微调 MedGemma 用于临床图像描述以增强马来西亚临床实践指南上的多模态 RAG
Fine-Tuning MedGemma for Clinical Captioning to Enhance Multimodal RAG over Malaysia CPGs
AI总结：
本文通过知识蒸馏构建合成数据集并用 QLoRA 微调 MedGemma，使其生成高保真临床图像描述，经分类准确率和 RAGAS 双重评估验证，显著提升了多模态 RAG 在马来西亚临床实践指南中的查询质量。
摘要：
检索增强生成系统对于基于马来西亚临床实践指南提供事实性指导至关重要。然而，它们在基于图像的查询中的有效性有限，因为通用视觉-语言模型生成的描述往往缺乏临床特异性和事实依据。本研究提出并验证了一个框架，用于专门化 MedGemma 模型，以生成高保真度的描述，从而作为更优质的查询。为克服数据稀缺问题，我们采用知识蒸馏流程，在皮肤科、眼底和胸部放射影像三个领域构建合成数据集，并使用参数高效的 QLoRA 方法对 MedGemma 进行微调。性能通过双重框架进行严格评估：一方面测量分类准确率，另一方面通过 RAGAS 框架的创新应用，评估描述的事实忠实度、相关性和正确性。微调后的模型在分类性能上表现出显著提升，同时 RAGAS 评估确认了描述事实忠实度和正确性的显著增益，验证了该模型生成可靠、有事实依据描述的能力。本研究建立了一条用于专门化医学视觉-语言模型的稳健流程，并验证了所得模型作为高质量查询生成器的有效性，为在循证临床决策支持中增强多模态 RAG 系统奠定了基础。
英文摘要：
Retrieval-Augmented Generation systems are essential for providing fact-based guidance from Malaysian Clinical Practice Guidelines. However, their effectiveness with image-based queries is limited, as general Vision-Language Model captions often lack clinical specificity and factual grounding. This study proposes and validates a framework to specialize the MedGemma model for generating high-fidelity captions that serve as superior queries. To overcome data scarcity, we employ a knowledge distillation pipeline to create a synthetic dataset across dermatology, fundus, and chest radiography domains, and fine-tune MedGemma using the parameter-efficient QLoRA method. Performance was rigorously assessed through a dual framework measuring both classification accuracy and, via a novel application of the RAGAS framework, caption faithfulness, relevancy, and correctness. The fine-tuned model demonstrated substantial improvements in classification performance, while RAGAS evaluation confirmed significant gains in caption faithfulness and correctness, validating the models ability to produce reliable, factually grounded descriptions. This work establishes a robust pipeline for specializing medical VLMs and validates the resulting model as a high-quality query generator, laying the groundwork for enhancing multimodal RAG systems in evidence-based clinical decision support.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
