Research map / Architectures and efficiency / CNN design
Johnson-Lindenstrauss Lemma Guided Network for Efficient 3D Medical Segmentation
Jinpeng Lu, Linghan Cai, Yinda Chen, Guo Tang, Songhan Jiang, Haoyuan Shi, Zhiwei Xiong
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 Gram-matrix supervision (SDKT) from texture teacher (ablation) | VeloxSeg · AutoPET-II | Dice (%): 59.71 -> 62.51 | Improved |
| add reconstruction texture teacher without Gram supervision (ablation) | VeloxSeg · AutoPET-II | Dice (%): 61.43 -> 59.64 | Hurt |
| use l1, Affinity or shared-ROI transfer loss instead of SDKT (ablation) | VeloxSeg · AutoPET-II | Dice (%): 59.71 -> 1.67 (l1), 41.44 (Affinity), 57.15 (Shared ROI) | Hurt |
| add Transformer (PWA) attention encoder to JLC convolution (ablation) | VeloxSeg · AutoPET-II | Dice (%): 55.84 -> 61.03 (with depth 1) | Improved |
| increase attention depth from 1 to 2 per stage (ablation) | VeloxSeg · AutoPET-II | Dice (%): 61.03 -> 59.56 | Hurt |
| unify upsampling strategy (ablation) | VeloxSeg · AutoPET-II | GFLOPs and GPU throughput: 2.84 G -> 1.79 G; 336.94 -> 599.06 patches/s | Improved |
| add modal interaction into PWA (ablation) | VeloxSeg · AutoPET-II multimodal input | Dice (%): +5.75% | Improved |
| use JL-guided group sizes instead of depth-wise convolution (ablation) | pure convolution network · Hecktor2022 and BraTS2021 | Dice (%): +6.25% Hecktor2022; +1.16% BraTS2021 | Improved |
| raise convolution group size from 1 to 4,8,8,16 (JL-guided) (ablation) | convolution encoder · AutoPET-II | Dice (%): 53.65 -> 55.84 | Improved |
| larger group size 16,32,64,128 versus 4,8,8,16 (ablation) | convolution encoder · AutoPET-II | Dice (%): 55.84 -> 56.20 with Params 1.18M -> 4.76M | No clear effect |
| replace l2 pruning of full convolution with JLC (ablation) | convolution network · BraTS2021 to BraTS2016 TCIA | Dice (%): 62.00 -> 79.34 | Improved |
| widen convolution channels from 16-128 to 32-256 (ablation) | convolution encoder · AutoPET-II | Dice (%): 50.10 -> 48.96 | 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.
- SuperLightNet — SuperLightNet: Lightweight Parameter Aggregation Network for Multimodal Brain Tumor Segmentation (CVPR 2025) · better on 2 of 3 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 0.87)
- EffiDec3D: An Optimized Decoder for High-Performance and Efficient 3D Medical Image Segmentation (CVPR 2025 · similarity 0.87)
- E2ENet: Dynamic Sparse Feature Fusion for Accurate and Efficient 3D Medical Image Segmentation (NeurIPS 2024 · similarity 0.87)
- SGTC: Semantic-Guided Triplet Co-training for Sparsely Annotated Semi-Supervised Medical Image Segmentation (AAAI 2025 · similarity 0.86)
- Combinatorial CNN-Transformer Learning with Manifold Constraints for Semi-supervised Medical Image Segmentation (AAAI 2024 · similarity 0.86)
- Revisiting MAE Pre-training for 3D Medical Image Segmentation (CVPR 2025 · 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.
