Reproduce “Constructing Ophthalmic MLLM for Positioning-diagnosis Collaboration Through Clinical Cognitive Chain Reasoning” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xinyao Liu, Diping Song
Researchers from Shanghai Artificial Intelligence Laboratory and the University of Science and Technology of China developed FundusExpert, an ophthalmology-specific multimodal large language model, and FundusGen, a dataset curated to enable positioning-diagnosis collaboration. The model integrates image localization with diagnostic reasoning via a "clinical cognitive chain," achieving superior performance in ophthalmic question-answering and zero-shot report generation with 77.0% clinical consistency.
Key insights
- Aligning MLLM training with a 'clinical cognitive chain' (moving from region localization to feature analysis and diagnostic reasoning) is crucial for achieving high precision, interpretability, and clinical consistency in specialized medical domains.
- Cognitively-aligned annotations within datasets significantly enhance data utilization efficiency, demonstrating a power-law scaling (L ```json { "summary": "Researchers from Shanghai Artificial Intelligence Laboratory and the University of Science and Technology of China developed FundusExpert, an ophthalmology-specific multimodal large language model, and FundusGen, a dataset curated to enable positioning-diagnosis collaboration. The model integrates image localization with diagnostic reasonin
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.