Research map / Detection and segmentation / CNN features
GeoSemba: Reconstructing State Space Model for Cross Paradigm Representation in Medical Image Segmentation
Xutao Sun, Jiarui Li, Junwen Liu, Yonggong Ren
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 Semantic Prototype Grouping to SSR (ablation) | GeoSemba SSR · ISIC2018 | DSC: 88.6 -> 90.5 | Improved |
| add Geometric-Semantic Nexus Attention to SSR (ablation) | GeoSemba SSR · ISIC2018 | DSC: 88.3 -> 90.5 | Improved |
| add macro-perception stage to CAR (ablation) | GeoSemba CAR · six datasets | average DSC: -2.13% average DSC when removed | Improved |
| add channel attention to CAR (ablation) | GeoSemba CAR · six datasets | average DSC: +1.88% | Improved |
| add Top-K sparsification in token aggregation (ablation) | GeoSemba CAR · ISIC2018 | DSC: 89.8 -> 90.7 | Improved |
| add learnable projection Wd in token aggregation (ablation) | GeoSemba CAR · ISIC2018 | DSC: 90.3 -> 90.7 | Improved |
| use 3 prototype nodes instead of 7 (ablation) | GeoSemba SSR · ISIC2018 | DSC: 91.1 -> 89.2 | Hurt |
| use 12 prototype nodes instead of 7 (ablation) | GeoSemba SSR · ISIC2018 | DSC: 91.1 -> 90.1; Infer 17.0 -> 19.8 ms | Hurt |
| use 6 or 18 channel groups instead of 12 (ablation) | GeoSemba CAR · ISIC2018 | DSC: 91.1 -> 89.5 (G=6), 90.2 (G=18) | Hurt |
| use Top-K ratio 0.25 or 0.75 instead of 0.50 (ablation) | GeoSemba CAR · ISIC2018 | DSC: 91.1 -> 90.0 (K=0.25), 90.3 (K=0.75) | Hurt |
| use GeoSemba instead of PVT-EMCAD on polyp data | CVC-ClinicDB · ClinicDB | DSC: 95.2 vs 94.8 | No clear effect |
| use single-scan GeoSemba instead of multi-directional scanning VM-UNet | Mamba segmentation · ISIC2018 | params / inference time: 54.86M -> 44.11M params; 0.023 -> 0.018 s/image | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- SCSegamba — SCSegamba: Lightweight Structure-Aware Vision Mamba for Crack Segmentation in Structures (CVPR 2025) · better on 1 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
- UniPro: Unified Multi-Mode Medical Image Segmentation from 2D Images to 3D Volumes via Propagation (NeurIPS 2026 · similarity 0.86)
- Mamba Goes HoME: Hierarchical Soft Mixture-of-Experts for 3D Medical Image Segmentation (NeurIPS 2025 · similarity 0.86)
- Multi-view Evidential Learning-based Medical Image Segmentation (AAAI 2025 · similarity 0.85)
- MAPSeg: Unified Unsupervised Domain Adaptation for Heterogeneous Medical Image Segmentation Based on 3D Masked Autoencoding and Pseudo-Labeling (CVPR 2024 · similarity 0.85)
- S³-Mamba: Small-Size-Sensitive Mamba for Lesion Segmentation (AAAI 2025 · similarity 0.85)
- S2S2: Semantic Stacking for Robust Semantic Segmentation in Medical Imaging (AAAI 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.
