Objective To construct an intelligent agent for the diagnosis and treatment of Alzheimer’s disease (AD) based on knowledge of renowned veteran traditional Chinese medicine (TCM) physicians’ knowledge by adopting artificial intelligence techniques including chain-of-thought reasoning, retrieval-augmented generation (RAG), and prompt engineering, so as to support clinical decision-making with expert TCM experience. Methods Relevant literature and monographs regarding AD treatment by prestigious TCM physicians were retrieved, and a total of 16 renowned veteran TCM physicians, including Tian Jinzhou, Zhou Zhongying, and Shen Baofan, were selected. Their medical records, clinical experience, and academic theories were systematically mined to establish a knowledge base of renowned veteran TCM physicians. To address inconsistencies in syndrome differentiation approaches among different physicians, a syndrome-element classification method was adopted to standardize syndrome differentiation.Based on a large language model and the constructed knowledge base, an intelligent agent for the diagnosis and treatment of AD was developed on the Dify 1.9.2 platform, incorporating both an agent-based individual decision-making scheme and an agent-based group decision-making scheme. Symptom clusters extracted from 108 real-world medical records were used as inputs to the agent, which generated syndrome differentiation results and recommended prescriptions, thereby creating an evaluation dataset. A total of 30 cases were randomly selected for a single-blind expert questionnaire evaluation. Experts evaluated the rationality of syndrome differentiation and prescription recommendations generated by three approaches: the original medical records, the agent-based individual decision-making scheme, and the agent-based group decision-making scheme. Results The agent-based group decision-making scheme achieved the highest estimated marginal mean score (3.93 ± 0.14), outperforming both the agent-based individual decision-making scheme (3.47 ± 0.14) and the original medical records (3.16 ± 0.14). The agent-based individual decision-making scheme also outperformed the original medical records. Conclusion By integrating large language model technology with a syndrome-element classification method, this study incorporated the syndrome differentiation experience of multiple renowned veteran TCM physicians and developed an intelligent agent for the diagnosis and treatment of AD that embodies their collective expertise. The agent-based group decision-making scheme achieved higher scores in expert evaluations, suggesting that this approach can leverage the advantages of collaborative decision-making among multiple renowned physicians. The development of this intelligent agent provides a reusable methodological framework and implementation pathway for applying renowned physicians’ knowledge to support clinical decision-making, and offers a model for developing disease-specific intelligent agents that leverage the advantages of group decision-making, thereby laying the foundation for a subsequent prospective validation platform.

Alzheimer’s disease  /  renowned veteran traditional Chinese medicine physicians  /  intelligent agent  /  retrieval-augmented generation  /  syndrome-element differentiation  /  group decision-making