This book is the result of collective endeavor from several noted engineering and computer scientists, mathematicians, medical doctors, physicists, and radiologists. The editors are indebted to all of their efforts and outstanding scientific contributions. The editors are particularly grateful to Drs. Petia Reveda, Alex Falco, Andrew Laine, David Breen, David Chopp, C. C. Lu, Gary Christensen, Dirk Vandermeulen, Aly Farag, Alejandro Frangi, Gilson Antonio Giraldi, Gabor Szekely, Pierre Hellier, Gabor Herman, Ardeshir Coshtasby, Jan Kybic, Jeff Weiss, Jean-Claude Klein, Majid Mirmehdi, Maria Kallergi, Yangming Zhu, Sunanda Mitra, Sameer Singh, Alessandro Sarti, Xioping Shen, Calvin R. Maurer, Jr., Yoshi-nobu Sato, Koon-Pong Wong, Avdhesh Sharma, Rakesh Sharma, and Chun Yuan and their team members for working with us so closely in meeting all of the deadlines of the book. We would like to express our appreciation to Kluwer Publishers for helping create this invitational handbook. We are particularly thankful to Aaron Johnson, the acquisition editor and Shoshana Sternlicht for their excellent coordination of the book at every stage.

Dr. Suri thanks Philips Medical Systems, Inc., for the MR datasets and encouragements during his experiments and research. Special thanks are due to Dr. Larry Kasuboski and Dr. Elaine Keeler from Philips Medical Systems, Inc., for their support and motivations. Thanks are also due to my past Ph.D. committee research professors, particularly Professors Linda Shapiro, Robert M. Haralick, Dean Lytle, and Arun Somani, for their encouragements.

We extend our appreciations to Drs. Ajit Singh, Siemens Medical Systems, George Thoma, chief, Imaging Science Division, National Institutes of Health, Dr. Sameer Singh, University of Exeter, UK, for his motivations.

Special thanks go to the book series editor, Professor Evangelia Micheli Tzanakou for advising us on all aspects of the book.

We thank the IEEE Press, Academic Press, Springer-Verlag Publishers, and several medical and engineering journals for permitting us to use some of the images previously published in these journals.

Finally, Jasjit Suri thanks his wife Malvika Suri for all the love and support she has showered over the years and to our baby Harman whose presence is always a constant source of pride and joy. I also express my gratitude to my father, a mathematician, who inspired me throughout my life and career, and to my late mother, who most unfortunately passed away a few days before my Ph.D. graduation, and who so much wanted to see me write this book. Special thanks to Pom Chadha and his family, who taught me life is not just books. He is my of my best friends. I would like to also thank my in-laws who have a special place for me in their hearts and have shown lots of love and care for me.

David Wilson acknowledges the support of the Department of Biomedical Engineering, Case Western Reserve University, in this endeavor. Special thanks are due to the many colleagues and students who make research in biomedical engineering an exciting, wondrous endeavor.

Swamy Laxminarayan expresses his loving acknowledgments to his wife Marijke and to his kids, Malini and Vinod, for always giving the strength of mind amidst all lifes frustrations. The book kindies fondest memories of my late parents who made many personal sacrifices that helped shape our careers and the support of my family members who were always there for me when I needed them most. I have shared many ideas and thoughts on the book with numerous of my friends and colleagues in the discipline. I acknowledge their friendship, feedbacks, and discussions with particular thanks to Professor David Kristol of the New Jersey Institute of Technology. Peter Brett of Ashton University, Ewart Carson of the City University, London, Laura Roa of the University of Sevilla in Spain, and Jean Louis Coatrieux of the University of Rennes in France for their constant support over the past two decades.


In Chapter 1 we present in detail a framework for fully automated brain tissue classification. The framework consists of a sequence of fully automated state of the art image registration (both rigid and nonrigid) and image segmentation algorithms. Models of the spatial distribution of brain tissues are combined with models of expected tissue intensities, including correction of MR bias fields and estimation of partial voluming. We also demonstrate how this framework can be applied in the presence of lesions.

Chapter 2 presents the intravascular ultrasound (IVUS), which is a tomographic imaging technique that has provided unique tool for observation and supervision of vessel structures and exact vascular dimensions. In this way, it has contributed to the better understanding of the coronary content and processes: vascular remodelling, plaque morphology, and evolution, etc. Most investigators are convinced that the best way to detect plaque ruptures is by IVUS sequences. At the same time, cardiologists confirm that due to the "speckle nature" of IVUS images, conventional IVUS imaging is difficult to clearly diagnose potentially vulnerable plaques due to the image resolution, lack of contours, speckle motion, etc. Advanced automatic classification techniques can significantly help the physicians take decisions about different classes of tissue morphology. The characterization of tissue and plaque involves different problems. Image feature space determines the reliable descriptions that should be sufficiently expressive to capture differences between different classes but at the same time should not increase unnecessarily the complexity of the classification problem. We consider and compare a wide set of different feature spaces (Gabor filters, DOG filters, cooccurrence matrices, binary local patterns, etc). In particular, we show that the binary local patterns represent an optimal description of ultrasound regions that at the same time allow real-time processing of images. After reviewing the IVUS classification works available in the bibliography, we present a comparison between classical and advanced classification techniques (principal component analysis, linear discriminant analysis, nonparametric discriminant analysis, Kernel principal component analysis, Kernel fisher analysis, etc.). The classification "goodness" of IVUS regions can be significantly improved by applying multiple classifiers (boosting, adaboost, etc.). The result of the classification techniques represents a map of classified pixels that still need to be organized in regions. The technique of snakes (deformable models) is a convenient way to organize regions of pixels with similar characteristics. Incorporating the classification map or the likelihood map into the snake framework, allows to organize pixels into compact image regions representing different plaque zones of IVUS images.

Chapter 3 is dedicated to functional imaging techniques. The last few decades of the twentieth century have witnessed significant advances in multidimensional medical imaging, which enabled us to view noninvasively, the anatomic structure of internal organs with unprecedented precision and to recognize any gross pathology of organs and diseases without the need to "open" the body. This marked a new era of medical diagnostics with many invasive and potentially morbid procedures being substituted by noninvasive cross-sectional imaging. Continuing advances in instrumentation and computer technologies also accelerated the development of various multidimensional imaging modalities that possess a great potential for providing, in addition to structural information, dynamic, and functional information on biochemical and pathophysiologic processes or organs of the human body. There is no doubt that substantial progress has been achieved in delivering health care more efficiently and in improving disease management, and that diagnostic imaging techniques have played a decisive role in routine clinical practice in almost all disciplines of contemporary medicine. With further development of functional imaging techniques, in conjunction with continuing progress in molecular biology and functional genomics, it is anticipated that we will be able to visualize and determine the actual molecular errors in a specific disease very soon, and be able to incorporate this biological information into clinical management of that particular group of patients. This is definitely not achievable with the use of structural imaging techniques. In this chapter, we will take a quick tour of a functional imaging technique called positron emission tomography (PET), which is a primer biologic imaging tool being able to provide in vivo quantitative functional information in most organ systems of the body. An overview of this imaging technique, including the basic principles and instrumentation, methods of image reconstruction from projections, some specific correction factors necessary to achieve quantitative images are presented. Basic assumptions and special requirements for quantitation are briefly discussed. Quantitative analysis techniques based on the framework of tracer kinetic modeling for absolute quantification of physiological parameters of interest are also introduced in this chapter.

Pancreatic cancer is a difficult to diagnose and lethal disease. In Chapter 4, we present the Helical computed tomography (CT), which is currently the imaging modality of choice for the detection, diagnosis, and evaluation of pancreatic tumors. Despite major technological advances, helical CT imaging still presents imaging limitations as well as significant challenges in the interpretation process. Computer methodologies could assist radiologists and oncologists in the interpretation of CT scans and improve the diagnosis and management of the patients with pancreatic cancer. However, few computer aided detection (CADetection) or diagnosis (CADiagnosis) techniques have been developed for pancreatic cancer and this area remains seriously understudied and unexplored. This chapter aims at introducing the problem of pancreatic cancer and the limitations of currently available imaging techniques with specific emphasis on helical CT. It also presents a novel CADiagnosis scheme for pancreatic tumor segmentation that is based on supervised or unsupervised fuzzy clustering techniques. The proposed algorithm aims at improving pancreatic tumor diagnosis and assessment of treatment effects by automatically segmenting the areas of the pancreas and associated tumor(s) from neighboring organs in CT slices as well as by classifying normal from abnormal pancreatic areas. Preliminary results from a pilot study of the proposed algorithm are presented and discussed including issues of segmentation validation and analysis that are critical to these types of CADiagnosis applications.

Chapter 5 presents the research in the area of flow-mediated dilation (FMD) that offers a mechanism to characterize endothelial function and therefore may play a role in the diagnosis of cardiovascular diseases. Computerized analysis techniques are very desirable to give accuracy and objectivity to the measurements. Virtually all methods proposed up to now to measure FMD rely on accurate edge detection of the arterial wall, and they are not always robust in the presence of poor image quality or image artifacts. A novel method for automatic dilation assessment based on a global image analysis strategy is presented. We model interframe arterial dilation as a superposition of a rigid motion model and a scaling factor perpendicular to the artery. Rigid motion can be interpreted as a global compensation for patient and probe movements, an aspect that has not been sufficiently studied before. The scaling factor explains arterial dilation. The ultrasound (US) sequence is analyzed in two phases using image registration to recover both transformation models. Temporal continuity in the registration parameters along the sequence is enforced with a Kalman filter since the dilation process is known to be a gradual physiological phenomenon. Comparing automated and gold standard measurements we found a negligible bias (0.04(1.14measurements (bias = 0.47 better reproducibility (CV = 0.46

In Chapter 6 we present the assessment of onset and progression of diseases from images of various modalities is critically dependent on identification of lesions or changes in structures and regions of interest. Mathematical modeling of such discrimination among regions as well as identification of changes in anatomical structures in an image result from the process of segmentation. For clinical applications of segmentation, a compromise between the accuracy and computational speed of segmentation techniques is needed. Optimal segmentation processes based on statistical and adaptive approaches and their applicability to clinical settings have been addressed using diverse modalities of images. Current drawbacks of automated segmentation methodologies stem mostly from nonuniform illumination, inhomogeneous structures, and the presence of noise in acquired images. The effect of preprocessing on the accuracy of segmentation has been discussed. The superior performance of advanced clustering algorithms based on statistical and adaptive approaches over traditional algorithms in medical image segmentation has been presented.

Chapter 7 presents the automatic analysis of color fundus images and with its application to the diagnosis of diabetic retinopathy, a severe and frequent eye disease. We give an overview of computer assistance in this domain and describe in detail some algorithms developed within this framework: the detection of main features in the human eye (vascular tree, the optic disc, and the macula) and the detection of retinal lesions like microaneurysms and hard exudates.

Chapter 8 presents the advanced atherosclerotic plaque that can lead to diseases, such as vessel lumen stenosis, thrombosis, and embolization, which are the leading causes of death and major disability among adults in the United

States. Previous studies have shown that plaque constituents are important determinants for plaque vulnerability and stenosis risk access. To identify and quantitatively measure the composition of atherosclerotic lesions in carotid arteries, plaque segmentation techniques will be discussed in this chapter. First, to extract the lumen contour and outer wall boundary of carotid artery accurately, we will discuss Active Contour Based boundary detection methods, including how to convert exerting energy design and searching process optimizations. Second part is about region-based image segmentation technique, such as Markov random fields, and its applications on image sequence processing. In recent study, plaque components identification with multiple contrast weightings MR images has shown more promising results than single contrast weightings images. In third part, we will introduce multiple contrast weighting MR image segmentation methods and its validation results by comparing with histology images. At last, a software package developed specifically for the quantitative analysis of atherosclerotic plaque by MRI, quantitative vascular analysis system (QVAS) will be presented.

Chapter 9 presents the pre- and postcontrast Gd-DTPA1 MR images of any body organ hold diagnostic utility in the area of medicine, particularly for breast lesion characterization. This paper reviews the state-of-the-art tools and techniques for lesion characterization, such as uptake curve estimation (functional segmentation), image subtraction, velocity thresholding, differential characteristics of lesions, such as maximum derivative of image sequence, steep slope and washout, fuzzy clustering, Markov random fields, and interactive deformable models such as Live-Wire. In first part of the paper, we discuss the MRI system and breast coils along with the MR breast data acquisition protocol for spatial and temporal MR data collection. Then the perfusion analysis tools are discussed for staging breast tumors. Here, the rate of absorption of the contrast agent (Gad) is used to stage abreast lesion. The differences in contrast enhancement have been shown to be able to help differentiate between benign and malignant lesions. Thus, oncologists, radiologists, and internists have shown great interest in such classification by examining the quantitative characteristics of the tissue signal enhancement. Then we discuss two other major tools for breast lesion characterization. The first set of tools is based on pixel-classification algorithms and second set of tools is based on user-based deformable models such as Live-Wire.

1 Gadolinium-Dithylene-THamine-Penta-Acetate, we will refer to it as Gad from now on.

Finally, the paper also presents the user-friendly Marconi Medical System's realtime MR Breast Perfusion2 Software Analysis System (BPAS), based on Motif using C/C++ and X window libraries that runs on Digital Unix and XP1000 workstations supporting Unix and Linux Operating Systems, respectively. This software was tested on 20 patient studies from the data collected from two major sites in the United States and Europe.

Chapter 10 presents methods for the enhancement and segmentation of 3D local structures, that is, line-like shapes such as blood vessels, sheet-like shapes such as articular cartilage, and blob-like shapes such as nodules in medical volume data. Firstly, a method for enhancement of 3D local structures with various widths is presented. Multiscale Gaussian filters and the eigenvalues of Hessian matrix of the volume function are combined to effectively enhance various widths of structures. The characteristics of multiscale filter responses are analysed to clarify the guidelines for the filter design. Secondly, methods for description and quantification of the 3D local structures are presented. Medial axis/surface elements are locally determined based on the second-order approximations of local intensity structures. Diameter/thickness quantification is performed based on detected medial axis/surface elements. Limits on the accuracy of thickness quantification from 3D MR data is analyzed based on a mathematical models of imaged structures, MR imaging and thickness measurement processes. The utility of the methods is demonstrated by examples using 3D CT and MR data of various parts of the body.

Chapter 11 presents work in the area of CAD design. Research into the computer-aided detection (CAD) of breast lesions from digitised mammograms has been extensive over the past 15 years. The large number of computer algorithms for mammogram contrast enhancement, segmentation, and region discrimination reflects the nontrivial nature in the problem of detecting cancer during breast screening. In addition, due to the enormous variability in the mammographic appearance of the breast, engineering a single CAD solution is formidable. Mammographic CAD must provide a high level of sensitivity for the detection of breast lesions, while maintaining a low number of false-positive regions for each image. This chapter describes an adaptive knowledge-based framework for the detection of breast cancer masses from digitised mammo-grams. The proposed framework accommodates a set of distinct contrast

2 Or Breast Uptake.

enhancement and image segmentation experts, used to learn an optimal pipeline of image processing operators for an individual mammogram. It is hypothesised that such an optimal flow will lead to an increase in the sensitivity in the detection of breast lesions. To facilitate efficient training of the adaptive knowledge-based model, a novel method of grouping mammograms on the basis of their mammo-graphic density is proposed. In addition, an automated mechanism for improving the quality of expert radiologist's lesion definitions is presented. To validate this work, 400 digitised mammograms are taken from the publicly available Digital Database of Screening Mammograms (DDSM). The 400 mammograms comprise 200 abnormal and 200 abnormal images complete with lesion ground-truth definitions provided by an expert radiologist. Following the evaluation of the knowledge-based framework in the contrast enhancement and segmentation of mammograms, it is shown that each knowledge-based component out-performs the single best performing expert. Following image segmentation and region prefiltering, a sensitivity of 0.81 with on average 8.65 false-positives per mam-mogram is reported.

A semiautomatic region growing algorithm, which employs the concept of fuzzy connectedness to perform the simultaneous segmentation of elements of an arbitrary set, is presented in Chapter 12. Because of its general nature, this algorithm can be applied to segmenting dots in the plane, pixels of an image or voxels of a three-dimensional volume. The algorithm is supplied some minimal input by the user and produces an M-segmentation that identifies a grade of membership of every element of the set in each of M objects. The algorithm is illustrated on both mathematically described images and on MR and CT reconstructions.

Chapter 13 presents research in the area of computer-aided diagnosis (CA-Diagnosis) techniques in medical imaging play the role of a "second opinion" and their goal is to assist the observer in the differenciation between benign and malignant image findings and lesions. CADiagnosis methodologies go beyond the task of automated detection of abnormalities in images and aim at predicting biopsy outcomes from image and/or patient characteristics. In mam-mography today, CADiagnosis is implemented for masses and calcifications and its output may be a binary one (benign or malignant assignment) or a likelihood (percentage) for a finding to be benign or malignant. This chapter presents a CADiagnosis algorithm developed for assigning a likelihood of malignancy to calcification clusters detected in mammograms. The algorithm is composed of several modules including image segmentation and classification steps that are based on multiresolution methods and artificial neural networks. Morphology, distribution, and demographics are the domains where features are determined from for the classification task. As a result, the segmentation and feature selection processes of the algorithm are critical in its performance and are the areas we focus on in this chapter. We look particularly into the segmentation aspects of the implementation and the impact of multiresolution filtering has on feature estimation and classification. General aspects of algorithm evaluation and, particularly, segmentation validation are presented using results of experiments conducted for the evaluation of our CADiagnosis scheme as the basis of discussion.

Chapter 14 presents research in the area of neuro segmentation.


1. Model-Based Brain Tissue Classification 1

Koen Van Leemput, Dirk Vandermeulen, Frederik Maes,

Siddharth Srivastava, Emiliano D'Agostino, and Paul Suetens

2. Supervised Texture Classification for Intravascular

Tissue Characterization 57

Oriol Pujol and Petia Radeva

3. Medical Image Segmentation: Methods and Applications in Functional Imaging 111

Koon-Pong Wong

4. Automatic Segmentation of Pancreatic Tumors in Computed Tomography 183 Maria Kallergi, Marla R. Hersh, and Anand Manohar

5. Computerized Analysis and Vasodilation Parameterization in Flow-Mediated Dilation Tests from Ultrasonic Image Sequences 229

Alejandro F. Frangi, Martin Laclaustra, and Jian Yang

6. Statistical and Adaptive Approaches for Optimal Segmentation in Medical Images 267

Shuyu Yang and Sunanda Mitra

7. Automatic Analysis of Color Fundus Photographs and Its Application to the Diagnosis of Diabetic Retinopathy 315

Thomas Walter and Jean-Claude Klein

8. Segmentation Issues in Carotid Artery Atherosclerotic Plaque Analysis with MRI 369

Dongxiang Xu, Niranjan Balu, William S. Kerwin, and Chun Yuan

9. Accurate Lumen Identification, Detection, and Quantification in MR

Plaque Volumes 451

Jasjit Suri, Vasanth Pappu, Olivier Salvado, Baowei Fei,

Swamy Laxminarayan, Shaoxiong Zhang, Jonathan Lewin,

Jeffrey Duerk, and David Wilson

10. Hessian-Based Multiscale Enhancement, Description, and Quantification of Second-Order 3-D Local Structures from Medical Volume Data 531

Yoshinobu Sato

11. A Knowledge-Based Scheme for Digital Mammography 591

Sameer Singh and Keir Bovis

12. Simultaneous Fuzzy Segmentation of Medical Images 661

Gabor T. Herman and Bruno M. Carvalho

13. Computer-Aided Diagnosis of Mammographic Calcification Clusters:

Impact of Segmentation 707

Maria Kallergi, John J. Heine, and Mugdha Tembey

14. Computer-Supported Segmentation of Radiological Data 753

Philippe Cattin, Matthias Harders, Johannes Hug, Raimundo Sierra, and Gabor Szekely

The Editors 799

Index 805

10 Ways To Fight Off Cancer

10 Ways To Fight Off Cancer

Learning About 10 Ways Fight Off Cancer Can Have Amazing Benefits For Your Life The Best Tips On How To Keep This Killer At Bay Discovering that you or a loved one has cancer can be utterly terrifying. All the same, once you comprehend the causes of cancer and learn how to reverse those causes, you or your loved one may have more than a fighting chance of beating out cancer.

Get My Free Ebook

Post a comment