Research map / Label-efficient and robust learning / Self-supervised and contrastive
InvCoSS: Inversion-driven Continual Self-supervised Learning in Medical Multi-modal Image Pre-training
Zihao Luo, Shaohao Rui, Zhenyu Tang, Guotai Wang, Xiaosong Wang
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 total-variation image prior Limg (ablation) | InvCoSS · six downstream tasks | AVG up: 86.12 -> 86.75 | Improved |
| add total-variation image prior Limg (ablation) | InvCoSS · six downstream tasks | AVG down: 27.12 -> 24.29 | Improved |
| add repulsive representation loss Lrep (ablation) | InvCoSS · six downstream tasks | AVG up: 86.75 -> 88.03 | Improved |
| add repulsive representation loss Lrep (ablation) | InvCoSS · RICORD | ACC: 82.14 -> 84.52 | Improved |
| add Lrep (ablation) | InvCoSS · six downstream tasks | AVG down: 24.29 -> 20.11 | Improved |
| use InvUNet instead of bottom-to-top generator (ablation) | InvCoSS · ChestXR (generator comparison) | ACC: 90.77 -> 92.27 | Improved |
| use InvUNet instead of larger bottom-to-top generator (ablation) | InvCoSS · ChestXR (generator comparison) | ACC: 91.21 -> 92.27 | Improved |
| add generator G (InvUNet) vs direct image optimization (ablation) | InvCoSS · QaTa | DSC: 74.42 -> 76.23 | Improved |
| add generator G for 3D data (ablation) | InvCoSS · CT and MRI tasks | feasibility | Improved |
| replace data replay with inversion-synthesized images | InvCoSS vs ER · nine downstream tasks | AVG up: 85.45 -> 89.17 | Improved |
| replace data replay with inversion-synthesized images | InvCoSS vs MedCoSS · nine downstream tasks | AVG up: 89.03 -> 89.17 | No clear effect |
| use InvCoSS instead of data-free CaSSLe | CSSL · nine downstream tasks | AVG up: 86.88 -> 89.17 | Improved |
Closest papers by meaning
- Continual Self-supervised Learning: Towards Universal Multi-modal Medical Data Representation Learning (CVPR 2024 · similarity 0.88)
- Model Inversion with Layer-Specific Modeling and Alignment for Data-Free Continual Learning (NeurIPS 2025 · similarity 0.87)
- Multi-modal Vision Pre-training for Medical Image Analysis (CVPR 2025 · similarity 0.86)
- Advancing Multiple Instance Learning with Continual Learning for Whole Slide Imaging (CVPR 2025 · similarity 0.85)
- FR²Seg: Continual Segmentation Across Multiple Sites via Fourier Style Replay and Adaptive Consistency Regularization (AAAI 2025 · similarity 0.85)
- Towards All-in-One Medical Image Re-Identification (CVPR 2025 · similarity 0.84)
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.
