Research map / Detection and segmentation / Open-set and few-shot detection
Med-SegLens: Latent-Level Model Diffing for Interpretable Medical Image Segmentation
Salma J. Ahmed, Emad Mohammed, Azam Asilian Bidgoli
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 |
|---|---|---|---|
| use cross-space consistent Hungarian matching instead of greedy, Sinkhorn or encoder-only matching (ablation) | SAE latent alignment · Adult vs Pediatric | Predictability: 78.7 -> 90.0 | Improved |
| use cross-space consistent matching instead of Sinkhorn (ablation) | SAE latent alignment · Adult vs Pediatric | Agreement: 98.9 -> 100.0 | Improved |
| use cross-space consistent matching instead of greedy (ablation) | SAE latent alignment · Adult vs Pediatric | Collision: 0.014 -> 0.000 | Improved |
| steer the most activated latent instead of the base model | SegFormer · Adult failure cases | mean Dice: 25.3 -> 43.0 | Improved |
| steer latent instead of base model | SegFormer · Pediatric failure cases | mean Dice: 58.8 -> 62.0 | Improved |
| steer latent instead of base model | SegFormer · SSA failure cases | mean Dice: 59.2 -> 63.8 | Improved |
| add connected-component filtering | SegFormer · Pediatric failure cases | mean Dice: 58.8 -> 47.4 | Hurt |
| add entropy-based uncertainty filtering | SegFormer · Pediatric failure cases | mean Dice: 58.8 -> 47.5 | Hurt |
| add connected-component filtering | SegFormer · Adult failure cases | mean Dice: 25.3 -> 25.7 | No clear effect |
| edema latent steering on Adult failures with edema Dice below 50% | SegFormer · Adult edema failure cases | Edema Dice: 65.8 -> 69.1 | Improved |
| add CCF or UF on Adult edema failures | SegFormer · Adult edema failure cases | Edema Dice: 65.8 -> 65.6 | No clear effect |
| suppress edema latents with additive steering in Adult->Pediatric | SegFormer · Adult->Pediatric | Class 2 Dice: 39.4 -> 74.2 | Improved |
Closest papers by meaning
- S2S2: Semantic Stacking for Robust Semantic Segmentation in Medical Imaging (AAAI 2025 · similarity 0.86)
- Steady Progress Beats Stagnation: Mutual Aid of Foundation and Conventional Models in Mixed Domain Semi-Supervised Medical Image Segmentation (CVPR 2025 · similarity 0.86)
- Advancing Generalizable Tumor Segmentation with Anomaly-Aware Open-Vocabulary Attention Maps and Frozen Foundation Diffusion Models (CVPR 2025 · similarity 0.86)
- LangDAug: Langevin Data Augmentation for Multi-Source Domain Generalization in Medical Image Segmentation (ICML 2025 · similarity 0.86)
- Generative Medical Segmentation (AAAI 2025 · similarity 0.86)
- Why Invariance is Not Enough for Biomedical Domain Generalization and How to Fix It (NeurIPS 2026 · 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.
