LLM-Powered Personalized Dietary Agent for Phenylketonuria
Pedro Henrique Palhano Modolo
Co-Presenters: Individual Presentation
College: The Dorothy and George Hennings College of Science, Mathematics and Technology
Major: Computer Science
Faculty Research Mentor: Malihe Aliasgari
Abstract:
Phenylketonuria (PKU) is a rare genetic disorder that requires meticulous lifelong dietary management to prevent severe neurological consequences. This study presents an innovative solution to the management of PKUs using AI-driven personalized diet recommendations. We propose a novel system that combines Large Language Models (LLMs), rule-based decision support systems, and patient profile to provide real-time individualized nutritional guidance. By integrating patient-specific data with established clinical guidelines, our AI-powered agent offers precise, dynamic recommendations that adapt to the unique needs of each patient, promoting better adherence to strict dietary requirements of PKU.The challenge of ensuring data accuracy and mitigating AI hallucinations - where incorrect recommendations are generated - remains a significant concern. To address these issues, we introduce Graph Retrieval-Augmented Generation (Graph RAG), a cutting-edge approach that enhances recommendation accuracy by aligning output with clinical guidelines and minimizing misinformation. Our results demonstrate improved patient adherence and significant health improvements through personalized dietary management.This work underscores the importance of AI in the transformation of PKU care, offering a scalable solution with the potential to benefit a larger patient population. We also highlight the need for robust AI-driven healthcare regulatory requirements to ensure consistent and high-quality care, setting the stage for future advances in personalized healthcare for patients with PKU.