Research map / Label-efficient and robust learning / Test-time adaptation
Memory-Supported Synergistic Adaptation for Training-Free Test-Time Medical Image Segmentation
Lingrui Li, Nan Pu, Dong Zhao, Wenjing Li, Andrew French, Xin Chen, Zhun Zhong
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 SAM majority voting (ablation) | MedCLIPv2 baseline · Optic disc ablation (avg) | mIoU: 58.0 -> 61.8 | Improved |
| add attention-map majority voting (ablation) | SAM-MV pipeline · Optic disc ablation (avg) | mIoU: 61.8 -> 62.9 | Improved |
| add Gaussian point selection (ablation) | SAM-MV + ATN-MV · Optic disc ablation (avg) | mIoU: 62.9 -> 64.7 | Improved |
| add noise-aware memory construction with similar-anchor selection to full image-text branch (ablation) | MSSA · Optic disc ablation (avg) | mIoU: 64.7 -> 73.1 | Improved |
| add similarity-based anchor selection (ablation) | MSSA without selection · Optic disc ablation (avg) | mIoU: 70.8 -> 73.1 | Improved |
| self-referential image-image branch without image-text stabilization (add select) (ablation) | Bank-only variant · Optic disc ablation (avg) | mIoU: 67.0 -> 59.7 | Hurt |
| select anchor by similarity instead of random (ablation) | MSSA · Optic disc anchor selection (avg) | mIoU: 70.8 -> 73.1 | Improved |
| select anchor by similarity instead of semantic alignment (ablation) | MSSA · Optic disc anchor selection (avg) | DSC: 82.8 -> 83.9 | Improved |
| combine semantic and smoothness scores instead of Q_sem only (ablation) | MSSA · Lung memory scoring (avg) | DSC: 69.7 -> 78.4 | Improved |
| combine semantic and smoothness scores instead of Q_smo only (ablation) | MSSA · Lung memory scoring (avg) | DSC: 69.9 -> 78.4 | Improved |
| apply training-based TTA (TENT/EATA/CoTTA/DELTA) to VLM segmentation | MedCLIPv2 · Optic disc, 5 domains avg | DSC: 71.4 -> 71.0 | No clear effect |
| apply UniVPT TTA | MedCLIPv2 · Optic disc, 5 domains avg | DSC: 71.4 -> 67.1 | Hurt |
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.
- TENT — Tent: Fully Test-Time Adaptation by Entropy Minimization (ICLR 2021) · better on 2 of 2 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
- Gradient Alignment Improves Test-Time Adaptation for Medical Image Segmentation (AAAI 2025 · similarity 0.92)
- Each Test Image Deserves A Specific Prompt: Continual Test-Time Adaptation for 2D Medical Image Segmentation (CVPR 2024 · similarity 0.91)
- Test-Time Adaptation of Vision-Language Models for Open-Vocabulary Semantic Segmentation (NeurIPS 2025 · similarity 0.90)
- Efficient Deformable Convolutional Prompt for Continual Test-Time Adaptation in Medical Image Segmentation (AAAI 2025 · similarity 0.88)
- Efficient Test-time Adaptation through Candidate Verification and Divergence Shifts (NeurIPS 2026 · similarity 0.87)
- S2S2: Semantic Stacking for Robust Semantic Segmentation in Medical Imaging (AAAI 2025 · similarity 0.87)
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.
