Deep learning driven personalized paths for TCM graduate education

Traditional Chinese Medicine (TCM) graduate education struggles to balance classical knowledge and modern medical integration. This study developed and validated a deep learning-based personalized learning path system. Due to the sequential nature of graduate program enrollment, a true randomized controlled trial was not feasible. We therefore conducted a rigorously controlled quasi-experimental non-randomized controlled trial involving 120 TCM graduate students (2019–2025, Anhui Medical University). Participants were assigned to the AI-personalized group (n = 60) or traditional learning group (n = 60) via stratified alternating allocation—a standard quasi-randomization method recommended by the Cochrane Handbook—stratified by prior education background and baseline TCM knowledge score to ensure baseline balance. The study followed the CONSORT 2010 Extension for Non-Randomized Trials (TREND statement) for standardized reporting. The system achieved 87.3% (95% CI:84.2–90.1) accuracy in predicting optimal learning paths. Students in the AI-personalized group showed significant improvements across all three domains: the proportion of students achieving “good” proficiency increased from an initial 26%–35% (TCM knowledge: 35.2%, clinical skills: 26.7%, research capability: 28.3%) to 46%–68% (TCM knowledge: 68.3%, clinical skills: 58.3%, research capability: 46.7%) (Cohen’s d = 0.71–0.91; all p < 0.001). Deep learning personalization enhances TCM education outcomes, suggesting a potentially scalable framework for integrative medicine programs.