Abstract
Recent advancements in artificial intelligence (AI), especially largelanguage models (LLMs), have significantly advanced healthcare applications anddemonstrated potentials in intelligent medical treatment. However, there areconspicuous challenges such as vast data volumes and inconsistent symptomcharacterization standards, preventing full integration of healthcare AIsystems with individual patients' needs. To promote professional andpersonalized healthcare, we propose an innovative framework, Heath-LLM, whichcombines large-scale feature extraction and medical knowledge trade-offscoring. Compared to traditional health management applications, our system hasthree main advantages: (1) It integrates health reports and medical knowledgeinto a large model to ask relevant questions to large language model fordisease prediction; (2) It leverages a retrieval augmented generation (RAG)mechanism to enhance feature extraction; (3) It incorporates a semi-automatedfeature updating framework that can merge and delete features to improveaccuracy of disease prediction. We experiment on a large number of healthreports to assess the effectiveness of Health-LLM system. The results indicatethat the proposed system surpasses the existing ones and has the potential tosignificantly advance disease prediction and personalized health management.