The ambiguity of symptom information in traditional Chinese medicine （TCM） syndrome differentiation can be amplified in the direct reasoning process from symptoms to syndromes in the absence of constraints， which affects the accuracy and stability of intelligent syndrome differentiation. TCM disease concepts， while historically rational， are structurally ambiguous in both their connotation and extension， making it difficult to serve as stable prior knowledge in intelligent modeling. In contrast， modern medical diseases， based on objective testing and quantifiable indicators， have relatively clear boundaries and reproducible standards. This study proposes a disease-syndrome combination model， adopting modern medical diseases as structural prior variables to reconstruct the hierarchical relationships among disease， symptoms， and syndromes. By applying disease constraints， effective screening of information from the four examinations and compressing the reasoning space are achieved. Furthermore， by integrating artificial intelligence technologies， such as multimodal fusion and knowledge graphs， an intelligent syndrome differentiation model driven by both prior knowledge and clinical data is constructed， providing a feasible path to enhance the accuracy of syndrome differentiation and realize the intelligentization of TCM diagnosis and treatment.