Research map / Graph learning / Graph self-supervised learning
Breaking Grid Constraints: Dynamic Graph Reconstruction Network for Multi-organ Segmentation
Junhao Xiao, Yang Wei, Jingyu Wang, Yongchao Wang, Xiuli Bi, Bin Xiao
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 |
|---|---|---|---|
| replace grid-aggregation CNN/Transformer with dynamic graph reconstruction (DGRNet) | DGRNet · BTCV | DSC: 0.7777 (UTNet) -> 0.8657 | Improved |
| DGRNet versus best baselines on ACDC and CMR | DGRNet · ACDC / CMR | DSC: ACDC 0.9285 -> 0.9454; CMR 0.9032 -> 0.9171 | Improved |
| use EdgeConv instead of GIN/GraphSAGE/MR GConv (ablation) | DGRNet · BTCV | DSC: 0.8598 (MR GConv) -> 0.8657 | Improved |
| use DGR decoder instead of interpolation or deconvolution upsampling (ablation) | DGRNet · BTCV | DSC: 0.8353 (deconv) -> 0.8657 | Improved |
| use dynamic neighbor count 9 to 18 instead of fixed K (ablation) | DGRNet · BTCV | DSC: 0.8506 (K=21) -> 0.8657 | Improved |
| add category-perception loss Lcp with dynamic weighting (ablation) | DGRNet · BTCV | DSC: 0.8426 -> 0.8657 | Improved |
| dynamic instead of static loss ratio between Dice-CE and Lcp (ablation) | DGRNet · BTCV | DSC: 0.8582 -> 0.8657 | Improved |
| graph construction in DGRNet | DGRNet · Model cost | FLOPs / Params: 10.8G FLOPs, 20.2M params versus TinyUNet 1.3G, 0.5M | Hurt |
Closest papers by meaning
- MagicNet: Semi-Supervised Multi-Organ Segmentation via Magic-Cube Partition and Recovery (CVPR 2023 · similarity 0.87)
- DoDNet: Learning To Segment Multi-Organ and Tumors From Multiple Partially Labeled Datasets (CVPR 2021 · similarity 0.86)
- Learning From Partially Labeled Data for Multi-Organ and Tumor Segmentation (TPAMI 2023 · similarity 0.86)
- Continual Segment: Towards a Single, Unified and Non-forgetting Continual Segmentation Model of 143 Whole-body Organs in CT Scans (ICCV 2023 · similarity 0.86)
- Teaching AI the Anatomy Behind the Scan: Addressing Anatomical Flaws in Medical Image Segmentation with Learnable Prior (ICCV 2025 · similarity 0.86)
- Efficient Connectivity-Preserving Instance Segmentation with Supervoxel-Based Loss Function (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.
