Research map / Label-efficient and robust learning / Self-supervised and contrastive
Structure-aware Semantic Discrepancy and Consistency for 3D Medical Image Self-supervised Learning
Tan Pan, Zhaorui Tan, Kaiyu Guo, Dongli Xu, Weidi Xu, Chen Jiang, Xin Guo, Yuan Qi, Yuan Cheng
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add patch-to-patch loss L_p2p to contrastive baseline (ablation) | S2DC · BTCV | Dice: 83.43 -> 83.98 | Improved |
| add patch-to-structure loss L_p2s to contrastive baseline (ablation) | S2DC · BTCV | Dice: 83.43 -> 84.02 | Improved |
| add L_p2s on top of L_p2p (ablation) | S2DC · BTCV | Dice: 83.98 -> 84.14 | Improved |
| add patch-to-patch loss L_p2p to contrastive baseline (ablation) | S2DC · AUTOPET | Dice: 45.72 -> 45.85 | Improved |
| add patch-to-structure loss L_p2s to contrastive baseline (ablation) | S2DC · AUTOPET | Dice: 45.72 -> 46.23 | Improved |
| add L_p2s on top of L_p2p (ablation) | S2DC · AUTOPET | Dice: 45.85 -> 46.47 | Improved |
| S2DC pretraining versus VoCo on BTCV | Swin-B · BTCV | Dice AVG: 83.85 -> 84.14 | Improved |
| S2DC pretraining versus training from scratch on small-target AUTOPET | Swin-UNETR · AUTOPET | Dice: 34.49 -> 46.47 | Improved |
| pretrain with S2DC for CT-to-CT classification on CC-CCII versus scratch | Swin-B · CC-CCII | accuracy: 88.04 -> 95.34 | Improved |
| pretrain with S2DC versus scratch for ADNI classification | Swin-B · ADNI-cls | accuracy: 56.17 -> 56.72 | No clear effect |
| pretrain on other SSL methods versus scratch in T1->T2 translation | Swin-UNETR · BraTs23 T1->T2 | NMSE: 0.1348 -> 0.1640 | Hurt |
| S2DC pretraining versus scratch in T1->T2 translation | Swin-UNETR · BraTs23 T1->T2 | NMSE: 0.1348 -> 0.1161 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- VoCo — VoCo: A Simple-yet-Effective Volume Contrastive Learning Framework for 3D Medical Image Analysis (CVPR 2024) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Autoregressive Sequence Modeling for 3D Medical Image Representation (AAAI 2025 · similarity 0.88)
- VoCo: A Simple-yet-Effective Volume Contrastive Learning Framework for 3D Medical Image Analysis (CVPR 2024 · similarity 0.87)
- Revisiting MAE Pre-training for 3D Medical Image Segmentation (CVPR 2025 · similarity 0.87)
- Knowledge Transfer Scaling Laws for 3D Medical Imaging (NeurIPS 2026 · similarity 0.86)
- DyCON: Dynamic Uncertainty-aware Consistency and Contrastive Learning for Semi-supervised Medical Image Segmentation (CVPR 2025 · similarity 0.86)
- S2S2: Semantic Stacking for Robust Semantic Segmentation in Medical Imaging (AAAI 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
