2024
Disentangled multimodal brain MR image translation via transformer-based modality infuser
Cho J, Liu X, Xing F, Ouyang J, Fakhri G, Park J, Woo J. Disentangled multimodal brain MR image translation via transformer-based modality infuser. Progress In Biomedical Optics And Imaging 2024, 12926: 129262h-129262h-6. DOI: 10.1117/12.3006502.Peer-Reviewed Original ResearchConvolutional neural networkBrain tumor segmentation taskModality-specific featuresTumor segmentation taskImage translationAdversarial networkSegmentation taskSynthesis qualityBrain MR imagesNeural networkMR modalitiesAcquired imagesExperimental resultsNetworkGlobal relationshipsDisease diagnosisImagesEncodingBraTSDatasetFeaturesTaskMethodSuperiorityMR imaging
2022
Posterior estimation using deep learning: a simulation study of compartmental modeling in dynamic positron emission tomography
Liu X, Marin T, Amal T, Woo J, Fakhri G, Ouyang J. Posterior estimation using deep learning: a simulation study of compartmental modeling in dynamic positron emission tomography. Medical Physics 2022, 50: 1539-1548. PMID: 36331429, PMCID: PMC10087283, DOI: 10.1002/mp.16078.Peer-Reviewed Original ResearchConceptsConditional variational auto-encoderDeep learning approachNeural networkDeep learningMarkov chain Monte CarloVariational Bayesian inference frameworkLearning approachDeep learning-based approachVariational auto-encoderDeep neural networksLearning-based approachDynamic brain PET imagingPosterior distributionEstimate posterior distributionsBayesian inference frameworkAuto-encoderMedical imagesInference frameworkNetworkSimulation studyBrain PET imagingLearningPosterior estimatesInferior performanceImagesLow-Dose Tau PET Imaging Based on Swin Restormer with Diagonally Scaled Self-Attention
Jang S, Lois C, Becker J, Thibault E, Li Y, Price J, Fakhri G, Li Q, Johnson K, Gong K. Low-Dose Tau PET Imaging Based on Swin Restormer with Diagonally Scaled Self-Attention. 2022, 00: 1-3. DOI: 10.1109/nss/mic44845.2022.10399169.Peer-Reviewed Original ResearchConvolutional neural networkSelf-attention mechanismSelf-attentionTransformer architectureComputer vision tasksLocal feature extractionLong-range informationVision tasksDenoising performanceSwin TransformerFeature extractionImage datasetsUNet structureNeural networkSwinComputational costReceptive fieldsImage qualityMap calculationNetwork structureArchitecturePET image qualityChannel dimensionsQuantitative evaluationDenoisingMeasuring strain in diffusion-weighted data using tagged magnetic resonance imaging
Xing F, Liu X, Reese T, Stone M, Wedeen V, Prince J, El Fakhri G, Woo J. Measuring strain in diffusion-weighted data using tagged magnetic resonance imaging. Proceedings Of SPIE--the International Society For Optical Engineering 2022, 12032: 1203205-1203205-7. PMID: 36777787, PMCID: PMC9911263, DOI: 10.1117/12.2610989.Peer-Reviewed Original ResearchMR spaceMotion fieldDeep neural networksEstimated motion fieldNeural networkDiffusion-weighted dataMedical imagesDynamic MR dataAlgorithm workflowStatic imagesDeformable organsTag dataData lack informationDiffeomorphic registrationEstimated strain valuesDiffusion tractographyInternal tongue musclesMR dataFiber tractographyMotion analysisTongue deformationAccurate strain measurementsMuscle fibersImaging excelsMagnetic resonanceVoxelHop: Successive Subspace Learning for ALS Disease Classification Using Structural MRI
Liu X, Xing F, Yang C, Kuo C, Babu S, Fakhri G, Jenkins T, Woo J. VoxelHop: Successive Subspace Learning for ALS Disease Classification Using Structural MRI. IEEE Journal Of Biomedical And Health Informatics 2022, 26: 1128-1139. PMID: 34339378, PMCID: PMC8807766, DOI: 10.1109/jbhi.2021.3097735.Peer-Reviewed Original ResearchConceptsConvolutional neural networkLearning modelsDimension reductionSubspace learning modelConcatenation of featuresState-of-the-artUnsupervised dimension reductionDeep learning modelsMedical image dataSupervised dimension reductionImage dataClassification of amyotrophic lateral sclerosisSubspace learningClassification taskDeep learningDataset sizeNeural networkSubspace approximationMemory requirementsTraining datasetClassification approachAUC scoreAccurate classificationDatasetExperimental resultsDeep Unsupervised Domain Adaptation: A Review of Recent Advances and Perspectives
Liu X, Yoo C, Xing F, Oh H, Fakhri G, Kang J, Woo J. Deep Unsupervised Domain Adaptation: A Review of Recent Advances and Perspectives. APSIPA Transactions On Signal And Information Processing 2022, 11: e25. DOI: 10.1561/116.00000192.Peer-Reviewed Original ResearchUnsupervised domain adaptationTarget domainLabeled source domain dataOut-of-distribution detectionUnlabeled target domain dataOut-of-distribution dataDomain dataTarget domain dataOut-of-distributionSource domain dataDeep neural networksNatural image processingMedical image analysisNatural language processingReal-world problemsDomain adaptationLabeled datasetSource domainDomain generalizationDeep learningNeural networkLanguage processingImpressive performanceTime series data analysisPerformance drop
2021
Segmentation of Cardiac Structures via Successive Subspace Learning with Saab Transform from Cine MRI
Liu X, Xing F, Gaggin H, Wang W, Kuo C, Fakhri G, Woo J. Segmentation of Cardiac Structures via Successive Subspace Learning with Saab Transform from Cine MRI. Annual International Conference Of The IEEE Engineering In Medicine And Biology Society (EMBC) 2021, 00: 3535-3538. PMID: 34892002, DOI: 10.1109/embc46164.2021.9629770.Peer-Reviewed Original ResearchConceptsConvolutional neural networkSegmentation of cardiac structuresSaab transformSubspace approximationAccurate segmentation of cardiac structuresDimension reductionConvolutional neural network modelConcatenation of featuresUnsupervised dimension reductionConditional random fieldPixel-wise classificationSupervised dimension reductionU-Net modelMachine learning modelsSubspace learningChannel-wiseSegmentation databaseSegmentation frameworkNeural networkU-NetEfficient segmentationAccurate segmentationLearning modelsCardiac MR imagesRandom fieldA deep joint sparse non-negative matrix factorization framework for identifying the common and subject-specific functional units of tongue motion during speech
Woo J, Xing F, Prince J, Stone M, Gomez A, Reese T, Wedeen V, El Fakhri G. A deep joint sparse non-negative matrix factorization framework for identifying the common and subject-specific functional units of tongue motion during speech. Medical Image Analysis 2021, 72: 102131. PMID: 34174748, PMCID: PMC8316408, DOI: 10.1016/j.media.2021.102131.Peer-Reviewed Original ResearchConceptsNon-negative matrix factorizationSparse Non-negative Matrix FactorizationIterative shrinkage-thresholding algorithmNon-negative matrix factorization frameworkDeep neural networksMatrix factorization frameworkDeep learning frameworkTongue motionIdentified functional unitsGraph regularizationClustering performanceWeight mapLearning frameworkSpectral clusteringNeural networkMatrix factorizationModular architectureIncreased interpretabilityMotion dataFactorization frameworkConvoluted natureComparison methodTagged magnetic resonance imagingMuscle coordination patternsSpeech
2019
Arterial spin labeling MR image denoising and reconstruction using unsupervised deep learning
Gong K, Han P, Fakhri G, Ma C, Li Q. Arterial spin labeling MR image denoising and reconstruction using unsupervised deep learning. NMR In Biomedicine 2019, 35: e4224. PMID: 31865615, PMCID: PMC7306418, DOI: 10.1002/nbm.4224.Peer-Reviewed Original ResearchConceptsSignal-to-noise ratioImage denoisingReconstruction frameworkDeep learning-based image denoisingDeep learning-based denoisersMR image denoisingLearning-based denoisingLow signal-to-noise ratioK-space dataNoisy imagesTraining labelsTraining pairsNetwork inputNeural networkDenoisingIn vivo experiment dataSuperior performanceImaging speedReconstruction processImage qualityLong imaging timesNetworkFrameworkImagesSpatial resolutionAttenuation correction using 3D deep convolutional neural network for brain 18F-FDG PET/MR: Comparison with Atlas, ZTE and CT based attenuation correction
Blanc-Durand P, Khalife M, Sgard B, Kaushik S, Soret M, Tiss A, Fakhri G, Habert M, Wiesinger F, Kas A. Attenuation correction using 3D deep convolutional neural network for brain 18F-FDG PET/MR: Comparison with Atlas, ZTE and CT based attenuation correction. PLOS ONE 2019, 14: e0223141. PMID: 31589623, PMCID: PMC6779234, DOI: 10.1371/journal.pone.0223141.Peer-Reviewed Original ResearchConceptsZero echo timeAC mapsAttenuation correctionPET attenuation correctionCT-based ACComputed tomographyAC methodPhoton attenuationZTE-ACInvestigation of suspected dementiaMR imagingBrain computed tomographyAtlas-ACBrain metabolismZTE-MRIConvolutional neural networkEcho timeHead atlasFDG-PET/MRPET imagingLow biasRegions-of-interestPatientsCorrectionNeural networkMAPEM-Net: an unrolled neural network for Fully 3D PET image reconstruction
Gong K, Wu D, Kim K, Yang J, Sun T, Fakhri G, Seo Y, Li Q. MAPEM-Net: an unrolled neural network for Fully 3D PET image reconstruction. Proceedings Of SPIE--the International Society For Optical Engineering 2019, 11072: 110720o-110720o-5. DOI: 10.1117/12.2534904.Peer-Reviewed Original ResearchPET image reconstructionNeural networkImage reconstructionImage denoising applicationDeep neural networksNeural network frameworkConvolutional neural networkDenoising applicationsDenoising methodNetwork frameworkUpdate stepData consistencyIll-posedNetworkClinical datasetsInverse problemMAPEMFrameworkAlgorithmDatasetDetected photonsReconstructionMethodSimulationGraph Convolutional Neural Networks For Alzheimer’s Disease Classification
Song T, Roy Chowdhury S, Yang F, Jacobs H, El Fakhri G, Li Q, Johnson K, Dutta J. Graph Convolutional Neural Networks For Alzheimer’s Disease Classification. 2011 IEEE International Symposium On Biomedical Imaging: From Nano To Macro 2019, 00: 414-417. PMID: 31327984, PMCID: PMC6641559, DOI: 10.1109/isbi.2019.8759531.Peer-Reviewed Original ResearchGraph convolutional neural networkConvolutional neural networkNeural networkCapabilities of convolutional neural networksGraph-structured dataNon-Euclidean domainsClassification capability of convolutional neural networksVector machine classifierGraph-based toolsData representationAudio signalsClassification capabilityMachine classifierClassifierPerformance gapImage dataNetworkConnected graphStructural connectivity graphsDisease classificationClassificationBrain connectivity studiesEuclidean domainsComplex systemsGraphComputational-efficient cascaded neural network for CT image reconstruction
Wu D, Kim K, Fakhri G, Li Q. Computational-efficient cascaded neural network for CT image reconstruction. Progress In Biomedical Optics And Imaging 2019, 10948: 109485z-109485z-6. DOI: 10.1117/12.2511526.Peer-Reviewed Original ResearchCascaded neural networkNeural networkImage reconstructionCT image reconstructionMemory consumptionDevelopment of deep learningDeep artificial neural networksState-of-the-artMedical image reconstructionReduce memory consumptionImage reconstruction qualitySparse-view samplingTraining ground truthUnrolling networkImage priorsImage quality improvementImage patchesReconstruction qualityDeep learningArtificial neural networkImage domainUndersampled projectionsTraining phaseTraining processParameter tuningEMnet: an unrolled deep neural network for PET image reconstruction
Gong K, Wu D, Kim K, Yang J, Fakhri G, Seo Y, Li Q. EMnet: an unrolled deep neural network for PET image reconstruction. Progress In Biomedical Optics And Imaging 2019, 10948: 1094853-1094853-6. DOI: 10.1117/12.2513096.Peer-Reviewed Original ResearchDeep neural networksPET image reconstructionNeural networkExpectation maximizationImage reconstructionImage denoising applicationNeural network frameworkNeural network denoisersDenoising applicationsDenoising methodNetwork denoisingNetwork trainingNetwork frameworkWhole graphUpdate stepData consistencyIll-posedNetworkInverse problemEMNETDenoisingSimulated dataFrameworkAlgorithmGraph
2018
Super-Resolution PET Using A Very Deep Convolutional Neural Network
Song T, Chowdhury S, Kim K, Gong K, Fakhri G, Li Q, Dutta J. Super-Resolution PET Using A Very Deep Convolutional Neural Network. 2018, 00: 1-2. DOI: 10.1109/nssmic.2018.8824683.Peer-Reviewed Original ResearchConvolutional neural networkNeural networkSuper-resolution convolutional neural networkDeep convolutional neural networkImage deblurring approachesInput image patchesBlur kernelResolution recovery techniquesSpatial location informationDeblurring approachDeblurring processImage patchesQuantitative accuracy of PETLocation informationSpatially-varying natureSuperior performanceNetworkRecovery techniquesDigital phantomDeblurringBrainWebInformationPartial volume effectsDeepBlurEnd-to-End Lung Nodule Detection in Computed Tomography
Wu D, Kim K, Dong B, Fakhri G, Li Q. End-to-End Lung Nodule Detection in Computed Tomography. Lecture Notes In Computer Science 2018, 11046: 37-45. DOI: 10.1007/978-3-030-00919-9_5.Peer-Reviewed Original ResearchDeep reconstruction networkLung nodule detectionReconstruction networkEnd-to-end detectorMedical imagesLung Image Database Consortium image collectionNodule detectionEfficient network trainingReconstructed imagesConvolutional neural networkEnd-to-endSuperior detection performanceRaw dataComputer visionCAD systemCNN detectorNetwork trainingImage collectionNeural networkDetection performanceImage spaceDetection taskDetection systemModern medical imagingFanbeam projectionsIterative PET Image Reconstruction Using Convolutional Neural Network Representation
Gong K, Guan J, Kim K, Zhang X, Yang J, Seo Y, Fakhri G, Qi J, Li Q. Iterative PET Image Reconstruction Using Convolutional Neural Network Representation. IEEE Transactions On Medical Imaging 2018, 38: 675-685. PMID: 30222554, PMCID: PMC6472985, DOI: 10.1109/tmi.2018.2869871.Peer-Reviewed Original ResearchConceptsPET image reconstructionNeural networkConvolutional neural network representationsDeep residual convolutional neural networkImage reconstructionResidual convolutional neural networkComputer vision tasksDeep neural networksConvolutional neural networkNeural network denoisersAlternating direction methodNeural network representationIterative reconstruction frameworkNeural network methodVision tasksImage representationNetwork denoisingReconstruction frameworkMultipliers algorithmMedical imagesOptimization problemNetwork methodPost-processing toolDirection methodNetwork representationDeep networks in identifying CT brain hemorrhage
Helwan A, El-Fakhri G, Sasani H, Uzun Ozsahin D. Deep networks in identifying CT brain hemorrhage. Journal Of Intelligent & Fuzzy Systems 2018, Preprint: 1-1. DOI: 10.3233/jifs-172261.Peer-Reviewed Original ResearchConvolutional neural networkStacked autoencoderDeep networksMedical image classificationDeep learning algorithmsMedical expert's experienceImage classificationTraining timeLearning algorithmsNeural networkAutoencoderExpert experienceBrain CT imagesCT imagesNetworkHigher accuracyLess errorAlgorithmImagesAccuracyErrorClassificationAttenuation correction for brain PET imaging using deep neural network based on Dixon and ZTE MR images
Gong K, Yang J, Kim K, Fakhri G, Seo Y, Li Q. Attenuation correction for brain PET imaging using deep neural network based on Dixon and ZTE MR images. Physics In Medicine And Biology 2018, 63: 125011. PMID: 29790857, PMCID: PMC6031313, DOI: 10.1088/1361-6560/aac763.Peer-Reviewed Original ResearchConceptsU-Net structureU-NetModified U-net structureAttenuation correctionDeep neural network methodBrain PET imagingPET attenuationDeep neural networksPatient data setsAttenuation coefficientDixon-based methodNeural network methodData setsConvolution moduleNetwork inputNeural networkDixon MRPET/MR hybrid systemImage reconstructionPET imagingNetwork methodNetworkNetwork approachNetwork structureQuantification errorsPenalized PET Reconstruction Using Deep Learning Prior and Local Linear Fitting
Kim K, Wu D, Gong K, Dutta J, Kim J, Son Y, Kim H, Fakhri G, Li Q. Penalized PET Reconstruction Using Deep Learning Prior and Local Linear Fitting. IEEE Transactions On Medical Imaging 2018, 37: 1478-1487. PMID: 29870375, PMCID: PMC6375088, DOI: 10.1109/tmi.2018.2832613.Peer-Reviewed Original ResearchConceptsDeep learningDenoising convolutional neural networkConvolutional neural networkDeep learning-basedPerformance of iterative reconstructionPotential of deep learningDeep networksNoise levelLearning-basedReconstruction frameworkDegradation of performanceNeural networkDnCNNMedical imagesDownsampled dataFitness functionPoisson thinningFull-dose imagesLow dose imagesNoise conditionsNetworkImage qualityPET reconstructionDose imagesDeep