Research map / Vision-language and multimodal / Cross-modal learning
MDCS-MoAME: Multi-directional Composite Scanning with Mixture of Attention and Mamba Experts for Cancer Survival Prediction
Linjie Qu, Jin Xiao, Xiangrong Liu, Changming Sun, Hui Cui, Yuqi Fang, Ran Su, Qiangguo Jin, Leyi Wei
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 multi-directional composite scanning module with original Mamba scanning (ablation) | MDCS-MoAME · LUAD / UCEC | c-index: LUAD 0.7079 -> 0.6695; UCEC 0.7263 -> 0.6752 | Improved |
| add expert-driven inter-modal interaction (MoAME) module (ablation) | MDCS-MoAME · LUAD / UCEC | c-index: LUAD 0.6894 -> 0.7079; UCEC 0.6929 -> 0.7263 | Improved |
| add inter-modal alignment loss Lcro (ablation) | MDCS-MoAME · LUAD / UCEC | c-index: LUAD 0.6864 -> 0.7079; UCEC 0.7143 -> 0.7263 | Improved |
| add region-patch redundancy loss Lintra (ablation) | MDCS-MoAME · LUAD / UCEC | c-index: LUAD 0.6955 -> 0.7079; UCEC 0.6981 -> 0.7263 | Improved |
| add patch-level WSI features alongside region-level (ablation) | MDCS-MoAME · LUAD / UCEC | c-index: LUAD 0.6745 -> 0.7079; UCEC 0.6901 -> 0.7263 | Improved |
| add region-level WSI features alongside patch-level (ablation) | MDCS-MoAME · LUAD / UCEC | c-index: LUAD 0.6739 -> 0.7079; UCEC 0.7017 -> 0.7263 | Improved |
| combine straight and oblique WSI scanning instead of a single direction (ablation) | MDCS-MoAME · LUAD | c-index: 0.6826 / 0.6863 -> 0.6905 | Improved |
| use gene interval scanning instead of forward scanning (ablation) | MDCS-MoAME · UCEC | c-index: 0.7032 -> 0.7065 | Improved |
| use all scanning directions for WSIs and genes together (ablation) | MDCS-MoAME · LUAD / UCEC | c-index: LUAD 0.7079; UCEC 0.7263 | Improved |
| increase number of experts in MoAME (single to three) (ablation) | MDCS-MoAME · LUAD / UCEC | c-index | Improved |
| pair CroMamFusion with StackedMamFusion experts (M+S) (ablation) | MoAME · LUAD / UCEC | c-index | Hurt |
| sparse gating selecting one expert per input instead of dense MoE | MDCS-MoAME · TCGA average | FLOPs (G): 5.67 vs MoME 6.95 and PAMoE 11.04 | Improved |
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.
- MoME — MoME: Mixture of Multimodal Experts for Generalist Multimodal Large Language Models (NeurIPS 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
- Multimodal Prototyping for cancer survival prediction (ICML 2024 · similarity 0.88)
- Prototypical Information Bottlenecking and Disentangling for Multimodal Cancer Survival Prediction (ICLR 2024 · similarity 0.88)
- CA-MLIF: Cross-Attention and Multimodal Low-Rank Interaction Fusion Framework for Tumor Prognostic Prediction (AAAI 2025 · similarity 0.87)
- Modeling Dense Multimodal Interactions Between Biological Pathways and Histology for Survival Prediction (CVPR 2024 · similarity 0.86)
- Robust Multimodal Survival Prediction with Conditional Latent Differentiation Variational AutoEncoder (CVPR 2025 · similarity 0.86)
- Leveraging Tumor Heterogeneity: Heterogeneous Graph Representation Learning for Cancer Survival Prediction in Whole Slide Images (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.
