Research map / Detection and segmentation / CNN features
Harnessing Lightweight Transformer With Contextual Synergic Enhancement for Efficient 3D Medical Image Segmentation
Xinyu Liu, Zhen Chen, Wuyang Li, Chenxin Li, Yixuan Yuan
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 random replacement of labeled patches into unlabeled images (ablation) | Light-UNETR · LA (4/76) | Dice: 75.55 -> 86.78 | Improved |
| use Attention-Guided Replacement instead of random replacement (ablation) | CSE · LA (4/76) | Dice: 86.78 -> 87.79 | Improved |
| add Spatial Masking Consistency alone (ablation) | Light-UNETR · LA (4/76) | Dice: 75.55 -> 83.59 | Improved |
| add Spatial Masking Consistency to AGR (ablation) | CSE · LA (4/76) | Dice: 87.79 -> 89.53 | Improved |
| add EMA teacher on top of shared weights (ablation) | CSE-Light-UNETR · LA (4/76) | Dice: 89.53 -> 86.85 | Hurt |
| replace shared weights with EMA-only teacher (ablation) | CSE-Light-UNETR · LA (4/76) | Dice: 89.53 -> 84.53 | Hurt |
| use two high-frequency branches with kernels 3,5 instead of one branch kernel 3 (ablation) | LIDR · LA (8/72) | Dice: 89.39 -> 90.50 | Improved |
| use LIDR instead of vanilla attention (ablation) | Light-UNETR · LA (4/76) | Dice and params: 2.68M -> 1.34M params; 89.05 -> 89.53 Dice | Improved |
| raise lambda_ext from 1.0 to 4.0 with lambda_int 1.0 (ablation) | CSE · LA (4/76) | Dice: 87.68 -> 89.53 | Improved |
| add overlap patch embedding to ViT baseline (ablation) | Light-UNETR roadmap · LA (5% labeled) | Dice: -> 88.38 | Improved |
| tune replacement patch ratio alpha from 0.45 to 0.85 (ablation) | AGR · LA (5% labeled) | Dice | Mixed |
| apply CSE to V-Net instead of labeled-only V-Net | V-Net · LA (8/72) | Dice: 82.74 -> 90.28 | Improved |
Closest papers by meaning
- nnWNet: Rethinking the Use of Transformers in Biomedical Image Segmentation and Calling for a Unified Evaluation Benchmark (CVPR 2025 · similarity 0.86)
- EffiDec3D: An Optimized Decoder for High-Performance and Efficient 3D Medical Image Segmentation (CVPR 2025 · similarity 0.86)
- UniPro: Unified Multi-Mode Medical Image Segmentation from 2D Images to 3D Volumes via Propagation (NeurIPS 2026 · similarity 0.86)
- DTMFormer: Dynamic Token Merging for Boosting Transformer-Based Medical Image Segmentation (AAAI 2024 · similarity 0.85)
- Rethinking Decoders for Transformer-based Semantic Segmentation: A Compression Perspective (NeurIPS 2024 · similarity 0.85)
- Upping the Game: How 2D U-Net Skip Connections Flip 3D Segmentation (NeurIPS 2024 · similarity 0.85)
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.
