Abstract:
Chronic diseases require long-term, individualized, and adaptive self-management; however, existing digital health
solutions often suffer from fragmented medical knowledge, limited personalization, and insufficient integration of
professional guidance with patients’ real-world experiences. In particular, the separation between Chinese medicine and
Western medicine further restricts the effectiveness of current self-management systems. To address these challenges, this
s
tudy proposes an AI agent–based self-management framework that integrates Chinese and Western medical knowledge,
using Type 2 diabetes as a representative chronic disease.
The proposed framework consists of three core components. First, a large language model (LLM)–driven module is
designed to systematically extract and organize heterogeneous knowledge from Chinese and Western medical experts,
clinical guidelines, and patients’ self-management experiences. Second, an integrated Chinese–Western medicine
knowledge graph is constructed to unify traditional syndrome patterns with modern biomedical concepts, enabling
s
tructured representation, semantic alignment, and cross-system reasoning. Third, a multi-agent AI system is developed
based on the knowledge graph, with specialized agents to generate personalized self-management plans, supported by
explicit safety constraints and a continuous feedback mechanism for dynamic adjustment.
A prototype system is implemented to demonstrate the feasibility and effectiveness of the proposed framework. The results
show that the system can provide individualized, explainable, and adaptive self-management guidance, while maintaining
medical safety and consistency across different medical paradigms. This work contributes a novel AI-enabled approach
for chronic disease self-management by bridging Chinese and Western medical knowledge through knowledge graphs
and multi-agent systems. It offers a practical and scalable reference for patient-centered healthcare systems that aim to
integrate heterogeneous medical knowledge and support sustainable long-term disease management.
