Research map / Detection and segmentation / Weakly supervised detection and segmentation
BackSplit: The Importance of Sub-dividing the Background in Biomedical Lesion Segmentation
Rachit Saluja, Asli Cihangir, Ruining Deng, Johannes C. Paetzold, Fengbei Liu, Mert R. Sabuncu
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 |
|---|---|---|---|
| sub-divide background into auxiliary anatomical classes (BackSplit) | U-Net · KiTS23 cyst | Dice: 0.1787 -> 0.4573 | Improved |
| BackSplit | U-Net, ResEncU-Net, SegResNet · PANTHER-MR tumor | Dice: U-Net 0.4784 -> 0.5251; ResEncU-Net 0.516 -> 0.5165 | Mixed |
| BackSplit | U-Net, ResEncU-Net, SegResNet · NSCLC-Radiomics GTV | Dice: 0.4969 -> 0.5256 (U-Net) | Improved |
| BackSplit with model-derived organ labels | U-Net · AutoPET tumor | Dice: 0.3881 -> 0.4435 | Improved |
| BackSplit with model-derived organ labels | U-Net · MSWAL all lesions | Dice: 0.2724 -> 0.3190 | Improved |
| BackSplit with model-derived organ labels | U-Net · MSWAL Kidney Stone | Dice: 0.195 -> 0.1639 | Hurt |
| use TotalSegmentator organ masks as auxiliary labels (ablation) | BackSplit U-Net · AutoPET tumor | Dice: 0.3921 -> 0.4456 | Improved |
| use VIBE-Segmentator organ masks as auxiliary labels (ablation) | BackSplit U-Net · MSWAL all lesions | Dice: 0.2518 -> 0.2646 | Improved |
| use noisy nnInteractive auxiliary masks (7 clicks) instead of GT (ablation) | BackSplit U-Net · KiTS23 cyst | Dice: 0.2033 -> 0.4919 | Improved |
| use noisy nnInteractive auxiliary masks (7 clicks) instead of GT (ablation) | BackSplit U-Net · MSWAL all lesions | Dice: 0.2518 -> 0.2714 | Improved |
| fine-tune pretrained binary model with auxiliary structures (ablation) | U-Net · KiTS23 cyst | Dice: nearly doubling by 250 epochs | Improved |
| partial auxiliary labels (small fraction) when training from scratch (ablation) | U-Net · KiTS23 cyst | Dice | Hurt |
Closest papers by meaning
- Minding Fuzzy Regions: A Data-driven Alternating Learning Paradigm for Stable Lesion Segmentation (CVPR 2025 · similarity 0.86)
- Unified Medical Lesion Segmentation via Self-referring Indicator (CVPR 2025 · similarity 0.86)
- EasyLens: A Training-Free Plug-and-Play Subtle-Lesion Representation Amplifier for Medical Vision-Language Models (NeurIPS 2026 · similarity 0.85)
- DyCON: Dynamic Uncertainty-aware Consistency and Contrastive Learning for Semi-supervised Medical Image Segmentation (CVPR 2025 · similarity 0.84)
- LeFusion: Controllable Pathology Synthesis via Lesion-Focused Diffusion Models (ICLR 2025 · similarity 0.84)
- Why Invariance is Not Enough for Biomedical Domain Generalization and How to Fix It (NeurIPS 2026 · 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.
