Featured Publications
Cross-Attention for Improved Motion Correction in Brain PET
Cai Z, Zeng T, Lieffrig E, Zhang J, Chen F, Toyonaga T, You C, Xin J, Zheng N, Lu Y, Duncan J, Onofrey J. Cross-Attention for Improved Motion Correction in Brain PET. Lecture Notes In Computer Science 2023, 14312: 34-45. PMID: 38174216, PMCID: PMC10758996, DOI: 10.1007/978-3-031-44858-4_4.Peer-Reviewed Original ResearchDeep learning networkCross-attention mechanismDeep learning benchmarksMotion correctionTraining data domainPET list-mode dataPET image reconstructionQuality of reconstructionData domainCross attentionLearning networkSupervised mannerLearning benchmarksReference imageMotion trackingInherent informationList-mode dataImage reconstructionBrain PET dataPrediction resultsDifferent scannersHead motionImproved motion correctionNetworkSpatial correspondence
2024
A Flow-based Truncated Denoising Diffusion Model for super-resolution Magnetic Resonance Spectroscopic Imaging
Dong S, Cai Z, Hangel G, Bogner W, Widhalm G, Huang Y, Liang Q, You C, Kumaragamage C, Fulbright R, Mahajan A, Karbasi A, Onofrey J, de Graaf R, Duncan J. A Flow-based Truncated Denoising Diffusion Model for super-resolution Magnetic Resonance Spectroscopic Imaging. Medical Image Analysis 2024, 99: 103358. PMID: 39353335, DOI: 10.1016/j.media.2024.103358.Peer-Reviewed Original ResearchDenoising diffusion modelsDeep learning-based super-resolution methodsLearning-based super-resolution methodsMulti-scale super-resolutionGenerative modelSuper-resolution methodsDeep learning modelsHigh-resolution magnetic resonance spectroscopic imagingHigh-quality imagesPost-processing approachSuper-resolutionFlow-based networksLearning modelsLow resolutionTruncation stepLow-resolution dataSharpness adjustmentNetworkSensitivity restrictionsUncertainty estimationDiffusion modelImagesCapabilitySampling processSpectroscopic imagingMonte-Carlo Frequency Dropout for Predictive Uncertainty Estimation in Deep Learning
Zeevi T, Venkataraman R, Staib L, Onofrey J. Monte-Carlo Frequency Dropout for Predictive Uncertainty Estimation in Deep Learning. 2024, 00: 1-5. DOI: 10.1109/isbi56570.2024.10635511.Peer-Reviewed Original ResearchArtificial neural networkState-of-the-artMedical image dataPredictive uncertainty estimationBiomedical image dataImage dataOptimal artificial neural networkMC dropoutDropout approachSource-codeDrop-connectDeep learningNeural networkSignal spaceMonte-CarloPrediction uncertaintyUncertainty estimationDiverse setComprehensive comparisonPrediction scenariosDeepPosterior predictive distributionRepositoryDecision-makingNetwork
2023
Predicting tumor recurrence on baseline MR imaging in patients with early-stage hepatocellular carcinoma using deep machine learning
Kucukkaya A, Zeevi T, Chai N, Raju R, Haider S, Elbanan M, Petukhova-Greenstein A, Lin M, Onofrey J, Nowak M, Cooper K, Thomas E, Santana J, Gebauer B, Mulligan D, Staib L, Batra R, Chapiro J. Predicting tumor recurrence on baseline MR imaging in patients with early-stage hepatocellular carcinoma using deep machine learning. Scientific Reports 2023, 13: 7579. PMID: 37165035, PMCID: PMC10172370, DOI: 10.1038/s41598-023-34439-7.Peer-Reviewed Original Research
2022
Co-attention spatial transformer network for unsupervised motion tracking and cardiac strain analysis in 3D echocardiography
Ahn S, Ta K, Thorn S, Onofrey J, Melvinsdottir I, Lee S, Langdon J, Sinusas A, Duncan J. Co-attention spatial transformer network for unsupervised motion tracking and cardiac strain analysis in 3D echocardiography. Medical Image Analysis 2022, 84: 102711. PMID: 36525845, PMCID: PMC9812938, DOI: 10.1016/j.media.2022.102711.Peer-Reviewed Original ResearchConceptsSpatial transformer networkMotion trackingNoisy displacement fieldReliable motion estimationMotion tracking methodCardiac strain analysisTransformer networkDisplacement fieldDisplacement pathsMotion fieldTracking methodMotion estimationExperimental resultsStrain analysisSuperior performanceTemporal constraintsCardiac motionTrackingRegularization functionDependent featuresEchocardiography imagesNetworkPrior assumptionsFieldAn Adaptive Patch Sampling Scheme for Deep Learning Based PET Image Denoising
Wu J, Tan H, Liu H, Liu C, Onofrey J. An Adaptive Patch Sampling Scheme for Deep Learning Based PET Image Denoising. 2022, 00: 1-3. DOI: 10.1109/nss/mic44845.2022.10399313.Peer-Reviewed Original ResearchOver-smoothing effectSignal-to-noise ratioU-NetImage denoisingDeep learning-based approachPET image denoisingL1 loss functionPatch-based strategyLearning-based approachSampling schemeMean square errorHigh signal-to-noise ratioDenoising performanceLow-dose PET imagesNetwork trainingWeight mapData augmentationDenoisingLoss functionNetworkImage noiseSquare errorPatch samplesSampling rateSchemeFlow-Based Visual Quality Enhancer for Super-Resolution Magnetic Resonance Spectroscopic Imaging
Dong S, Hangel G, Chen E, Sun S, Bogner W, Widhalm G, You C, Onofrey J, de Graaf R, Duncan J. Flow-Based Visual Quality Enhancer for Super-Resolution Magnetic Resonance Spectroscopic Imaging. Lecture Notes In Computer Science 2022, 13609: 3-13. DOI: 10.1007/978-3-031-18576-2_1.Peer-Reviewed Original ResearchAdversarial networkVisual qualityDeep learning-based super-resolution methodsLearning-based super-resolution methodsFlow-based modelImage visual qualityGenerative adversarial networkHigh visual qualitySuper-resolution methodSuper-resolved imagesGenerative modelHigh-resolution imagesImage modalitiesFlow-based methodNetworkLow spatial resolutionUncertainty estimationImagesPromising resultsEnhancer networkAnatomical informationHigh fidelityEssential toolDatasetQuality adjustmentDual-Branch Squeeze-Fusion-Excitation Module for Cross-Modality Registration of Cardiac SPECT and CT
Chen X, Zhou B, Xie H, Guo X, Zhang J, Sinusas A, Onofrey J, Liu C. Dual-Branch Squeeze-Fusion-Excitation Module for Cross-Modality Registration of Cardiac SPECT and CT. Lecture Notes In Computer Science 2022, 13436: 46-55. DOI: 10.1007/978-3-031-16446-0_5.Peer-Reviewed Original ResearchConvolutional neural networkCross-modality registrationFeature fusionPrevious convolutional neural networkEarly feature fusionCross-modality informationMultiple convolutional layersMedical image registrationLow registration errorCardiac SPECTConvolutional layersCNN moduleImage featuresLate fusionSource codeNeural networkExcitation moduleInput modalitiesImage registrationSpatial featuresMultiple modalitiesRegistration errorPrevious methodsRigid registrationNetwork
2021
TVnet: Automated Time-Resolved Tracking of the Tricuspid Valve Plane in MRI Long-Axis Cine Images with a Dual-Stage Deep Learning Pipeline
Gonzales R, Lamy J, Seemann F, Heiberg E, Onofrey J, Peters D. TVnet: Automated Time-Resolved Tracking of the Tricuspid Valve Plane in MRI Long-Axis Cine Images with a Dual-Stage Deep Learning Pipeline. Lecture Notes In Computer Science 2021, 12906: 567-576. DOI: 10.1007/978-3-030-87231-1_55.Peer-Reviewed Original Research
2019
Generalizable Multi-Site Training and Testing Of Deep Neural Networks Using Image Normalization
Onofrey JA, Casetti-Dinescu DI, Lauritzen AD, Sarkar S, Venkataraman R, Fan RE, Sonn GA, Sprenkle PC, Staib LH, Papademetris X. Generalizable Multi-Site Training and Testing Of Deep Neural Networks Using Image Normalization. 2011 IEEE International Symposium On Biomedical Imaging: From Nano To Macro 2019, 00: 348-351. PMID: 32874427, PMCID: PMC7457546, DOI: 10.1109/isbi.2019.8759295.Peer-Reviewed Original ResearchDeep neural networksNeural networkDeep learning algorithmsProstate gland segmentationImage normalization methodGland segmentationLearning algorithmImage normalizationMulti-site dataIntensity normalization methodNormalization methodSingle-site dataAlgorithmNetworkPotential solutionsEquipment sourcesClinical adoptionSegmentationTrainingIntensity characteristicsRobustnessDataSite trainingMethodAdoption
2018
Deep Learning Retinal Vessel Segmentation from a Single Annotated Example: An Application of Cyclic Generative Adversarial Neural Networks
Sadda P, Onofrey J, Papademetris X. Deep Learning Retinal Vessel Segmentation from a Single Annotated Example: An Application of Cyclic Generative Adversarial Neural Networks. Lecture Notes In Computer Science 2018, 11043: 82-91. DOI: 10.1007/978-3-030-01364-6_10.Peer-Reviewed Original ResearchGenerative adversarial neural networksAdversarial neural networkGround truth segmentationNeural networkTruth segmentationMedical image segmentation tasksImage segmentation tasksConvolutional neural networkDeep learning methodsRetinal vessel segmentationConvolutional networkSegmentation taskTraining examplesAnnotated examplesTraining dataLearning methodsVessel segmentationSegmentationSynthetic examplesNetworkLarge amountDatasetTaskExampleAccuracy