Research map / Detection and segmentation / Open-set and few-shot detection
From Minimal Clinical Prompts to 3D: Spacing-Aware Prompt Propagation for Multimodal Prostate Lesion Segmentation in bpMRI
Jiacheng Wang, Heinrich von Busch, Robert Grimm, Ipek Oguz, Dorin Comaniciu, Ali Kamen, Bin Lou
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 adaptive multimodal fusion with cross-attention instead of late fusion (ablation) | SAPP · Oracle-mask 3D, held-out cohorts | 3D DSC: 0.78 -> 0.85 | Improved |
| add cross-attention context fusion instead of early channel stacking (ablation) | SAPP · Oracle-mask 3D, held-out cohorts | 3D DSC: 0.75 -> 0.85 | Improved |
| add spacing-decayed memory attention (ablation) | SAPP · Oracle-mask 3D, held-out cohorts | 3D DSC: 0.74 -> 0.85 | Improved |
| add soft-to-hard stop gate (ablation) | SAPP · Oracle-mask 3D, held-out cohorts | 3D DSC: 0.76 -> 0.85 | Improved |
| add Module 2 (spacing decay plus stop gate) (ablation) | SAPP · Oracle-mask 3D, held-out cohorts | 3D DSC: 0.73 -> 0.85 | Improved |
| add gland-prior input (ablation) | SAPP · Oracle-mask 3D, held-out cohorts | 3D DSC: ~0.08 DSC reduction when removed | Improved |
| add prompt augmentation during training (ablation) | SAPP · Random lesion slice prompt | 3D DSC: 0.63 -> 0.71 | Improved |
| use box prompt instead of diameter prompt | SAPP · Prompted-slice, 11 cohorts avg | DSC: 0.82 -> 0.86 | Improved |
| prompt on random lesion slice instead of largest slice | SAPP · Prompt-to-volume 3D | 3D DSC: 0.87 -> 0.71 | Hurt |
| prompt on second-largest slice instead of largest slice | SAPP · Prompt-to-volume 3D | 3D DSC: 0.86 -> 0.82 | Hurt |
| add SAPP propagation over strongest baseline MedSAM2, diameter prompt | MedSAM2 · Prompt-to-volume 3D, diameter prompt | 3D DSC: 0.60 -> 0.78 | Improved |
| coarser slice spacing (1.5 mm to 6.0 mm) degrades SAM 2 more than SAPP | SAM 2 vs SAPP · Oracle-mask 3D by spacing | 3D DSC: SAM 2 ~0.78 -> ~0.51; SAPP ~0.86 -> ~0.72 | 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.
- I-MedSAM — I-MedSAM: Implicit Medical Image Segmentation with Segment Anything (ECCV 2024) · 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
- NaMa: Neighbor-Aware Multi-Modal Adaptive Learning for Prostate Tumor Segmentation on Anisotropic MR Images (AAAI 2024 · similarity 0.87)
- CAT: Coordinating Anatomical-Textual Prompts for Multi-Organ and Tumor Segmentation (NeurIPS 2024 · similarity 0.85)
- UniPro: Unified Multi-Mode Medical Image Segmentation from 2D Images to 3D Volumes via Propagation (NeurIPS 2026 · similarity 0.85)
- Unleashing the Potential of SAM for Medical Adaptation via Hierarchical Decoding (CVPR 2024 · similarity 0.84)
- Towards a Comprehensive, Efficient and Promptable Anatomic Structure Segmentation Model Using 3D Whole-Body CT Scans (AAAI 2025 · similarity 0.84)
- Knowledge Transfer Scaling Laws for 3D Medical Imaging (NeurIPS 2026 · similarity 0.83)
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.
