Medical Image Segmentation2015en
Beyond The Cranial Vault Multi-Organ CT Segmentation
30 abdominal CT scans with 13 organ annotations. MICCAI 2015 challenge hosted on Synapse platform.
Current State of the Art
STU-Net-H
Ziyan Huang et al.
85.38
mean-dsc
mean-dsc Progress Over Time
Showing 3 breakthroughs from Oct 2021 to Apr 2023
Key Milestones
Jan 2022
Swin UNETR
Mean DSC on BTCV 13-organ CT. Table 3, Swin UNETR arxiv:2201.01266. Won BTCV challenge 2021.
79.1
+6.2%
Apr 2023
STU-Net-HCurrent SOTA
Mean DSC on BTCV 13-organ CT. Table 4, STU-Net-H (1.4B params) arxiv:2304.06716. Best BTCV result in paper.
85.4
+7.9%
Total Improvement
14.6%
Time Span
1y 6m
Breakthroughs
3
Current SOTA
85.4
Top Models Performance Comparison
Top 6 models ranked by mean-dsc
Best Score
85.4
Top Model
STU-Net-H
Models Compared
6
Score Range
10.9
mean-dscPrimary
| # | Model | Score | Paper / Code | Date |
|---|---|---|---|---|
| 1 | STU-Net-HOpen Source Ziyan Huang et al. | 85.38 | Apr 2023 | |
| 2 | STU-Net-LOpen Source Ziyan Huang et al. | 83.54 | Apr 2023 | |
| 3 | nnU-Net v2Open Source German Cancer Research Center (DKFZ) | 81.81 | Apr 2023 | |
| 4 | Swin UNETROpen Source NVIDIA (MONAI) | 79.13 | Jan 2022 | |
| 5 | SAM-Med3DOpen Source University Medical Center Hamburg-Eppendorf et al. | 78.94 | Aug 2023 | |
| 6 | UNETROpen Source NVIDIA (MONAI) | 74.52 | Oct 2021 |
Related Papers4
SAM-Med3D: Towards General-purpose Segmentation Models for Volumetric Medical Images
Aug 2023Models: SAM-Med3D
STU-Net: Scalable and Transferable Medical Image Segmentation Models Empowered by Large-Scale Supervised Pre-training
Apr 2023Models: STU-Net-H, STU-Net-L, nnU-Net v2
Self-Supervised Pre-Training of Swin Transformers for 3D Medical Image Analysis
Jan 2022Models: Swin UNETR
UNETR: Transformers for 3D Medical Image Segmentation
Oct 2021Models: UNETR