Research map / Vision-language and multimodal / Multimodal LLMs
MARE: Multimodal Analogical Reasoning for Disease Evolution-Aware Radiology Report Generation
Qingqing Gao, Tengfei Liu, Xiaoyan Li, Xiaodan Zhang, Zhongfan Sun, Boyue Wang, Baocai Yin, Zhaohui Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes MARE, an end-to-end framework that formulates longitudinal radiology report generation as multimodal analogical reasoning to model disease evolution.
Problem
Radiology report generation from longitudinal medical data needs explicit disease evolution modeling and temporal reasoning, but existing methods mainly rely on multimodal feature fusion.
Approach
Formulates longitudinal report generation as multimodal analogical reasoning so the model explicitly reasons about how disease evolves across time, rather than only fusing features from multiple timepoints.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Adaptive Region Alignment (ARA) (ablation) | MARE · Longitudinal-MIMIC test | CIDEr: 0.243 -> 0.271 | Improved |
| add Adaptive Region Alignment (ARA) (ablation) | MARE · Longitudinal-MIMIC test | F-1: 0.353 -> 0.375 | Improved |
| add Adaptive Region Alignment (ARA) (ablation) | MARE · Longitudinal-MIMIC test | BLEU-4: 0.129 -> 0.133 | Improved |
| add both DEC losses (ablation) | MARE · Longitudinal-MIMIC test | F-1: 0.330 -> 0.375 | Improved |
| add image-evolution consistency loss alone (ablation) | MARE · Longitudinal-MIMIC test | F-1: 0.330 -> 0.362 | Improved |
| add report-evolution consistency loss alone (ablation) | MARE · Longitudinal-MIMIC test | F-1: 0.330 -> 0.345 | Improved |
| add report-evolution loss on top of image-evolution loss (ablation) | MARE · Longitudinal-MIMIC test | CIDEr: 0.265 -> 0.271 | Improved |
| add both DEC losses (ablation) | MARE · Longitudinal-MIMIC test | CIDEr: 0.251 -> 0.271 | Improved |
| increase ARA learnable queries from 16 to 32 (ablation) | MARE · Longitudinal-MIMIC test | CIDEr | Improved |
| increase ARA learnable queries beyond 32 (ablation) | MARE · Longitudinal-MIMIC test | CIDEr | Hurt |
| add historical report Yp as input | R2GenGPT · Longitudinal-MIMIC test | F-1: 0.247 -> 0.336 | Improved |
| add historical image Xp as input | R2GenGPT · Longitudinal-MIMIC test | F-1: 0.247 -> 0.250 | No clear effect |
Closest papers by meaning
- BiOTPrompt: Bidirectional Optimal Transport Guided Prompting for Disease Evolution-aware Radiology Report Generation (CVPR 2026 · similarity 0.90)
- S2D-Align: Shallow-to-Deep Auxiliary Learning for Anatomically-Grounded Radiology Report Generation (AAAI 2026 · similarity 0.88)
- HC-LLM: Historical-Constrained Large Language Models for Radiology Report Generation (AAAI 2025 · similarity 0.88)
- DART: Disease-aware Image-Text Alignment and Self-correcting Re-alignment for Trustworthy Radiology Report Generation (CVPR 2025 · similarity 0.87)
- LLM-RG4: Flexible and Factual Radiology Report Generation Across Diverse Input Contexts (AAAI 2025 · similarity 0.87)
- Enhanced Contrastive Learning with Multi-view Longitudinal Data for Chest X-ray Report Generation (CVPR 2025 · similarity 0.86)
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.
