Evidence-Based Intelligent Diagnostic and Therapeutic Visualization System with Large Language Models: Multi-Turn Interaction and Multimodal Treatment Plan Generation

Researchers developed an AI-enhanced diagnostic system for traditional Chinese medicine that combines Neo4j knowledge graphs, large language models, and multimodal visualization to improve diagnostic transparency and treatment planning. The system demonstrated a 32% reduction in non-standard outputs and significantly improved diagnostic trust and credibility compared to existing tools.

This research represents a meaningful advancement in applying artificial intelligence to medical diagnostics, specifically addressing longstanding criticisms of opaque AI reasoning in healthcare contexts. The system architecture addresses three core problems: the black-box nature of AI medical tools, passive user interaction limiting diagnostic accuracy, and inadequate treatment presentation for patient understanding and clinician confidence.

The technical approach combines structured knowledge representation through a Neo4j knowledge graph containing 241 syndromes and 1,263 symptoms with adaptive LLM verification, creating a hybrid system that leverages both symbolic AI rigor and neural network flexibility. The four-stage symptom matching pipeline and genetic algorithm-optimized questioning strategy reflect sophisticated system design aimed at reducing diagnostic error through both human-centered interaction and computational validation.

The clinical validation results demonstrate measurable improvements: a Cohen's d of 1.82 for diagnostic trust indicates substantial effect size, while reduced cognitive load across multiple dimensions suggests practical usability benefits. These metrics matter because healthcare AI adoption requires demonstrable improvements in both accuracy and practitioner confidence—areas where many current systems underperform.

The implications extend beyond traditional Chinese medicine to broader healthcare AI implementation. As regulatory bodies worldwide demand greater interpretability and trustworthiness in medical AI systems, this evidence-based approach combining knowledge graphs with multimodal presentation provides a replicable framework. The system's success in educational contexts suggests potential applications in medical training and knowledge transfer, while its performance improvements could influence adoption rates among clinical practitioners skeptical of AI-assisted diagnosis.

• →Knowledge graph constraints reduced non-standard AI outputs by 32%, improving diagnostic reliability
• →System achieved significant diagnostic trust improvement (Cohen's d = 1.82, p < 0.001) versus baseline approaches
• →Multimodal treatment presentation integrating AI illustrations, 3D models, and evidence-based literature enhances interpretability
• →Proactive questioning strategy optimized with genetic algorithms improves diagnostic accuracy through adaptive interaction
• →Framework demonstrates practical applicability across patient self-assessment, clinical diagnosis, and medical education