Leveraging Large Language Models for Generating Personalized Care Recommendations in Dementia
Hsiang‐Wei Hu, Yu‐Chun Lin, Chang-Hung Chia, Ethan Chuang, Yang Cheng Ru · 2024
As dementia cases surge globally, with projections reaching 78 million by 2030, innovative care solutions are urgently needed. This study introduces a groundbreaking approach to personalized dementia care by integrating Large Language Models (LLMs) with Retrieval-Augmented Generation (RAG) technology. Our method leverages AI to enhance cognitive state assessment accuracy and generate tailored care recommendations. The system fine-tunes a GPT-4 based LLM, integrating it with a vector database using RAG to optimize care plan personalization. AI-generated diagnostic reports based on key clinical parameters achieved a 90% accuracy and 88% readability score when evaluated by clinical physicians. Personalized care plans encompassing cognitive training, music therapy, dietary control, and routine management were developed and rigorously assessed by healthcare professionals. These plans demonstrated exceptional performance, scoring between 7.66 and 8.93 out of 10 across suitability, scientific basis, operability, and comprehensiveness. This research addresses the critical need for scalable, personalized dementia care. By synergizing AI technology with clinical expertise, our approach has the potential to transform care paradigms, improve patient outcomes, and alleviate pressure on global healthcare systems. This integration of LLMs with RAG represents a significant advancement in medical AI, opening new avenues for personalized medicine in neurodegenerative disorders.