DEMENTIA-PLAN: An Agent-Based Framework for Multi-Knowledge Graph Retrieval-Augmented Generation in Dementia Care

This project develops DEMENTIA-PLAN, an agent-based framework that uses large language models and multiple knowledge graphs to support patients with mild-stage dementia. By integrating daily routine graphs and life memory graphs, the system provides both practical care assistance and emotional support through memory-based conversations. A self-reflection planning agent coordinates knowledge retrieval, dynamically balancing routine and memory information to generate responses that stabilize mood and strengthen memory recall. DEMENTIA-PLAN demonstrates how LLM-powered, knowledge-graph–driven agents can bridge the gap between clinical interventions and everyday caregiver support, advancing the future of dementia care.

Publications:


Full paper

Project information

  • Category: Future Healthcare, Large Language Models, and Personal Health Models
  • Contact Person: Amir M. Rahmani
  • DEMENTIA-PLAN: An Agent-Based Framework for Multi-Knowledge Graph Retrieval-Augmented Generation in Dementia Care

info@futurehealth.uci.edu

© Copyright 2021 UCI Institute for Future Health - All Rights Reserved