Intelligent Systems Reference Library 140 Kenji Suzuki Yisong Chen E ditors Artificial Intelligence in Decision Support Systems for Diagnosis in Medical Imaging Intelligent Systems Reference Library Volume 140 Series editors Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland e-mail: [email protected] Lakhmi C. Jain, University of Canberra, Canberra, Australia; Bournemouth University, UK; KES International, UK e-mail: [email protected]; [email protected] URL: http://www.kesinternational.org/organisation.php The aim of this series is to publish a Reference Library, including novel advances and developments in all aspects of Intelligent Systems in an easily accessible and well structured form. The series includes reference works, handbooks, compendia, textbooks,well-structuredmonographs,dictionaries,andencyclopedias.Itcontains well integrated knowledge and current information in the field of Intelligent Systems. The series covers the theory, applications, and design methods of IntelligentSystems.Virtuallyalldisciplinessuchasengineering,computerscience, avionics, business, e-commerce, environment, healthcare, physics and life science are included. The list of topics spans all the areas of modern intelligent systems such as: Ambient intelligence, Computational intelligence, Social intelligence, Computational neuroscience, Artificial life, Virtual society, Cognitive systems, DNA and immunity-based systems, e-Learning and teaching, Human-centred computing and Machine ethics, Intelligent control, Intelligent data analysis, Knowledge-based paradigms, Knowledge management, Intelligent agents, Intelli- gent decision making, Intelligent network security, Interactive entertainment, Learningparadigms,Recommendersystems,RoboticsandMechatronicsincluding human-machine teaming, Self-organizing and adaptive systems, Soft computing including Neural systems, Fuzzy systems, Evolutionary computing and the Fusion of these paradigms, Perception and Vision, Web intelligence and Multimedia. More information about this series at http://www.springer.com/series/8578 ⋅ Kenji Suzuki Yisong Chen Editors fi Arti cial Intelligence in Decision Support Systems for Diagnosis in Medical Imaging 123 Editors KenjiSuzuki YisongChen Medical Imaging Research Center Key Laboratoryof MachinePerception andDepartmentof Electrical and (Ministry of Education), Schoolof Computer Engineering Electronics Engineering andComputer Illinois Institute of Technology Science Chicago, IL PekingUniversity USA Beijing China and World Research HubInitiative (WRHI), Institute of Innovative Research (IIR) Tokyo Institute of Technology Yokohama,Kanagawa Japan ISSN 1868-4394 ISSN 1868-4408 (electronic) Intelligent Systems Reference Library ISBN978-3-319-68842-8 ISBN978-3-319-68843-5 (eBook) https://doi.org/10.1007/978-3-319-68843-5 LibraryofCongressControlNumber:2017957701 ©SpringerInternationalPublishingAG2018 Thisworkissubjecttocopyright.AllrightsarereservedbythePublisher,whetherthewholeorpart of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission orinformationstorageandretrieval,electronicadaptation,computersoftware,orbysimilarordissimilar methodologynowknownorhereafterdeveloped. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publicationdoesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfrom therelevantprotectivelawsandregulationsandthereforefreeforgeneraluse. 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 authorsortheeditorsgiveawarranty,expressorimplied,withrespecttothematerialcontainedhereinor for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictionalclaimsinpublishedmapsandinstitutionalaffiliations. Printedonacid-freepaper ThisSpringerimprintispublishedbySpringerNature TheregisteredcompanyisSpringerInternationalPublishingAG Theregisteredcompanyaddressis:Gewerbestrasse11,6330Cham,Switzerland I dedicate this book to my wife, Harumi Suzuki, for her support, encouragement, and love,andtomydaughters,MineruSuzukiand Juno Suzuki, for their love. Kenji Suzuki I dedicate this book to my wife, Hong Cui, for her unwavering support and love. Yisong Chen Foreword Medical images of CT, MRI, PET, ultrasonography, etc., prevail in clinical diag- nosis.Usually,ittakesyearsoftrainingforphysicianstoconducttheinterpretation of medical images. Still, it may happen that interpretations may change from physician to physician. Computer-aided diagnosis (CAD) is a technology to help physicians efficiently ease the interpretation procedure. The first CAD product for detection of breast lesions in mammography was approved by the FDA about 20 years ago. Many CAD systems have been installed in hospitals or shipped with scanners and help improve the diagnostic performance of physicians. The need of healthcareorganizationsforCADsystemsisevergrowing.Inadditiontoitssuccess in clinics, CAD can also be used for education and for remote diagnose, among others. CAD is an interdisciplinary field of traditional disciplines, such as statistics, mathematics, physics, medicine, computer and imaging technologies, and new disciplines, such as artificial intelligence, especially, pattern recognition and machine learning. It usually takes a long leaning curve to train sophisticated researcher and engineers for CAD. Many scientific discoveries and technological inventions have resulted during the interdisciplinary collaborations. This book demonstrates the picture of CAD by contributions from active groups and state-of-the-arttechniquesforCAD.Studentsmayfindhis/herownpathbyreading this book if they are to pursue a career in CAD. ThisbookisatimelypublicationforCAD.Itisonlyundertheappreciatedeffort of the well-respected CAD researchers Drs. Suzuki and Chen that it is possible to bring together the state-of-the-art advances in CAD in one readily accessible source. This book demonstrates a number of CAD applications in a variety of diseases and body regions, and for different imaging modalities. This book covers also a number of machine-learning methods and their CAD applications. Researchers, engineers, and professionals may find this book a comprehensive reference on CAD. With the performance of AlphaGo, a deep learning approach is becoming popular in CAD because it is able to learn from examples and prior knowledge. Several successful applications have been reported. A number of start-ups and big vii viii Foreword vendors are investing in this technology. Although most of current CAD systems areorgan-specific,modality-tailored,anddisease-specific,thediseases,organs,and imaging modalities covered within CAD systems are rapidly expanding. It is expected that CAD systems will evolve with thedevelopment ofmachine learning to cover whole-body and cross-modalities and to incorporate other patients’ symptoms and clinical tests. Will physicians lose theirjobs withthedevelopment of CAD? Very unlikely in my opinion. Although machine learning is becoming ubiquitous, it is necessary to understand what the machines are doing to trust it. The problem with the deep learning is that knowledge is trained into a neural network of a huge number of parameters, rather than into us. It works like a black box at present. Moreover, successfulexamplesofCADareonlyfortypicaldiseases,butnotforrarediseases becauseofinsufficientdatatotrainthenetwork.Insuchrarecases,theengagement of physicians is indispensable. There are, and will remain, debates if machine intelligence will perform over human beings and whether CAD will replace physicians in the future. This book is a valuable source for reads to envisage the future of CAD while reading the vivid stories in this book. Beijing, China Ming Jiang Peking University Preface A decision support system for diagnosis is a crucial element in medicine and patients’ healthcare, because diagnosis is a complex task that is often difficult for even experienced physicians. Medical imaging offers useful information on patients’ medical conditions and clues to causes of their symptoms and diseases. Thus, medical imaging is indispensable for accurate medical decision making in modern medicine. Medical images, however, provide a large number of images, which physicians must interpret. That would lead to “information overload” for physicians, and it can complicate medical decision making further. Therefore, intelligent decision support systems are increasingly demanded to help physicians intheirdecisionmakingindiagnosisandtreatmentthroughmedicalimages.Inthe computer aids in medical decision making, computational intelligence plays very important roles to intelligently support physicians’ decision making. The areas of researchinthisfieldincludecomputer-aideddiagnosis,computer-aidedsurgeryand therapy, medical image analysis, automated organ/lesion segmentation, automated image fusion, automated image annotation, and content-based image retrieval. As this field of intelligent decision support systems for diagnosis through medical images is one of the most promising, growing fields, the efforts are cur- rently scattered, and a large number of researchers participated in the field and developed a number of intelligent diagnosis methods based on medical imaging. This book covers the state-of-the-art technologies and recent advances in the field of intelligent decision support systems for diagnosis through medical images. We expectthatthisbookwillbeusefulforprofessors,students,researchers,engineers, andprofessionals intheirstudies, research, development aswellasdailyworkand practice. Leading researchers in the field contributed 13 chapters to this book in which they describe their cutting-edge techniques and studies on intelligent decision support systems for diagnosis through medical images. The 13 chapters are orga- nized infivepartsthat representfivemajor research areas inthe fieldofintelligent decision support systems. Part I contains three chapters provided by leading researchers in the research area of advanced machine learning in computer-aided systems. ix x Preface In Chapter “Multi-modality Feature Learning in Diagnoses of Alzheimer’s Disease,”Dr.DaoqiangZhangintroducesalabel-alignedmultitaskfeatureselection methodthatcanfullyexploretherelationshipsacrossbothmodalitiesandsubjects, andthen,theyproposeadiscriminativemultitaskfeatureselectionmethodtoselect the most discriminative features for multimodality-based classification. The experimental results on magnetic resonance imaging (MRI), fluorodeoxyglucose positron emission tomography (FDG-PET), and cerebrospinal fluid (CSF) data of subjects from the Alzheimer’s disease neuroimaging initiative (ADNI) database demonstrate the effectiveness of their proposed method. InChapter“AComparativeStudyofModernMachineLearningApproachesfor Focal Lesion Detection and Classification in Medical Images: BoVW, CNN and MTANN,”Drs.NimaTajbakhshandKenjiSuzukireviewandcomparetwomajor end-to-end machine-learning models, massive-training artificial neural networks (MTANNs) and convolutional neural networks (CNNs), and a well-known non-end-to-end machine-learning model, a bag of visual words (BoVW) with Fisher vectors in detection and classification of focal lesions in medical images. They show that MTANNs outperform CNNs and BoVW in detection of lung nodules and colorectal polyps and classification of lung nodules in CT. In Chapter “Introduction to Binary Coordinate Ascent: New Insights into EfficientFeatureSubsetSelectionforMachineLearning,”Drs.AminZarshenasand Kenji Suzuki describe a novel optimization technique based on their originally developedoptimizationalgorithm,thecoordinatedescentalgorithm.Thealgorithm is an iterative deterministic local optimization approach that can be coupled with wrapper, filter, or hybrid feature selection techniques. The algorithm searches throughout the space of binary-coded input variables by iteratively optimizing the objective function in each dimension at a time. With their new technique, the efficiencyintermsofthenumberofsubsetevaluationswasimprovedsubstantially. PartIIcontainstwochaptersprovidedbytwoleadinggroupsworkinginthearea of computer-aided detection. In Chapter “Automated Lung Nodule Detection Using Positron Emission Tomography/Computed Tomography,” Drs. Atsushi Teramoto and Hiroshi Fujita highlighttheirrecentcontributionstoahybriddetectionschemeoflungnodulesin PET/CT images. The method detects lung nodules by using both the anatomical information obtained by CT and the functional information obtained with PET. Their results demonstrate that the proposed hybrid method would be useful for computer-aided detection of lung cancer in clinical practice. In Chapter “Detecting Mammographic Masses via Image Retrieval and Discriminative Learning,” Drs. Menglin Jiang, Shaoting Zhang, and Dimitris N. Metaxas introduce an automatic computer-aided diagnosis (CAD) approach that integrates content-based image retrieval (CBIR) and discriminative learning. Their approach achieves a high mass detection accuracy and retrieval precision, com- paring favorably with traditional methods. Compared with CBIR-based CAD methods, their approach serves as a fully automated “double reading” aid without radiologists’ labeling of suspicious regions. Part III contains four chapters in the area of computer-aided diagnosis.