Research map / Label-efficient and robust learning / Domain adaptation and generalization
CoSMIC: Continual Self-supervised Learning for Multi-Domain Medical Imaging via Conditional Mutual Information Maximization
Yihang Liu, Ying Wen, Longzhen Yang, Lianghua He, Heng Tao Shen
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 reconstruction loss L_Recon (ablation) | CoSMIC · Retinoscopy D1 | A (avg): 85.79 -> 87.70 | Improved |
| add the ([XZ],Y) similarity term in L_CoCst (ablation) | CoSMIC · Retinoscopy D1 | A (avg): 78.64 -> 87.70 | Improved |
| add the (X,Y) similarity term in L_CoCst (ablation) | CoSMIC · Retinoscopy D1 | A (avg): 83.63 -> 87.70 | Improved |
| add Anatomy-Guided Calibration (ablation) | CoSMIC · Retinoscopy D1 | A (avg): 85.04 -> 87.70 | Improved |
| add Anatomy-Guided Calibration (ablation) | CoSMIC · Retinoscopy D1 | KB: 0.91 -> 0.21 | Improved |
| add Hierarchical Alignment (ablation) | CoSMIC · Retinoscopy D1 | A (avg): 86.59 -> 87.70 | Improved |
| add the ([XZ],Y) similarity term in L_CoCst (ablation) | CoSMIC · Retinoscopy D1 | KB: 8.07 -> 0.21 | Improved |
| add the ([XZ],Y) similarity term in L_CoCst (ablation) | CoSMIC · Histopathology D5 (unseen) | KF: -9.43 -> -0.71 | Improved |
| reverse pre-training order to D4-D3-D2-D1 (ablation) | CoSMIC · Retinoscopy D1 | A (avg): 87.70 -> 87.32 | No clear effect |
| use CoSMIC instead of MedCoss (rehearsal) when moving from D0 to D1 | continual SSL · D1 knowledge backward | KB: 7.33 -> 0.21 | Improved |
| use CoSMIC instead of RETFound on APTOS | fundus pre-training · APTOS | AUC (%): 94.48 -> 95.39 | Improved |
| use CoSMIC instead of RETFound on PAPILA | fundus pre-training · PAPILA | AUC (%): 84.51 -> 80.10 | Hurt |
Closest papers by meaning
- Continual Self-supervised Learning: Towards Universal Multi-modal Medical Data Representation Learning (CVPR 2024 · similarity 0.86)
- Perturbating, Tuning, and Collaborating: Harnessing Vision Foundation Models for Single Domain Generalization on Medical Imaging (AAAI 2025 · similarity 0.86)
- Advancing Multiple Instance Learning with Continual Learning for Whole Slide Imaging (CVPR 2025 · similarity 0.85)
- Knowledge Transfer Scaling Laws for 3D Medical Imaging (NeurIPS 2026 · similarity 0.85)
- TinyMIG: Transferring Generalization from Vision Foundation Models to Single-Domain Medical Imaging (ICML 2025 · similarity 0.84)
- A Textbook Remedy for Domain Shifts: Knowledge Priors for Medical Image Analysis (NeurIPS 2024 · 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.
