Islem Rekik Ehsan Adeli Sang Hyun Park (Eds.) 3 4 8 Predictive Intelligence 1 1 S C in Medicine N L Second International Workshop, PRIME 2019 Held in Conjunction with MICCAI 2019 Shenzhen, China, October 13, 2019 Proceedings Lecture Notes in Computer Science 11843 Founding Editors Gerhard Goos Karlsruhe Institute of Technology, Karlsruhe, Germany Juris Hartmanis Cornell University, Ithaca, NY, USA Editorial Board Members Elisa Bertino Purdue University, West Lafayette, IN, USA Wen Gao Peking University, Beijing, China Bernhard Steffen TU Dortmund University, Dortmund, Germany Gerhard Woeginger RWTH Aachen, Aachen, Germany Moti Yung Columbia University, New York, NY, USA More information about this series at http://www.springer.com/series/7412 Islem Rekik Ehsan Adeli (cid:129) (cid:129) Sang Hyun Park (Eds.) Predictive Intelligence in Medicine Second International Workshop, PRIME 2019 Held in Conjunction with MICCAI 2019 Shenzhen, China, October 13, 2019 Proceedings 123 Editors Islem Rekik Ehsan Adeli BASIRA StanfordUniversity Istanbul TechnicalUniversity Stanford, CA,USA Istanbul,Turkey SangHyun Park Daegu GyeongbukInstitute of ScienceandTechnology Daegu,Korea (Republic of) ISSN 0302-9743 ISSN 1611-3349 (electronic) Lecture Notesin Computer Science ISBN 978-3-030-32280-9 ISBN978-3-030-32281-6 (eBook) https://doi.org/10.1007/978-3-030-32281-6 LNCSSublibrary:SL6–ImageProcessing,ComputerVision,PatternRecognition,andGraphics ©SpringerNatureSwitzerlandAG2019 Thisworkissubjecttocopyright.AllrightsarereservedbythePublisher,whetherthewholeorpartofthe 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 or information storageandretrieval,electronicadaptation,computersoftware,orbysimilarordissimilarmethodologynow knownorhereafterdeveloped. Theuseofgeneraldescriptivenames,registerednames,trademarks,servicemarks,etc.inthispublication doesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfromtherelevant protectivelawsandregulationsandthereforefreeforgeneraluse. Thepublisher,theauthorsandtheeditorsaresafetoassumethattheadviceandinformationinthisbookare believedtobetrueandaccurateatthedateofpublication.Neitherthepublishernortheauthorsortheeditors give a warranty, expressed or implied, with respect to the material contained herein or for any errors or omissionsthatmayhavebeenmade.Thepublisherremainsneutralwithregardtojurisdictionalclaimsin publishedmapsandinstitutionalaffiliations. ThisSpringerimprintispublishedbytheregisteredcompanySpringerNatureSwitzerlandAG Theregisteredcompanyaddressis:Gewerbestrasse11,6330Cham,Switzerland Preface Itwouldconstituteastunningprogressinmedicineif,inafewyears,wecontributeto engineering a ‘predictive intelligence’ able to predict missing clinical data with high precision.Giventheoutburstofbigandcomplexmedicaldatawithmultiplemodalities (e.g.,structuralmagneticresonanceimaging(MRI)andrestingfunctionMRI(rsfMRI)) and multiple acquisition time-points (e.g., longitudinal data), more intelligent predictivemodelsareneededtoimprovediagnosisofawidespectrumofdiseasesand disorders while leveraging minimal medical data. Basically, predictive intelligence in medicine (PRIME workshop) primarily aims to diagnose in the earliest stage using minimal clinically non-invasive data. For instance, PRIME would constitute a breakthrough in early neurological disorder diagnosis as it would allow accurate early diagnosis using multimodal MRI data (e.g., diffusion and functional MRIs) and follow-up observations all predicted from only T1-weighted MRI acquired at baseline time-point. Existing computer-aided diagnosis methods can be divided into two main categories: (1) analytical methods and (2) predictive methods. While analytical methodsaimtoefficientlyanalyze,represent,andinterpretdata(staticorlongitudinal), predictivemethodsleveragethedatacurrentlyavailabletopredictobservationsatlater time-points(i.e.,forecastingthefuture)orpredictingobservationsatearliertime-points (i.e.,predictingthepastformissingdatacompletion).Forinstance,amethodthatonly focusesonclassifyingpatientswithmildcognitiveimpairment(MCI)andpatientswith Alzheimer’s disease (AD) is an analytical method, while a method which predicts if a subject diagnosed with MCI will remain stable or convert to AD over time is a predictive method. Similar examples can be established for various neurodegenerative or neuropsychiatric disorders, degenerative arthritis, or in cancer studies, in which the disease or disorder develops over time. Following the success of the first edition of PRIME-MICCAI in 2018, the second editionofPRIME2019aimedtodrivethefieldof‘high-precisionpredictivemedicine,’ where late medical observations are predicted with high precision, while providing explanation via machine and deep learning, and statistically, mathematically, or physically-based models of healthy, disordered development, and aging. Despite the terrific progress that analytical methods have made in the last 20 years in medical image segmentation, registration, or other related applications, efficient predictive intelligent models and methods are somewhat lagging behind. As such predictive intelligence develops and improves (and this is likely to do so exponentially in the coming years), this will have far-reaching consequences for the development of new treatment procedures and novel technologies. These predictive models will begin to shedlightononeofthemostcomplexhealthcareandmedicalchallengeswehaveever encountered, and, in doing so, change our basic understanding of who we are. vi Preface What Are the Key Challenges We Aim to Address? The main aim of PRIME-MICCAI is to propel the advent of predictive models in a broadsense,withapplicationtomedicaldata.Particularly,theworkshopaccepted8-to 12-page papers describing newcutting-edgepredictive modelsandmethods thatsolve challenging problems in the medical field. We hope that the PRIME workshop becomes a nest for high-precision predictive medicine – one that is set to transform multiplefieldsofhealthcaretechnologiesinunprecedentedways.Topicsofinterestfor the workshop included but were not limited to predictive methods dedicated to the following topics: – Modeling and predicting disease development or evolution from a limited number of observations – Computer-aided prognostic methods (e.g., for brain diseases, prostate cancer, cervical cancer, dementia, acute disease, neurodevelopmental disorders) – Forecasting disease or cancer progression over time – Predicting low-dimensional data (e.g., behavioral scores, clinical outcome, age, gender) – Predicting the evolution or development of high-dimensional data (e.g., shapes, graphs, images, patches, abstract features, learned features) – Predicting high-resolution data from low-resolution data – Prediction methods using 2D, 2D+t, 3D, 3D+t, ND, and ND+t data – Predicting data of one image modality from a different modality (e.g., data synthesis) – Predicting lesion evolution – Predicting missing data (e.g., data imputation or data completion problems) – Predicting clinical outcome from medical data (genomic, imaging data, etc) Key Highlights This workshop mediated ideas from both machine learning and mathematical, statistical,andphysicalmodelingresearchdirectionsinthehopeofprovidingadeeper understanding of thefoundations ofpredictive intelligence developedfor medicine, as well as to where we currently stand and what we aspire to achieve through this field. PRIME-MICCAI 2019 featured a single-track workshop with keynote speakers with deepexpertiseinhigh-precisionpredictivemedicine usingmachine learning andother modelingapproacheswhicharebelievedtostandatopposingdirections.Ourworkshop also included technical paper presentations, poster sessions, and demonstrations. Eventually,thishelpssteerawidespectrumofMICCAIpublicationsfrombeing‘only analytical’ to being ‘jointly analytical and predictive.’ We received a total of 18 submissions. All papers underwent a rigorous double-blinded review process by at least 2 members (mostly 3 members) of the Program Committee composed of 26 well-known research experts in the field. Preface vii Theselectionofthepaperswasbasedontechnicalmerit,significanceofresults,and relevance and clarity of presentation. Based on the reviewing scores and critiques, all PRIME2019submissionsscoredhighbyreviewers,i.e.,allhadanaveragescoreofat least above the acceptance threshold. Hence all 18 submissions were approved for publication in the present proceedings. August 2019 Islem Rekik Ehsan Adeli Sang Hyun Park Organization Chairs Islem Rekik Istanbul Technical University, Turkey Ehsan Adeli Stanford University, USA Sang Hyun Park DGIST, South Korea Program Committee Amir Alansary Imperial College London, UK Changqing Zhang Tianjin University, China Dong Nie University of North Carolina, USA Duygu Sarikaya University of Rennes 1, France Gerard Sanroma Pompeu Fabra University, Spain Guorong Wu University of North Carolina, USA Heung-Il Suk Korea University, South Korea Ilwoo Lyu Vanderbilt University, USA Ipek Oguz University of Pennsylvania, USA Jaeil Kim Kyungpook National University, South Korea Le Lu PAII Inc., USA Lichi Zhang Shanghai Jiao Tong University, China Mara Valds Hernndez University of Edinburgh, UK Minjeong Kim University of North Carolina at Greensboro, USA Nesrine Bnouni National Engineering School of Sousse (ENISo), Tunisia Pew-Thian Yap University of North Carolina, USA Qian Wang Shanghai Jiao Tong University, China Qingyu Zhao Stanford University, USA Seyed Mostafa Kia Donders Institute, The Netherlands Stefanie Demirci Technische Universität München, Germany Sophia Bano University College London, UK Ulas Bagci University of Central Florida, USA Xiaohuan Cao United Imaging Intelligence, China Yu Zhang Stanford University, USA Yue Gao Tsinghua University, China Ziga Spiclin University of Ljubljana, Slovenia Contents TADPOLE Challenge: Accurate Alzheimer’s Disease Prediction Through Crowdsourced Forecasting of Future Data . . . . . . . . . . . . . . . . . . . . . . . . . 1 Răzvan V. Marinescu, Neil P. Oxtoby, Alexandra L. Young, Esther E.Bron,Arthur W.Toga,Michael W.Weiner, FrederikBarkhof, Nick C. Fox, Polina Golland, Stefan Klein, and Daniel C. Alexander Inter-fractionalRespiratoryMotionModellingfromAbdominalUltrasound: A Feasibility Study. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 Alina Giger, Christoph Jud, Damien Nguyen, Miriam Krieger, Ye Zhang, Antony J. Lomax, Oliver Bieri, Rares Salomir, and Philippe C. Cattin Adaptive Neuro-Fuzzy Inference System-Based Chaotic Swarm Intelligence Hybrid Model for Recognition of Mild Cognitive Impairment from Resting-State fMRI. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 Ahmed M. Anter and Zhiguo Zhang Deep Learning via Fused Bidirectional Attention Stacked Long Short-Term Memory for Obsessive-Compulsive Disorder Diagnosis and Risk Screening. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Chiyu Feng, Lili Jin, Chuangyong Xu, Peng Yang, Tianfu Wang, Baiying Lei, and Ziwen Peng Modeling Disease Progression in Retinal OCTs with Longitudinal Self-supervised Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 Antoine Rivail, Ursula Schmidt-Erfurth, Wolf-Dieter Vogel, Sebastian M. Waldstein, Sophie Riedl, Christoph Grechenig, Zhichao Wu, and Hrvoje Bogunovic Predicting Response to the Antidepressant Bupropion Using Pretreatment fMRI. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 Kevin P. Nguyen, Cherise Chin Fatt, Alex Treacher, Cooper Mellema, Madhukar H. Trivedi, and Albert Montillo Progressive Infant Brain Connectivity Evolution Prediction from Neonatal MRI Using Bidirectionally Supervised Sample Selection. . . . . 63 Olfa Ghribi, Gang Li, Weili Lin, Dinggang Shen, and Islem Rekik