Research map / Image generation / Autoregressive and latent generation
TIM: Temporal Decoupling with Iterative Mutual-Refinement Model for Longitudinal Radiology Report Generation
Yiheng Dong, Yi Lin, Shilong Huang, Xiyan Yang, Xin Yang
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 dynamic progression modeling branch with perception frames and a video encoder (ablation) | Base single-timepoint LRRG model · Longitudinal-MIMIC test | F1: 0.404 -> 0.427 | Improved |
| add dual-timepoint report generation (ablation) | Model with DPM · Longitudinal-MIMIC test | F1: 0.427 -> 0.446 | Improved |
| add static pathology contrastive loss Lspr (ablation) | Model with DPM and DRG · Longitudinal-MIMIC test | F1: 0.446 -> 0.460 | Improved |
| add progression alignment loss Ldpm (ablation) | Model with DPM and DRG · Longitudinal-MIMIC test | F1: 0.446 -> 0.451 | Improved |
| add both Lspr and Ldpm (full Stage I) (ablation) | Model with DPM and DRG · Longitudinal-MIMIC test | F1: 0.446 -> 0.483 | Improved |
| add prior difference information (error triplets) as Stage II input (ablation) | TIM Stage II · Longitudinal-MIMIC test | F1: 0.453 -> 0.487 | Improved |
| add similarity-based mutual-refinement loss L∆ (ablation) | TIM Stage II · Longitudinal-MIMIC test | F1: 0.487 -> 0.511 | Improved |
| apply Stage II mutual refinement on top of Stage I | TIM · Longitudinal-MIMIC test | F1: 0.483 -> 0.511 | Improved |
| use 2 perception frames rather than none (ablation) | TIM Stage I · Longitudinal-MIMIC test | F1: 0.454 -> 0.483 | Improved |
| increase perception frames from 2 to 10 (ablation) | TIM Stage I · Longitudinal-MIMIC test | F1: 0.483 -> 0.443 | Hurt |
| increase refinement iterations from 3 to 5 (ablation) | TIM Stage II · Longitudinal-MIMIC test | F1: 0.511 -> 0.510 | No clear effect |
| use larger LLM backbone (Vicuna-13B vs LLaMA2-7B) (ablation) | TIM · Longitudinal-MIMIC test | F1 (Stage II): 0.511 -> 0.529 | Improved |
Closest papers by meaning
- HC-LLM: Historical-Constrained Large Language Models for Radiology Report Generation (AAAI 2025 · similarity 0.90)
- Bootstrapping Large Language Models for Radiology Report Generation (AAAI 2024 · similarity 0.88)
- 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 0.87)
- S2D-Align: Shallow-to-Deep Auxiliary Learning for Anatomically-Grounded Radiology Report Generation (AAAI 2026 · similarity 0.86)
- DART: Disease-aware Image-Text Alignment and Self-correcting Re-alignment for Trustworthy Radiology Report Generation (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.
