SOTA medical image segmentation methods based on various challenges

  • A+
所属分类:医信大学

State-of-the-art medical image segmentation methods based on various challenges! (Updated 2022-01)

Contents

Head and Neck

  • 2020 MICCAI: Retinal Fundus Glaucoma Challenge Edition2 (REFUGE2) (Results)
  • 2020 MICCAI: Brain Tumor Segmentation Challenge (BraTS) (Results)
  • 2020 MICCAI: CATARACTS Semantic Segmentation
  • 2020 MICCAI: Anatomical Brain Barriers to Cancer Spread: Segmentation from CT and MR Images (ABCs) (Results)
  • 2020 MICCAI: 3D Head and Neck Tumor Segmentation (HECKTOR) (Results)
  • 2020 MICCAI: Cerebral Aneurysm Segmentation (CADA) (Results)
  • 2020 MICCAI: Aneurysm Detection And segMenation Challenge 2020 (ADAM) (Results)
  • 2020 MICCAI: Thyroid nodule segmentation and classification challenge (TN-SCUI 2020)(Results)
  • 2020 ICIAR: Automatic Lung Cancer Patient Management (LNDb) (LNDb)
  • 2019 MICCAI: Multimodal Brain Tumor Segmentation Challenge (BraTS2019) (Results)
  • 2019 MICCAI: 6-month Infant Brain MRI Segmentation from Multiple Sites (iSeg2019) (Results)
  • 2019 MICCAI: Automatic Structure Segmentation for Radiotherapy Planning Challenge (Results)
  • 2018 MICCAI: Multimodal Brain Tumor Segmentation Challenge
  • 2018 MICCAI: Ischemic stroke lesion segmentation
  • 2018 MICCAI Grand Challenge on MR Brain Image Segmentation

Chest & Abdomen

  • 2022 MICCAI: Multi-Modality Abdominal Multi-Organ Segmentation Challenge (AMOS22) (Results)
  • 2021 MICCAI: Kidney Tumor Segmentation Challenge (KiTS21) (Results)
  • 2020 MICCAI: Large Scale Vertebrae Segmentation Challenge (VerSe2020) (Results)
  • 2020 MICCAI: Automatic Evaluation of Myocardial Infarction from Delayed-Enhancement Cardiac MRI (EMIDEC)
  • 2020 MICCAI: Automated Segmentation of Coronary Arteries (ASOCA) (Results)
  • 2020 MICCAI: MyoPS 2020: Myocardial pathology segmentation combining multi-sequence CMR (Homepage)
  • 2019 MICCAI: VerSe2019: Large Scale Vertebrae Segmentation Challenge (Results)
  • 2019 MICCAI: Multi-sequence Cardiac MR Segmentation Challenge
  • 2018 MICCAI: Left Ventricle Full Quantification Challenge 
  • 2018 MICCAI: Atrial Segmentation Challenge
  • 2019 MICCAI: Kidney Tumor Segmentation Challenge
  • 2019 ISBI: Segmentation of THoracic Organs at Risk in CT images
  • 2017 ISBI & MICCAI: Liver tumor segmentation challenge 
  • 2012 MICCAI: Prostate MR Image Segmentation 

Others

  • 2018 MICCAI: Medical Segmentation Decathlon (MSD) (Results)
  • 2020 MICCAI: Quantification of Uncertainties in Biomedical Image Quantification Challenge (QUBIQ) (Results)
  • Awesome Open Source Tools
  • Loss Odyssey in Medical Image Segmentation

Survey

@article{Ma-SOTASeg2020,
  title={Cutting-edge 3D Medical Image Segmentation Methods in 2020: Are Happy Families All Alike?},
  author={Jun Ma},
  journal={arXiv preprint arXiv:2101.00232},
  year={2021}
}

Ongoing Challenges

2022 MICCAI: Multi-Modality Abdominal Multi-Organ Segmentation Challenge (AMOS22) (Results)

DateFirst AuthorTitleTask 1-DSCTask 1-NSDTask 2-DSCTask 2-NSDRemark
202209Fabian Isensee, Constantin Ulrich and Tassilo WaldExtending nnU-Net is all you need (paper) (code)TBATBATBATBA1st Place in MICCAI 2022

2021 ISBI: MitoEM Challenge: Large-scale 3D Mitochondria Instance Segmentation (MitoEM) (Results)

DateFirst AuthorTitleMitoEM-RMitoEM-HAverageRemark
202104Mingxing LiAdvanced Deep Networks for 3D Mitochondria Instance Segmentation (paper) (code)0.8510.8290.8401st Place in ISBI 2021

2021 MICCAI: Fast and Low GPU memory Abdominal oRgan sEgmentation (FLARE) (Results)

DateFirst AuthorTitleDSCNSDTimeGPU MemoryRemark
202110Fan ZhangEfficient Context-Aware Network for Abdominal Multi-organ Segmentation (paper) (code)0.8950.7969.3211771st Place in MICCAI 2021

2021 MICCAI: Kidney Tumor Segmentation Challenge (KiTS) (Results)

DateFirst AuthorTitleDSCNSDRemark
202110Zhaozhong ChenA Coarse-to-fine Framework for The 2021 Kidney and Kidney Tumor Segmentation Challenge (paper)0.90770.82621st Place in MICCAI 2021

2020 MICCAI: Cerebral Aneurysm Segmentation (CADA) (Results)

DateFirst AuthorTitleIoUHDMDRemark
20201008MedicloudsTBA0.7582.8661.6181st Place in MICCAI 2020
20201008Jun MaExploring Large Context for Cerebral Aneurysm Segmentation (arxiv) (Code)0.7594.9673.5352nd Place in MICCAI 2020

2020 MICCAI: Automatic Evaluation of Myocardial Infarction from Delayed-Enhancement Cardiac MRI (EMIDEC)

DateFirst AuthorTitleMyoInfarctionRe-flowRemark
20201008Yichi ZhangCascaded Convolutional Neural Network for Automatic Myocardial Infarction Segmentation from Delayed-Enhancement Cardiac MRI (arxiv)0.87860.71240.78511st Place in MICCAI 2020
20201008Jun MaCascaded Framework for Automatic Evaluation of Myocardial Infarction from Delayed-Enhancement Cardiac MRI (arxiv)0.86280.62240.77762nd Place in MICCAI 2020
20201008Xue FengAutomatic Scar Segmentation from DE-MRI Using 2D Dilated UNet with Rotation-based Augmentation (paper)0.83560.45680.72223rd Place in MICCAI 2020

Metrics: DSC

Aneurysm Detection And segMenation Challenge 2020 (ADAM) (Results)

DateFirst AuthorTitleDSCMHDVSRemark
20201008Jun MaLoss Ensembles for Intracranial Aneurysm Segmentation: An Embarrassingly Simple Method (Code)0.418.960.501st Place in MICCAI 2020
20201008Yuexiang LiAutomatic Aneurysm Segmentation via 3D U-Net Ensemble0.408.670.482nd Place in MICCAI 2020
20201008Riccardo De FeoMulti-loss CNN ensemblesfor aneurysm segmentation0.2818.130.393rd Place in MICCAI 2020

Multi-Centre, Multi-Vendor & Multi-Disease Cardiac Image Segmentation Challenge (M&Ms) (Results)

DateFirst AuthorTitleLVMYORVRemark
20201004Peter FullThe effect of Data Augmentation on Robustness against Domain Shifts in cMRI Segmentation0.9100.8490.8841st Place in MICCAI 2020
20201004Yao ZhangSemi-Supervised Cardiac Image Segmentation via Label Propagation and Style Transfer0.9060.8400.8782nd Place in MICCAI 2020
20201004Jun MaHistogram Matching Augmentation for Domain Adaptation (code)0.9020.8350.8743rd Place in MICCAI 2020

Dice values are reported. Video records are available on pathable. All the papers are in press

2020 MICCAI: 3D Head and Neck Tumor Segmentation in PET/CT (HECKTOR 2020)(Results)

DateFirst AuthorTitleDSCRemark
20201004Andrei IantsenSqueeze-and-Excitation Normalization for Automated Delineation of Head and Neck Primary Tumors in Combined PET and CT Images (paper)0.7591st Place in MICCAI 2020
20201004Jun MaCombining CNN and Hybrid Active Contours for Head and Neck Tumor Segmentation in CT and PET Images (paper)0.7522nd Place in MICCAI 2020

2020 MICCAI: Thyroid nodule segmentation and classification challenge (TN-SCUI 2020)(Results)

DateFirst AuthorTitleIoURemark
20201004Mingyu WangA Simple Cascaded Framework for Automatically Segmenting Thyroid Nodules (code)0.82541st Place in MICCAI 2020
20201004Huai ChenLRTHR-Net: A Low-Resolution-to-High-Resolution Framework to Iteratively Refine the Segmentation of Thyroid Nodule in Ultrasound Images0.81962nd Place in MICCAI 2020
20201004Zhe TangCoarse to Fine Ensemble Network for Thyroid Nodule Segmentation0.81943rd Place in MICCAI 2020

Video records are available on pathable

Endoscopy Computer Vision Challenge (EndoCV2020)

DateFirst AuthorTitleAvg F1 and F2Remark
202004Vajira ThambawitaDivergentNets: Medical Image Segmentation by Network Ensemble (paper) (code)0.8231st Place in ISBI EndoCV 2020

2020 ICIAR: Automatic Lung Cancer Patient Management (LNDb) (LNDb)

Results

DateFirst AuthorTitleIoURemark
20200625Alexandr G. RassadinDeep Residual 3D U-Net for Joint Segmentation and Texture Classification of Nodules in Lung (arxiv)0.52211st Place in Seg. Task

Challenges on Open Leaderboard Phase

2019 MICCAI: Kidney Tumor Segmentation Challenge (KiTS19)

Leaderboard (2019/07/30)

DateFirst AuthorTitleComposite DiceKidney DiceTumor Dice
202004Fabian IsenseeAutomated Design of Deep Learning Methods for Biomedical Image Segmentation (arxiv)0.91680.97930.8542
20190730Fabian IsenseeAn attempt at beating the 3D U-Net (paper)0.91230.97370.8509
20190730Xiaoshuai HouCascaded Semantic Segmentation for Kidney and Tumor (paper)0.90640.96740.8454
20190730Guangrui MuSegmentation of kidney tumor by multi-resolution VB-nets (paper)0.90250.97290.8321

2017 ISBI & MICCAI: Liver tumor segmentation challenge (LiTS)

Summary: The Liver Tumor Segmentation Benchmark (LiTS), Patrick Bilic et al. 201901 (arxiv)

DateFirst AuthorTitleLiver Per Case DiceLiver Global DiceTumor Per Case DiceTumor Global Dice
202004Fabian IsenseeAutomated Design of Deep Learning Methods for Biomedical Image Segmentation (arxiv)0.9670.9700.7630.858
201909Xudong WangVolumetric Attention for 3D Medical Image Segmentation and Detection (MICCAI2019)--0.741-
201908Jianpeng ZhangLight-Weight Hybrid Convolutional Network for Liver Tumor Segmentation (IJCAI 2019)0.9650.9680.7300.820
202007Youbao TangE^2Net: An Edge Enhanced Network for Accurate Liver and Tumor Segmentation on CT Scans (arXiv)0.9660.9680.7240.829
201709Xiaomeng LiH-DenseUNet: Hybrid Densely Connected UNet for Liver and Tumor Segmentation from CT Volumes, (TMI)(Keras code)0.9610.9650.7220.824

2012 MICCAI: Prostate MR Image Segmentation (PROMISE12)

DateFirst AuthorTitleWhole DiceOverall Score
201904Anonymous3D segmentation and 2D boundary network (paper)-90.34
201902Qikui ZhuBoundary-weighted Domain Adaptive Neural Network for Prostate MR Image Segmentation (paper)91.4189.59

Others

2018 MICCAI Medical Segmentation Decathlon

Recent results can be found here.

TaskData InfoFabian Isensee et al. (paper)nnUNet v2Qihang Yu et al. (paper)
BratsMultimodal multisite MRI data (FLAIR, T1w, T1gd,T2w), (484 Training + 266 Testing)0.68/0.48/0.6868/46.8/68.4667.6/48.6/69.7
HeartMono-modal MRI (20 Training + 10 Testing)0.9396.7492.49
Hippocampus head and bodyMono-modal MRI (263 Training + 131 Testing)0.90/0.8990/88.6989.37/87.96
Liver & TumorPortal venous phase CT (131 Training + 70 Testing)0.95/0.7495.75/75.9794.98/72.89
LungCT (64 Training + 32 Testing)0.6973.9770.44
Pancreas & TumorPortal venous phase CT (282 Training +139 Testing)0.80/0.5281.64/52.7880.76/54.41
Prostate central gland and peripheralMultimodal MR (T2, ADC) (32 Training + 16 Testing)0.76/0.9076.59/89.6274.88/88.75
Hepatic vessel& TumorCT, (303 Training + 140 Testing)0.63/0.6966.46/71.7864.73/71
SpleenCT (41 Training + 20 Testing)0.9697.4396.28
ColonCT (41 Training + 20 Testing)0.5658.3358.90

Only showing Dice Score.

Recent papers on Medical Segmentation Decathlon

DateFirst AuthorTitleScore
20181129Yingda Xia3D Semi-Supervised Learning with Uncertainty-Aware Multi-View Co-Training (paper)no test set score
20190606Zhuotun ZhuV-NAS: Neural Architecture Search for Volumetric Medical Image Segmentation (arxiv)Lung tumor: 55.27; Pancreas and tumor: 79.94, 37.78 (4-fold CV)

Past Challenges (New submission closed)

2020 MICCAI-MyoPS: Myocardial pathology segmentation combining multi-sequence CMR (MyoPS 2020)

DateFirst AuthorTitleScarScar+EdemaRemark
20201004Shuwei ZhaiMyocardial Edema and Scar Segmentation using a Coarse-to-Fine Framework with Weighted Ensemble (paper in press)0.672 (0.244)0.731 (0.109)1st Place in MICCAI 2020

2019 MICCAI: Structure Segmentation for Radiotherapy Planning (StructSeg)

Results

DateFirst AuthorTitleHead & Neck OARHead & Neck GTVChest OARChest GTV
20191001Huai ChenTBD0.81090.66660.90110.5406
20191001Fabian IsenseennU-Net0.79880.63980.90830.5343
20191001Yujin HuTBD0.79560.62450.90240.5447
20191001Xuechen LiuTBD--0.9066-

2019 MICCAI: Multi-sequence Cardiac MR Segmentation Challenge (MS-CMRSeg)

Multi-sequence ventricle and myocardium segmentation.

DateFirst AuthorTitleLVMyoRV
20190821Chen ChenUnsupervised Multi-modal Style Transfer for Cardiac MR Segmentation (arxiv)0.920.830.88

2019 Kaggle SIIM-ACR Pneumothorax Segmentation

DateFirst AuthorTitleDice
20190905Aimoldin AnuarSIIM-ACR Pneumothorax Challenge - 1st place solution (pytorch)0.8679

2019 ISBI: Segmentation of THoracic Organs at Risk in CT images (SegTHOR)

DateFirst AuthorTitleEsophagusHeartTracheaAorta
20190320Miaofei HanSegmentation of CT thoracic organs by multi-resolution VB-nets (paper)86959294
20190606Shadab KhanExtreme Points Derived Confidence Map as a Cue For Class-Agnostic Segmentation Using Deep Neural Network (paper)89.8795.9791.8794

Challenge results

2018 MICCAI: Multimodal Brain Tumor Segmentation Challenge(BraTS)

Summary: Identifying the Best Machine Learning Algorithms for Brain Tumor Segmentation, Progression Assessment, and Overall Survival Prediction in the BRATS Challenge Spyridon Bakas et al. 201811, (arxiv)

Rank(18)First AuthorTitleVal. WT/EN/TC DiceTest Val. WT/ET/TC Dice
1Andriy Myronenko3D MRI Brain Tumor Segmentation Using Autoencoder Regularization (paper)0.91/0.823/0.8670.884/0.766/0.815
2Fabian IsenseeNo New-Net (paper)0.913/0.809/0.8630.878/0.779/0.806
3Richard McKinleyEnsembles of Densely-Connected CNNs with Label-Uncertainty for Brain Tumor Segmentation (paper)0.903/0.796/0.8470.886/0.732/0.799
3Chenhong ZhouLearning Contextual and Attentive Information for Brain Tumor Segmentation (paper)0.9095/0.8136/0.86510.8842/0.7775/0.7960
NewXuhua RenTask Decomposition and Synchronization for Semantic Biomedical Image Segmentation (paper)0.915/0.832/0.883-

2018 MICCAI: Ischemic stroke lesion segmentation (ISLES )

DateFirst AuthorTitleDice
20190605Yu ChenOctopusNet: A Deep Learning Segmentation Network for Multi-modal Medical Images (paper)57.90 (5-fold CV)
201812Hoel KervadecBoundary loss for highly unbalanced segmentation (paper)(pytorch 1.0 code)65.6
201809Tao Song3D Multi-scale U-Net with Atrous Convolution for Ischemic Stroke Lesion Segmentation, (paper)55.86
201809Pengbo LiuStroke Lesion Segmentation with 2D Convolutional Neutral Network and Novel Loss Function, (paper)55.23
201809Yu ChenEnsembles of Modalities Fused Model for Ischemic Stroke Lesion Segmentation, (paper)-

2018 MICCAI Grand Challenge on MR Brain Image Segmentation (MRBrainS18)

  • Eight Label Segmentation Results (201809)
RankFirst AuthorTitleScore
1Miguel Luna3D Patchwise U-Net with Transition Layers for MR Brain Segmentation (paper)9.971
2Alireza MehrtashU-Net with various input combinations (paper)9.915
3Xuhua RenEnsembles of Multiple Scales, Losses and Models for Segmentation of Brain Area (paper)9.872
201906Xuhua RenBrain MR Image Segmentation in Small Dataset with Adversarial Defense and Task Reorganization (arxiv )5 fold CV Dice: 84.46
  • Three Label Segmentation Results (201809)
RankFirst AuthorTitleGM/WM/CSF DiceScore
1Liyan SunBrain Tissue Segmentation Using 3D FCN with Multi-modality Spatial Attention (paper)0.86/0.889/0.85011.272

2018 MICCAI: Left Ventricle Full Quantification Challenge (LVQuan18)

RankFirst AuthorTitle
1Jiahui LiLeft Ventricle Full Quantification Using Deep Layer Aggregation Based Multitask Relationship Learning, (paper)
2Eric KerfootLeft-Ventricle Quantification Using Residual U-Net, (paper)
3Fumin GuoCardiac MRI Left Ventricle Segmentation and Quantification: A Framework Combining U-Net and Continuous Max-Flow (paper)

2018 MICCAI: Atrial Segmentation Challenge (AtriaSeg)

RankFirst AuthorTitleScore
1Qing XiaAutomatic 3D Atrial Segmentation from GE-MRIs Using Volumetric Fully Convolutional Networks (paper)0.932
2Cheng BianPyramid Network with Online Hard Example Mining for Accurate Left Atrium Segmentation (paper)0.926
2Sulaiman VesalDilated Convolutions in Neural Networks for Left Atrial Segmentation in 3D Gadolinium Enhanced-MR (paper)0.926

Awesome Open Source Tools

TaskFirst AuthorTitleNotes
Detection&SegmentationPaul F. JaegerRetina U-Net: Embarrassingly Simple Exploitation of Segmentation Supervision for Medical Object Detection, (paper)(code)pytorch
Medical Image AnalysisMany excellent contributorsMONAI: Medical Open Network for AI (code)pytorch
SegmentationChristian S. PeroneMedicalTorchpytorch
SegmentationFabian IsenseennU-Net (paper) (code)pytorch
MedImgIOFernando Pérez GarcíaTorchIO: tools for loading, augmenting and writing 3D medical images on PyTorch (code)pytorch
SegmentationDLinRadiologyMegSeg: a free segmentation tool for radiological images (CT and MRI)homepage
SegmentationAdaloglou NikolaosA 3D multi-modal medical image segmentation library in PyTorch (code)pytorch

Segmentation Loss Odyssey (paper & code)](https://github.com/JunMa11/SegLoss)

  • HealthIT.CN小程序
  • 产学研医创新平台
  • weinxin
  • 健康数据
  • HealthIT公众号
  • weinxin
admin

发表评论

:?: :razz: :sad: :evil: :!: :smile: :oops: :grin: :eek: :shock: :???: :cool: :lol: :mad: :twisted: :roll: :wink: :idea: :arrow: :neutral: :cry: :mrgreen: