Nandinee Haq · Patricia Johnson · Andreas Maier · Chen Qin · Tobias Würfl · Jaejun Yoo (Eds.) Machine Learning 7 8 5 for Medical 3 1 S C Image Reconstruction N L 5th International Workshop, MLMIR 2022 Held in Conjunction with MICCAI 2022 Singapore, September 22, 2022, Proceedings Lecture Notes in Computer Science 13587 FoundingEditors GerhardGoos KarlsruheInstituteofTechnology,Karlsruhe,Germany JurisHartmanis CornellUniversity,Ithaca,NY,USA EditorialBoardMembers ElisaBertino PurdueUniversity,WestLafayette,IN,USA WenGao PekingUniversity,Beijing,China BernhardSteffen TUDortmundUniversity,Dortmund,Germany MotiYung ColumbiaUniversity,NewYork,NY,USA Moreinformationaboutthisseriesathttps://link.springer.com/bookseries/558 · · Nandinee Haq Patricia Johnson · · · Andreas Maier Chen Qin Tobias Würfl Jaejun Yoo (Eds.) Machine Learning for Medical Image Reconstruction 5th International Workshop, MLMIR 2022 Held in Conjunction with MICCAI 2022 Singapore, September 22, 2022 Proceedings Editors NandineeHaq PatriciaJohnson Hitachi NYUGrossmanSchoolofMedicine Montreal,Canada NewYork,NY,USA AndreasMaier ChenQin Friedrich-Alexander-Universität UniversityofEdinburgh Erlangen,Bayern,Germany Edinburgh,UK TobiasWürfl JaejunYoo SiemensHealthineers UlsanNationalInstituteofScience Erlangen,Germany andTechnology Ulsan,Korea(Republicof) ISSN 0302-9743 ISSN 1611-3349 (electronic) LectureNotesinComputerScience ISBN 978-3-031-17246-5 ISBN 978-3-031-17247-2 (eBook) https://doi.org/10.1007/978-3-031-17247-2 ©TheEditor(s)(ifapplicable)andTheAuthor(s),underexclusivelicense toSpringerNatureSwitzerlandAG2022 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,theauthors,andtheeditorsaresafetoassumethattheadviceandinformationinthisbookare 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 We are proud to present the proceedings for the Fifth Workshop on Machine Learning for Medical Image Reconstruction (MLMIR 2022) which was held on September 22, 2022, at the Resorts World Convention Centre in Singapore, as part ofthe25thMedicalImageComputingandComputerAssistedIntervention(MICCAI 2022)conference. Imagereconstructioncommonlyreferstosolvinganinverseproblem,recoveringa latent image of some physical parameter from a set of noisy measurements assuming aphysicalmodelofthegeneratingprocessbetweentheimageandthemeasurements. In medical imaging two particular widespread applications are computed tomography (CT)andmagneticresonanceimaging(MRI).Usingthosetwomodalitiesasexamples, conditions have been established under which the associated reconstruction problems can be solved uniquely. However, in many cases there is a need to recover solutions fromfewermeasurementstoreducethedoseappliedtopatientsortoreducethemea- surementtime.Thetheoryofcompressedsensingshowedhowtopursuethiswhilestill enablingaccuratereconstructionbyusingpriorknowledgeabouttheimagedobjects.A criticalquestionistheconstructionofsuitablemodelsofpriorknowledgeaboutimages. Researchhasdepartedfromconstructingexplicitpriorsforimagesandmovedtowards learningsuitablepriorsfromlargedatasetsusingmachinelearning(ML). After four previous successful workshops, we found that ML approaches havefoundtheirwayintomultipleproductsfordifferentmodalities.Itscross-modality approachbringstogetherresearchersfromvariousmodalitiesrangingfromCTandMRI tomicroscopyandX-rayfluoroscopy.Webelievejointdiscussionfostersthetranslation ofalgorithmsbetweenmodalities. WewerefortunatethatShanshanWang(PaulC.LauterburResearchCenter,Chinese Academy of Sciences, China) and Jong Chul Ye (Kim Jaechul Graduate School of AI, KAIST, South Korea) accepted our invitation as keynote speakers and presented fascinatingkeynotelecturesaboutthestateoftheartinthisfield.Forthisfirstin-person iteration of the workshop after the start of the COVID-19 pandemic, we received 19 submissions and accepted 15 papers for inclusion in the workshop. The topics of the acceptedpapersarestilldominatedbyMRIreconstructionbutcoverabroadrangeof applicationsoverCT,PET,ultrasound,fluoroscopy,andmagneticparticleimaging. August2022 NandineeHaq PatriciaJohnson AndreasMaier ChenQin TobiasWürfl JaejunYoo Organization WorkshopOrganizers NandineeHaq Hitachi,Canada PatriciaJohnson NewYorkUniversity,USA AndreasMaier Friedrich-Alexander-University Erlangen-Nuremberg,Germany ChenQin UniversityofEdinburgh,UK TobiasWürfl SiemensHealthineers,Germany JaejunYoo UlsanNationalInstituteofScienceand Technology,SouthKorea ScientificProgramCommittee DelaramBehnami UniversityofBritishColumbia,Canada TolgaCukur BilkentUniversity,Turkey KerstinHammernik ImperialCollegeLondon,UK EssamRashed UniversityofHyogo,Japan AndrewReader King’sCollegeLondon,UK ZhengguoTan Friedrich-Alexander-University Erlangen-Nuremberg,Germany GeWang RensselaerPolytechnicInstitute,USA ShanshanWang PaulC.LauterburResearchCenter,Chinese AcademyofSciences,China PengweiWu GEGlobalResearch,USA GuangYang ImperialCollegeLondon,UK JongChulYe KimJaechulGraduateSchoolofAI,KAIST, SouthKorea CanZhao NVIDIA,USA Contents DeepLearningforMagneticResonanceImaging RethinkingtheOptimizationProcessforSelf-supervisedModel-Driven MRIReconstruction ................................................... 3 WeijianHuang, ChengLi, WenxinFan, ZiyaoZhang, TongZhang, YongjinZhou,QiegenLiu,andShanshanWang NPB-REC: Non-parametric Assessment of Uncertainty inDeep-Learning-BasedMRIReconstructionfromUndersampledData ....... 14 SamahKhawaledandMotiFreiman AdversarialRobustnessofMRImageReconstructionUnderRealistic Perturbations ......................................................... 24 JanNikolasMorshuis, SergiosGatidis, MatthiasHein, andChristianF.Baumgartner High-FidelityMRIReconstructionwiththeDenselyConnectedNetwork CascadeandFeatureResidualDataConsistencyPriors ...................... 34 JingshuaiLiu,ChenQin,andMehrdadYaghoobi Metal Artifact Correction MRI Using Multi-contrast Deep Neural NetworksforDiagnosisofDegenerativeSpinalDiseases .................... 44 Jaa-YeonLee, MinAYoon, ChoongGuenChee, JaeHwanCho, JinHoonPark,andSung-HongPark Segmentation-AwareMRIReconstruction ................................. 53 MertAcar,TolgaÇukur,andI˙lkayÖksüz MRIReconstructionwithConditionalAdversarialTransformers ............. 62 YilmazKorkmaz,MuzafferÖzbey,andTolgaCukur DeepLearningforGeneralImageReconstruction ANoise-Level-AwareFrameworkforPETImageDenoising ................. 75 YeLi,JiananCui,JunyuChen,GuodongZeng,ScottWollenweber, FlorisJansen, Se-InJang, KyungsangKim, KuangGong, andQuanzhengLi DuDoTrans: Dual-Domain Transformer for Sparse-View CT Reconstruction ........................................................ 84 CeWang,KunShang,HaimiaoZhang,QianLi,andS.KevinZhou viii Contents Deep Denoising Network for X-Ray Fluoroscopic Image Sequences ofMovingObjects ..................................................... 95 WonjinKim, WonkyeongLee, Sun-YoungJeon, NayeonKang, GeonhuiJo,andJang-HwanChoi PP-MPI: A Deep Plug-and-Play Prior for Magnetic Particle Imaging Reconstruction ........................................................ 105 BarisAskin, AlperGüngör, DamlaAlptekinSoydan, EmineUlkuSaritas,CanBarıs¸Top,andTolgaCukur Learning While Acquisition: Towards Active Learning Framework forBeamforminginUltrasoundImaging .................................. 115 MayankKatare, MaheshRaveendranathaPanicker, A.N.Madhavanunni,andGayathriMalamal DPDudoNet:Deep-PriorBasedDual-DomainNetworkforLow-Dose ComputedTomographyReconstruction ................................... 123 TemitopeEmmanuelKomolafe, YuhangSun, NizhuanWang, KaicongSun,GuohuaCao,andDinggangShen MTD-GAN: Multi-task Discriminator Based Generative Adversarial NetworksforLow-DoseCTDenoising ................................... 133 SungguKyung, JongJunWon, SeongyongPak, Gil-sunHong, andNamkugKim Uncertainty-InformedBayesianPETImageReconstructionUsingaDeep ImagePrior ........................................................... 145 ViswanathP.Sudarshan, K.PavanKumarReddy, MohanaSingh, JayavardhanaGubbi,andArpanPal AuthorIndex ......................................................... 157 Deep Learning for Magnetic Resonance Imaging