Bin Zhou, MS
StatisticianCards
About
Research
Publications
2026
HiFusion: Hierarchical Intra-Spot Alignment and Regional Context Fusion for Spatial Gene Expression Prediction from Histopathology
Weng Z, Fang Y, Qian J, Wang X, Cooper L, Cai W, Zhou B. HiFusion: Hierarchical Intra-Spot Alignment and Regional Context Fusion for Spatial Gene Expression Prediction from Histopathology. Proceedings Of The AAAI Conference On Artificial Intelligence 2026, 40: 10630-10637. DOI: 10.1609/aaai.v40i13.38036.Peer-Reviewed Original ResearchGene expression predictionState-of-the-art performanceWhole-slide imagesState-of-the-artDeep learning frameworkGene expressionExpression predictionSpatial transcriptomicsContext fusionCross-attentionFusion moduleAlignment lossSemantic consistencyLearning frameworkScalable solutionContextual informationRepresentational capacityAdoption barriersMicroenvironmental cuesST datasetsCross-validationMorphological representationH&E-stained whole-slide imagesExpressionComputational methodsBurnout in Sickle Cell Disease-Focused Hematology-Oncology Trained Physicians: A National Cross-Sectional Study
Restrepo V, Marshall A, Feder K, Afranie-Sakyi J, Carrithers B, Jones-Wonni B, Mensah C, Zhou B, Mistry R, McGann P, Azar S, De Castro L, Calhoun C, Bozang K, Brown T, Lee A, Van Doren L. Burnout in Sickle Cell Disease-Focused Hematology-Oncology Trained Physicians: A National Cross-Sectional Study. Blood Advances 2026 PMID: 41817318, DOI: 10.1182/bloodadvances.2025018338.Peer-Reviewed Original ResearchThis study investigates burnout in U.S. hematology-oncology physicians specializing in sickle cell disease, showing higher burnout rates linked to systemic factors, reduced recreation, and lower job pride.AMA-SAM: Adversarial multi-Domain alignment of segment anything model for high-Fidelity histology nuclei segmentation
Qian J, Fang Y, Hao J, Zhou B. AMA-SAM: Adversarial multi-Domain alignment of segment anything model for high-Fidelity histology nuclei segmentation. Medical Image Analysis 2026, 111: 104011. PMID: 41812368, DOI: 10.1016/j.media.2026.104011.Peer-Reviewed Original ResearchDomain shiftDomain-invariant representation learningState-of-the-art approachesAccurate segmentationGradient reversal layerState-of-the-artMulti-domain settingLow resolution outputNuclei segmentation methodRepresentation learningAlignment moduleCell nuclei segmentation methodSegmentation mapDiverse domainsHistopathological imagesPerformance dropSegmentation methodAuxiliary datasetsReversal layerNuclei segmentationMultiple datasetsDatasetNegative transferSegmentation of cell nucleiPrimary datasetParametric Cardiac Imaging with 18F-Flutemetamol PET to Evaluate the Impact of Tafamidis in Patients with Transthyretin Cardiac Amyloidosis.
Liu Q, Shi T, Gravel P, Sharma A, De Freitas C, Fazzone-Chettiar R, Van Laere K, Baldick A, Kattan C, Guo X, Guo L, Xie H, Chen X, Zhou B, Liu Y, Carson R, Liu C, Miller E. Parametric Cardiac Imaging with 18F-Flutemetamol PET to Evaluate the Impact of Tafamidis in Patients with Transthyretin Cardiac Amyloidosis. Journal Of Nuclear Medicine 2026, jnumed.125.270003. PMID: 41748295, DOI: 10.2967/jnumed.125.270003.Peer-Reviewed Original ResearchTransthyretin cardiac amyloidosisATTR-CACardiac amyloidosisMethods:Results:ATTR-CA patientsImpact of tafamidisMultilinear analysis 1Blood volume fractionBlood-to-plasma ratioImage-derived input functionTreatment-related changesBlood-to-plasmaMyocardial blood flowCardiac imagingMyocardial blood volume fractionBlood flowInput functionSensitive to treatment-related changesTafamidisPET dataPatientsAmyloid burdenAmyloidosisTracer kineticsDOSTA-Net: Domain-Shuffle Temporal Attention Network for Vessel Extraction in X-Ray Coronary Angiography Using Synthetic Data
Hao J, Cantrell D, Abdalla R, Ansari S, Zhou B. DOSTA-Net: Domain-Shuffle Temporal Attention Network for Vessel Extraction in X-Ray Coronary Angiography Using Synthetic Data. IEEE Transactions On Medical Imaging 2026, PP: 1-1. PMID: 41615973, DOI: 10.1109/tmi.2026.3659754.Peer-Reviewed Original ResearchTemporal attention networkState-of-the-art methodsLack of large-scale datasetsAttention networkDeep learning-based methodsTraining deep learning modelsUnlabeled real dataTemporal feature learningState-of-the-artImage quality assessmentVessel segmentation performanceLearning-based methodsLarge-scale datasetsHigh-quality annotationsDevelopment of deep learning-based methodsDeep learning modelsReal dataDomain gapPseudo-labelsDomain discrepancyFeature learningSynthetic dataHuman annotatorsSegmentation performanceLoss functionCoronary artery segmentation in non-contrast cardiac CT using anatomy-informed contrastive learning and synthetic data
Hao J, He X, Durak G, Aktas H, Bagci U, Shah N, Zhou B. Coronary artery segmentation in non-contrast cardiac CT using anatomy-informed contrastive learning and synthetic data. Physics In Medicine And Biology 2026, 71: 025017. PMID: 41534216, DOI: 10.1088/1361-6560/ae387c.Peer-Reviewed Original ResearchConceptsSynthetic dataDomain adaptation approachAccurate coronary artery segmentationState-of-the-artContrastive learning strategyDeep learning frameworkContrast-to-noise ratioContrastive learningGeneralization capabilityLearning frameworkGeneration pipelineManual annotationNon-contrast imagesModel weightsAutomatic segmentationAnatomical priorsDatasetClinical datasetsRadiation doseExperimental resultsBackground structureLearning strategiesFalse positivesTraditional methodsAnnotation
2025
FairREAD: Re-fusing demographic attributes after disentanglement for fair medical image classification
Gao Y, Hao J, Zhou B. FairREAD: Re-fusing demographic attributes after disentanglement for fair medical image classification. Medical Image Analysis 2025, 107: 103858. PMID: 41202615, DOI: 10.1016/j.media.2025.103858.Peer-Reviewed Original ResearchConceptsOut-of-distribution testsMedical image classificationOut-of-distributionX-ray datasetAdversarial trainingImage representationImage classificationDeep learningSensitive attributesMedical imagesEfficient frameworkOrthogonality constraintsDemographic attributesEquitable performanceThreshold adjustmentImage dataPerformance disparityRelevant informationFairnessUnfairnessAttributesClinical profile and long-term outcomes of chest pain patients with coronary microvascular dysfunction from the emergency department – results from the Yale-CMD registry
Safdar B, Zhou B, Li F, Camici P, Dziura J, Jastreboff A, Lansky A, Shah S, Sinusas A, Spatz E, D'Onofrio G. Clinical profile and long-term outcomes of chest pain patients with coronary microvascular dysfunction from the emergency department – results from the Yale-CMD registry. Microvascular Research 2025, 163: 104878. PMID: 41110546, DOI: 10.1016/j.mvr.2025.104878.Peer-Reviewed Original ResearchConceptsCMD patientsCoronary microvascular dysfunctionLong-term prognosisLong-term outcomesCoronary artery diseaseCoronary flow reserveEmergency departmentMicrovascular dysfunctionProspective cohortHeart failureHealthcare utilizationMedian follow-up timeHigh riskCohort of ED patientsMyocardial infarctionCardiac positron emission tomographyHigher adverse eventsChest pain patientsFollow-up timeAdverse cardiac eventsHigher MACE riskHigher healthcare utilizationAnnual follow-upSeattle Angina QuestionnaireProspective cohort of ED patientsLeqMod: Adaptable Lesion-Quantification-Consistent Modulation for Deep Learning Low-Count PET Image Denoising
Xia M, Xie H, Liu Q, Zhou B, Wang H, Li B, Rominger A, Li Q, Badawi R, Shi K, Fakhri G, Liu C. LeqMod: Adaptable Lesion-Quantification-Consistent Modulation for Deep Learning Low-Count PET Image Denoising. IEEE Transactions On Medical Imaging 2025, 45: 1115-1126. PMID: 41052161, PMCID: PMC12910708, DOI: 10.1109/tmi.2025.3618247.Peer-Reviewed Original ResearchPeak signal-to-noise ratioImage denoisingPET image denoisingLow-count imagesDenoising frameworkDenoised imageInference phaseSignal-to-noise ratioSegmentation networkModel architectureModel trainingDenoisingPositron emission tomography datasetsComputational burdenOptimization procedureNoise levelImagesAuxiliary toolFrameworkModulationDeepQumodesDatasetArchitectureVendorsS 2 CAC: Semi-supervised coronary artery calcium segmentation via scoring-driven consistency and negative sample boosting
Hao J, Shah N, Zhou B. S 2 CAC: Semi-supervised coronary artery calcium segmentation via scoring-driven consistency and negative sample boosting. Medical Image Analysis 2025, 107: 103823. PMID: 41045882, DOI: 10.1016/j.media.2025.103823.Peer-Reviewed Original ResearchConceptsPixel-level segmentationState-of-the-art performanceSemi-supervised learning frameworkLarge-scale annotated datasetsDynamically weighted loss functionNegative samplesState-of-the-artDown-sampling operationsWeighted loss functionAutomatic segmentation modelUnlabeled dataBaseline methodsAnnotated datasetsLearning frameworkSegmentation mapAnnotation processModel collapseOver-segmentationInformation lossManual annotationLoss functionSegmentation modelTraining processConsistency mechanismRobust performance
Get In Touch
Contacts
Email
Academic Office Number