/SOTA-MedSeg

SOTA medical image segmentation methods based on various challenges

Awesome medical image segmentation methods based on various challenges! (Updated 2023-12)

Overview of medical image segmentation challenges in MICCAI 2023.

For each competition, we present the segmentation target, image modality, dataset size, and the base network architecture in the winning solution. The competitions cover different modalities and segmentation targets with various challenging characteristics. U-Net and its variants still dominate the winning solutions.

miccai23

Contents

Head and Neck

  • Brain Tumor Segmentation: BraTS 2019, 2020, 2021, 2022
  • Intracranial Hemorrhage Segmentation Challenge on Non-Contrast head CT (INSTANCE)
  • Retinal Fundus Glaucoma Challenge Edition2 (REFUGE2)
  • CATARACTS Semantic Segmentation
  • Anatomical Brain Barriers to Cancer Spread: Segmentation from CT and MR Images (ABCs)
  • 3D Head and Neck Tumor Segmentation: HECKTOR 2020, 2021, 2022
  • Cerebral Aneurysm Segmentation (CADA)
  • Aneurysm Detection And segMenation Challenge 2020 (ADAM)
  • Thyroid nodule segmentation and classification challenge (TN-SCUI 2020)
  • Automatic Lung Cancer Patient Management (LNDb) (LNDb)
  • 6-month Infant Brain MRI Segmentation from Multiple Sites: iSeg2019, cSeg2022

Heart

  • Automatic Evaluation of Myocardial Infarction from Delayed-Enhancement Cardiac MRI (EMIDEC)
  • Automated Segmentation of Coronary Arteries (ASOCA) (Results)
  • MyoPS 2020: Myocardial pathology segmentation combining multi-sequence CMR (Homepage)

Chest & Abdomen

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

Ongoing Challenges

2022 MICCAI: Intracranial Hemorrhage Segmentation Challenge on Non-Contrast head CT (INSTANCE)

Date First Author Title DSC NSD RVD HD Remark
202301 Xiangyu Li The state-of-the-art 3D anisotropic intracranial hemorrhage segmentation on non-contrast head CT: The INSTANCE challenge (paper) 0.7912 0.5026 0.21 29.02 Summary paper

2022 MICCAI: Brain Tumor Segmentation (BraTS2022)

Date First Author Title ET DSC TC DSC WT DSC
202209 Ramy A. Zeineldin Multimodal CNN Networks for Brain Tumor Segmentation in MRI: A BraTS 2022 Challenge Solution (paper) 0.8438 0.8753 0.9271

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

Date First Author Title Task 1-DSC Task 1-NSD Task 2-DSC Task 2-NSD Remark
202209 Fabian Isensee, Constantin Ulrich and Tassilo Wald Extending nnU-Net is all you need (paper) (code) TBA TBA TBA TBA 1st Place in MICCAI 2022
202303 Saikat Roy MedNeXt: Transformer-driven Scaling of ConvNets for Medical Image Segmentation (paper) (code) 89.87 92.95 TBA TBA Improve nnUNet by ~1%

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

Date First Author Title MitoEM-R MitoEM-H Average Remark
202104 Mingxing Li Advanced Deep Networks for 3D Mitochondria Instance Segmentation (paper) (code) 0.851 0.829 0.840 1st Place in ISBI 2021

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

Date First Author Title DSC NSD Time GPU Memory Remark
202110 Fan Zhang Efficient Context-Aware Network for Abdominal Multi-organ Segmentation (paper) (code) 0.895 0.796 9.32 1177 1st Place in MICCAI 2021

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

Date First Author Title DSC NSD Remark
202110 Zhaozhong Chen A Coarse-to-fine Framework for The 2021 Kidney and Kidney Tumor Segmentation Challenge (paper) 0.9077 0.8262 1st Place in MICCAI 2021

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

Date First Author Title IoU HD MD Remark
20201008 Mediclouds TBA 0.758 2.866 1.618 1st Place in MICCAI 2020
20201008 Jun Ma Exploring Large Context for Cerebral Aneurysm Segmentation (arxiv) (Code) 0.759 4.967 3.535 2nd Place in MICCAI 2020

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

Date First Author Title Myo Infarction Re-flow Remark
20201008 Yichi Zhang Cascaded Convolutional Neural Network for Automatic Myocardial Infarction Segmentation from Delayed-Enhancement Cardiac MRI (arxiv) 0.8786 0.7124 0.7851 1st Place in MICCAI 2020
20201008 Jun Ma Cascaded Framework for Automatic Evaluation of Myocardial Infarction from Delayed-Enhancement Cardiac MRI (arxiv) 0.8628 0.6224 0.7776 2nd Place in MICCAI 2020
20201008 Xue Feng Automatic Scar Segmentation from DE-MRI Using 2D Dilated UNet with Rotation-based Augmentation (paper) 0.8356 0.4568 0.7222 3rd Place in MICCAI 2020

Metrics: DSC

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

Date First Author Title DSC MHD VS Remark
20201008 Jun Ma Loss Ensembles for Intracranial Aneurysm Segmentation: An Embarrassingly Simple Method (Code) 0.41 8.96 0.50 1st Place in MICCAI 2020
20201008 Yuexiang Li Automatic Aneurysm Segmentation via 3D U-Net Ensemble 0.40 8.67 0.48 2nd Place in MICCAI 2020
20201008 Riccardo De Feo Multi-loss CNN ensemblesfor aneurysm segmentation 0.28 18.13 0.39 3rd Place in MICCAI 2020

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

Date First Author Title LV MYO RV Remark
20201004 Peter Full The effect of Data Augmentation on Robustness against Domain Shifts in cMRI Segmentation 0.910 0.849 0.884 1st Place in MICCAI 2020
20201004 Yao Zhang Semi-Supervised Cardiac Image Segmentation via Label Propagation and Style Transfer 0.906 0.840 0.878 2nd Place in MICCAI 2020
20201004 Jun Ma Histogram Matching Augmentation for Domain Adaptation (code) 0.902 0.835 0.874 3rd 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)

Date First Author Title DSC Remark
20201004 Andrei Iantsen Squeeze-and-Excitation Normalization for Automated Delineation of Head and Neck Primary Tumors in Combined PET and CT Images (paper) 0.759 1st Place in MICCAI 2020
20201004 Jun Ma Combining CNN and Hybrid Active Contours for Head and Neck Tumor Segmentation in CT and PET Images (paper) 0.752 2nd Place in MICCAI 2020

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

Date First Author Title IoU Remark
20201004 Mingyu Wang A Simple Cascaded Framework for Automatically Segmenting Thyroid Nodules (code) 0.8254 1st Place in MICCAI 2020
20201004 Huai Chen LRTHR-Net: A Low-Resolution-to-High-Resolution Framework to Iteratively Refine the Segmentation of Thyroid Nodule in Ultrasound Images 0.8196 2nd Place in MICCAI 2020
20201004 Zhe Tang Coarse to Fine Ensemble Network for Thyroid Nodule Segmentation 0.8194 3rd Place in MICCAI 2020

Video records are available on pathable

Endoscopy Computer Vision Challenge (EndoCV2020)

Date First Author Title Avg F1 and F2 Remark
202004 Vajira Thambawita DivergentNets: Medical Image Segmentation by Network Ensemble (paper) (code) 0.823 1st Place in ISBI EndoCV 2020

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

Results

Date First Author Title IoU Remark
20200625 Alexandr G. Rassadin Deep Residual 3D U-Net for Joint Segmentation and Texture Classification of Nodules in Lung (arxiv) 0.5221 1st Place in Seg. Task

Challenges on Open Leaderboard Phase

2019 MICCAI: Kidney Tumor Segmentation Challenge (KiTS19)

Leaderboard (2019/07/30)

Date First Author Title Composite Dice Kidney Dice Tumor Dice
202004 Fabian Isensee Automated Design of Deep Learning Methods for Biomedical Image Segmentation (arxiv) 0.9168 0.9793 0.8542
20190730 Fabian Isensee An attempt at beating the 3D U-Net (paper) 0.9123 0.9737 0.8509
20190730 Xiaoshuai Hou Cascaded Semantic Segmentation for Kidney and Tumor (paper) 0.9064 0.9674 0.8454
20190730 Guangrui Mu Segmentation of kidney tumor by multi-resolution VB-nets (paper) 0.9025 0.9729 0.8321

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

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

Date First Author Title Liver Per Case Dice Liver Global Dice Tumor Per Case Dice Tumor Global Dice
202004 Fabian Isensee Automated Design of Deep Learning Methods for Biomedical Image Segmentation (arxiv) 0.967 0.970 0.763 0.858
201909 Xudong Wang Volumetric Attention for 3D Medical Image Segmentation and Detection (MICCAI2019) - - 0.741 -
201908 Jianpeng Zhang Light-Weight Hybrid Convolutional Network for Liver Tumor Segmentation (IJCAI 2019) 0.965 0.968 0.730 0.820
202007 Youbao Tang E^2Net: An Edge Enhanced Network for Accurate Liver and Tumor Segmentation on CT Scans (arXiv) 0.966 0.968 0.724 0.829
201709 Xiaomeng Li H-DenseUNet: Hybrid Densely Connected UNet for Liver and Tumor Segmentation from CT Volumes, (TMI), (Keras code) 0.961 0.965 0.722 0.824

2012 MICCAI: Prostate MR Image Segmentation (PROMISE12)

Date First Author Title Whole Dice Overall Score
201904 Anonymous 3D segmentation and 2D boundary network (paper) - 90.34
201902 Qikui Zhu Boundary-weighted Domain Adaptive Neural Network for Prostate MR Image Segmentation (paper) 91.41 89.59

Others

Recent results can be found here.

Task Data Info Fabian Isensee et al. (paper) nnUNet v2 Qihang Yu et al. (paper)
Brats Multimodal multisite MRI data (FLAIR, T1w, T1gd,T2w), (484 Training + 266 Testing) 0.68/0.48/0.68 68/46.8/68.46 67.6/48.6/69.7
Heart Mono-modal MRI (20 Training + 10 Testing) 0.93 96.74 92.49
Hippocampus head and body Mono-modal MRI (263 Training + 131 Testing) 0.90/0.89 90/88.69 89.37/87.96
Liver & Tumor Portal venous phase CT (131 Training + 70 Testing) 0.95/0.74 95.75/75.97 94.98/72.89
Lung CT (64 Training + 32 Testing) 0.69 73.97 70.44
Pancreas & Tumor Portal venous phase CT (282 Training +139 Testing) 0.80/0.52 81.64/52.78 80.76/54.41
Prostate central gland and peripheral Multimodal MR (T2, ADC) (32 Training + 16 Testing) 0.76/0.90 76.59/89.62 74.88/88.75
Hepatic vessel& Tumor CT, (303 Training + 140 Testing) 0.63/0.69 66.46/71.78 64.73/71
Spleen CT (41 Training + 20 Testing) 0.96 97.43 96.28
Colon CT (41 Training + 20 Testing) 0.56 58.33 58.90

Only showing Dice Score.

Recent papers on Medical Segmentation Decathlon

Date First Author Title Score
20181129 Yingda Xia 3D Semi-Supervised Learning with Uncertainty-Aware Multi-View Co-Training (paper) no test set score
20190606 Zhuotun Zhu V-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)

Date First Author Title Scar Scar+Edema Remark
20201004 Shuwei Zhai Myocardial 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

Date First Author Title Head & Neck OAR Head & Neck GTV Chest OAR Chest GTV
20191001 Huai Chen TBD 0.8109 0.6666 0.9011 0.5406
20191001 Fabian Isensee nnU-Net 0.7988 0.6398 0.9083 0.5343
20191001 Yujin Hu TBD 0.7956 0.6245 0.9024 0.5447
20191001 Xuechen Liu TBD - - 0.9066 -

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

Multi-sequence ventricle and myocardium segmentation.

Date First Author Title LV Myo RV
20190821 Chen Chen Unsupervised Multi-modal Style Transfer for Cardiac MR Segmentation (arxiv) 0.92 0.83 0.88
Date First Author Title Dice
20190905 Aimoldin Anuar SIIM-ACR Pneumothorax Challenge - 1st place solution (pytorch) 0.8679

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

Date First Author Title Esophagus Heart Trachea Aorta
20190320 Miaofei Han Segmentation of CT thoracic organs by multi-resolution VB-nets (paper) 86 95 92 94
20190606 Shadab Khan Extreme Points Derived Confidence Map as a Cue For Class-Agnostic Segmentation Using Deep Neural Network (paper) 89.87 95.97 91.87 94

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 Author Title Val. WT/EN/TC Dice Test Val. WT/ET/TC Dice
1 Andriy Myronenko 3D MRI Brain Tumor Segmentation Using Autoencoder Regularization (paper) 0.91/0.823/0.867 0.884/0.766/0.815
2 Fabian Isensee No New-Net (paper) 0.913/0.809/0.863 0.878/0.779/0.806
3 Richard McKinley Ensembles of Densely-Connected CNNs with Label-Uncertainty for Brain Tumor Segmentation (paper) 0.903/0.796/0.847 0.886/0.732/0.799
3 Chenhong Zhou Learning Contextual and Attentive Information for Brain Tumor Segmentation (paper) 0.9095/0.8136/0.8651 0.8842/0.7775/0.7960
New Xuhua Ren Task Decomposition and Synchronization for Semantic Biomedical Image Segmentation (paper) 0.915/0.832/0.883 -

2018 MICCAI: Ischemic stroke lesion segmentation (ISLES )

Date First Author Title Dice
20190605 Yu Chen OctopusNet: A Deep Learning Segmentation Network for Multi-modal Medical Images (paper) 57.90 (5-fold CV)
201812 Hoel Kervadec Boundary loss for highly unbalanced segmentation (paper), (pytorch 1.0 code) 65.6
201809 Tao Song 3D Multi-scale U-Net with Atrous Convolution for Ischemic Stroke Lesion Segmentation, (paper) 55.86
201809 Pengbo Liu Stroke Lesion Segmentation with 2D Convolutional Neutral Network and Novel Loss Function, (paper) 55.23
201809 Yu Chen Ensembles 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)
Rank First Author Title Score
1 Miguel Luna 3D Patchwise U-Net with Transition Layers for MR Brain Segmentation (paper) 9.971
2 Alireza Mehrtash U-Net with various input combinations (paper) 9.915
3 Xuhua Ren Ensembles of Multiple Scales, Losses and Models for Segmentation of Brain Area (paper) 9.872
201906 Xuhua Ren Brain MR Image Segmentation in Small Dataset with Adversarial Defense and Task Reorganization (arxiv ) 5 fold CV Dice: 84.46
  • Three Label Segmentation Results (201809)
Rank First Author Title GM/WM/CSF Dice Score
1 Liyan Sun Brain Tissue Segmentation Using 3D FCN with Multi-modality Spatial Attention (paper) 0.86/0.889/0.850 11.272

2018 MICCAI: Left Ventricle Full Quantification Challenge (LVQuan18)

Rank First Author Title
1 Jiahui Li Left Ventricle Full Quantification Using Deep Layer Aggregation Based Multitask Relationship Learning, (paper)
2 Eric Kerfoot Left-Ventricle Quantification Using Residual U-Net, (paper)
3 Fumin Guo Cardiac MRI Left Ventricle Segmentation and Quantification: A Framework Combining U-Net and Continuous Max-Flow (paper)

2018 MICCAI: Atrial Segmentation Challenge (AtriaSeg)

Rank First Author Title Score
1 Qing Xia Automatic 3D Atrial Segmentation from GE-MRIs Using Volumetric Fully Convolutional Networks (paper) 0.932
2 Cheng Bian Pyramid Network with Online Hard Example Mining for Accurate Left Atrium Segmentation (paper) 0.926
2 Sulaiman Vesal Dilated Convolutions in Neural Networks for Left Atrial Segmentation in 3D Gadolinium Enhanced-MR (paper) 0.926

Awesome Open Source Tools

Task First Author Title Notes
Detection&Segmentation Paul F. Jaeger Retina U-Net: Embarrassingly Simple Exploitation of Segmentation Supervision for Medical Object Detection, (paper), (code) pytorch
Medical Image Analysis Many excellent contributors MONAI: Medical Open Network for AI (code) pytorch
Segmentation Christian S. Perone MedicalTorch pytorch
Segmentation Fabian Isensee nnU-Net (paper) (code) pytorch
MedImgIO Fernando Pérez García TorchIO: tools for loading, augmenting and writing 3D medical images on PyTorch (code) pytorch
Segmentation DLinRadiology MegSeg: a free segmentation tool for radiological images (CT and MRI) homepage
Segmentation Adaloglou Nikolaos A 3D multi-modal medical image segmentation library in PyTorch (code) pytorch

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

Contribute

Contributions are most welcome!

⬆ back to top