward. Chapters 1 through
4 are fairlybasic. Chapter 1 covers topics related to analog signal
processing and data acqui-sition while Chapter 2 includes topics
that are basic to all aspects of signal andimage processing.
Chapters 3 and 4 cover classical spectral analysis and basicdigital
filtering, topics fundamental to any signal processing course.
Advancedspectral methods, covered in Chapter 5, are important due
to their widespreaduse in biomedical engineering. Chapter 6 and the
first part of Chapter 7 covertopics related to spectral analysis
when the signals spectrum is varying in time,a condition often
found in biological signals. Chapter 7 also covers both contin-uous
and discrete wavelets, another popular technique used in the
analysis ofbiomedical signals. Chapters 8 and 9 feature advanced
topics. In Chapter 8,optimal and adaptive filters are covered, the
latters inclusion is also motivatedby the time-varying nature of
many biological signals. Chapter 9 introducesmultivariate
techniques, specifically principal component analysis and
indepen-dent component analysis, two analysis approaches that are
experiencing rapidgrowth with regard to biomedical applications.
The last four chapters coverimage processing, with the first of
these, Chapter 10, covering the conventionsused by MATLABs Imaging
Processing Toolbox. Image processing is a vastarea and the material
covered here is limited primarily to areas associated withmedical
imaging: image acquisition (Chapter 13); image filtering,
enhancement,and transformation (Chapter 11); and segmentation, and
registration (Chapter 12).
Many of the chapters cover topics that can be adequately covered
only ina book dedicated solely to these topics. In this sense,
every chapter representsa serious compromise with respect to
comprehensive coverage of the associated
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
-
topics. My only excuse for any omissions is that classroom
experience with thisapproach seems to work: students end up with a
working knowledge of a vastarray of signal and image processing
tools. A few of the classic or major bookson these topics are cited
in an Annotated bibliography at the end of the book.No effort has
been made to construct an extensive bibliography or reference
listsince more current lists would be readily available on the
Web.
TEXTBOOK PROTOCOLS
In most early examples that feature MATLAB code, the code is
presented infull, while in the later examples some of the routine
code (such as for plotting,display, and labeling operation) is
omitted. Nevertheless, I recommend that stu-dents carefully label
(and scale when appropriate) all graphs done in the prob-lems. Some
effort has been made to use consistent notation as described
inTable 1. In general, lower-case letters n and k are used as data
subscripts, andcapital letters, N and K are used to indicate the
length (or maximum subscriptvalue) of a data set. In
two-dimensional data sets, lower-case letters m and nare used to
indicate the row and column subscripts of an array, while
capitalletters M and N are used to indicate vertical and horizontal
dimensions, respec-tively. The letter m is also used as the index
of a variable produced by a transfor-mation, or as an index
indicating a particular member of a family of relatedfunctions.*
While it is common to use brackets to enclose subscripts of
discretevariables (i.e., x[n]), ordinary parentheses are used here.
Brackets are reservedto indicate vectors (i.e., [x1, x2, x3 , . . .
]) following MATLAB convention.Other notation follows standard
conventions.
Italics () are used to introduce important new terms that should
be incor-porated into the readers vocabulary. If the meaning of
these terms is not obvi-ous from their use, they are explained
where they are introduced. All MATLABcommands, routines, variables,
and code are shown in the Courier typeface.Single quotes are used
to highlight MATLAB filenames or string variables.Textbook
protocols are summarized in Table 1.
I wish to thank Susanne Oldham who managed to edit this book,
andprovided strong, continuing encouragement and support. I would
also like toacknowledge the patience and support of Peggy Christ
and Lynn Hutchings.Professor Shankar Muthu Krishnan of Singapore
provided a very thoughtfulcritique of the manuscript which led to
significant improvements. Finally, Ithank my students who provided
suggestions and whose enthusiasm for thematerial provided much
needed motivation.
*For example, m would be used to indicate the harmonic number of
a family of harmonically relatedsine functions; i.e., fm(t) = sin
(2 m t).
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
-
TABLE 1 Textbook Conventions
Symbol Description/General usage
x(t), y(t) General functions of time, usually a waveform or
signalk, n Data indices, particularly for digitized time dataK, N
Maximum index or size of a data setx(n), y(n) Waveform variable,
usually digitized time variables (i.e., a dis-
creet variable)m Index of variable produced by transformation,
or the index of
specifying the member number of a family of functions
(i.e.,fm(t))
X(f), Y(f) Frequency representation (complex) of a time
functionX(m), Y(m) Frequency representation (complex) of a discreet
variableh(t) Impulse response of a linear systemh(n) Discrete
impulse response of a linear systemb(n) Digital filter coefficients
representing the numerator of the dis-
creet Transfer Function; hence the same as the impulse
re-sponse
a(n) Digital filter coefficients representing the denominator of
the dis-creet Transfer Function
Courier font MATLAB command, variable, routine, or
program.Courier font MATLAB filename or string variable
John L. Semmlow
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
-
Contents
Preface
1 Introduction
Typical Measurement SystemsTransducers
Further Study: The TransducerAnalog Signal ProcessingSources of
Variability: Noise
Electronic NoiseSignal-to-Noise Ratio
Analog Filters: Filter BasicsFilter TypesFilter BandwidthFilter
OrderFilter Initial Sharpness
Analog-to-Digital Conversion: Basic ConceptsAnalog-to-Digital
Conversion Techniques
Quantization ErrorFurther Study: Successive Approximation
Time Sampling: BasicsFurther Study: Buffering and Real-Time Data
Processing
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
-
Data BanksProblems
2 Basic Concepts
NoiseEnsemble AveragingMATLAB Implementation
Data Functions and TransformsConvolution, Correlation, and
Covariance
Convolution and the Impulse ResponseCovariance and
CorrelationMATLAB Implementation
Sampling Theory and Finite Data ConsiderationsEdge Effects
Problems
3 Spectral Analysis: Classical Methods
IntroductionThe Fourier Transform: Fourier Series Analysis
Periodic FunctionsSymmetry
Discrete Time Fourier AnalysisAperiodic Functions
Frequency ResolutionTruncated Fourier Analysis: Data
WindowingPower Spectrum
MATLAB ImplementationDirect FFT and WindowingThe Welch Method
for Power Spectral Density DeterminationWidow Functions
Problems
4 Digital Filters
The Z-TransformDigital Transfer FunctionMATLAB
Implementation
Finite Impulse Response (FIR) FiltersFIR Filter Design
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
-
Derivative Operation: The Two-Point Central
DifferenceAlgorithm
MATLAB ImplementationInfinite Impulse Response (IIR)
FiltersFilter Design and Application Using the MATLAB Signal
Processing ToolboxFIR Filters
Two-Stage FIR Filter DesignThree-Stage Filter Design
IIR FiltersTwo-Stage IIR Filter DesignThree-Stage IIR Filter
Design: Analog Style Filters
Problems
5 Spectral Analysis: Modern Techniques
Parametric Model-Based MethodsMATLAB Implementation
Non-Parametric Eigenanalysis Frequency EstimationMATLAB
Implementation
Problems
6 TimeFrequency Methods
Basic ApproachesShort-Term Fourier Transform: The
SpectrogramWigner-Ville Distribution: A Special Case of Cohens
ClassChoi-Williams and Other Distributions
Analytic SignalMATLAB Implementation
The Short-Term Fourier TransformWigner-Ville
DistributionChoi-Williams and Other Distributions
Problems
7 The Wavelet Transform
IntroductionThe Continuous Wavelet Transform
Wavelet TimeFrequency CharacteristicsMATLAB Implementation
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
-
The Discrete Wavelet TransformFilter Banks
The Relationship Between Analytical Expressions andFilter
Banks
MATLAB ImplementationDenoisingDiscontinuity DetectionFeature
Detection: Wavelet Packets
Problems
8 Advanced Signal Processing Techniques:Optimal and Adaptive
Filters
Optimal Signal Processing: Wiener FiltersMATLAB
Implementation
Adaptive Signal ProcessingAdaptive Noise CancellationMATLAB
Implementation
Phase Sensitive DetectionAM ModulationPhase Sensitive
DetectorsMATLAB Implementation
Problems
9 Multivariate Analyses: Principal Component Analysisand
Independent Component Analysis
IntroductionPrincipal Component Analysis
Order SelectionMATLAB Implementation
Data RotationPrincipal Component Analysis Evaluation
Independent Component AnalysisMATLAB Implementation
Problems
10 Fundamentals of Image Processing: MATLAB ImageProcessing
Toolbox
Image Processing Basics: MATLAB Image FormatsGeneral Image
Formats: Image Array Indexing
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
-
Data Classes: Intensity Coding SchemesData FormatsData
ConversionsImage DisplayImage Storage and RetrievalBasic Arithmetic
Operations
Advanced Protocols: Block ProcessingSliding Neighborhood
OperationsDistinct Block Operations
Problems
11 Image Processing: Filters, Transformations,and
Registration
Spectral Analysis: The Fourier TransformMATLAB
Implementation
Linear FilteringMATLAB Implementation
Filter DesignSpatial Transformations
MATLAB ImplementationAffine TransformationsGeneral Affine
TransformationsProjective Transformations
Image RegistrationUnaided Image RegistrationInteractive Image
Registration
Problems
12 Image Segmentation
Pixel-Based MethodsThreshold Level AdjustmentMATLAB
Implementation
Continuity-Based MethodsMATLAB Implementation
Multi-ThresholdingMorphological Operations
MATLAB ImplementationEdge-Based Segmentation
MATLAB ImplementationProblems
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
-
13 Image Reconstruction
CT, PET, and SPECTFan Beam GeometryMATLAB Implementation
Radon TransformInverse Radon Transform: Parallel Beam
GeometryRadon and Inverse Radon Transform: Fan Beam Geometry
Magnetic Resonance ImagingBasic PrinciplesData Acquisition:
Pulse SequencesFunctional MRIMATLAB ImplementationPrincipal
Component and Independent Component Analysis
Problems
Annotated Bibliography
Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.
BIOSIGNAL AND BIOMEDICAL IMAGE PROCESSINGSERIES
INTRODUCTIONPREFACECONTENTSCHAPTER 1: INTRODUCTIONCHAPTER 2: BASIC
CONCEPTSCHAPTER 3: SPECTRAL ANALYSIS: CLASSICAL METHODSCHAPTER 4:
DIGITAL FILTERSCHAPTER 5: SPECTRAL ANALYSIS: MODERN
TECHNIQUESCHAPTER 6: TIMEFREQUENCY METHODSCHAPTER 7: THE WAVELET
TRANSFORMCHAPTER 8: ADVANCED SIGNAL PROCESSING TECHNIQUES: OPTIMAL
AND ADAPTIVE FILTERSCHAPTER 9: MULTIVARIATE ANALYSES: PRINCIPAL
COMPONENT ANALYSIS AND INDEPENDENT COMPONENT ANALYSISCHAPTER 10:
FUNDAMENTALS OF IMAGE PROCESSING: MATLAB IMAGE PROCESSING
TOOLBOXCHAPTER 11: IMAGE PROCESSING: FILTERS, TRANSFORMATIONS, AND
REGISTRATIONCHAPTER 12: IMAGE SEGMENTATIONCHAPTER 13: IMAGE
RECONSTRUCTIONANNOTATED BIBLIOGRAPHY