Correlation between TCM Syndromes and Type 2 Diabetic Comorbidities Based on Fully Connected Neural Network Prediction Model

Abstract

Objective. To predict the major comorbidities of type 2 diabetes based on the distribution characteristics of syndromes, and to explore the relationship between TCM syndromes and comorbidities of type 2 diabetes. Methods. Based on the electronic medical record data of 3413 outpatient visits from 995 type 2 diabetes patients with comorbidities, descriptive statistical methods were used to analyze the basic characteristics of the population, the distribution characteristics of comorbidities, and TCM syndromes. A neural network model for the prediction of type 2 diabetic comorbidities based on TCM syndromes was constructed. Results. Patients with TCM syndrome of blood amassment in the lower jiao were diagnosed with renal insufficiency with 95% test sensitivity. The patients with spleen deficiency combined with ascending counterflow of stomach qi and cold-damp patterns were diagnosed with gastrointestinal lesions with 92% sensitivity. The patients with TCM syndrome group of spleen heat and exuberance of heart fire were diagnosed as type 2 diabetes complicated with hypertension with a sensitivity of 91%. In addition, the prediction accuracy of combined neuropathy, heart disease, liver disease, and lipid metabolism disorder reached 70∼90% in TCM syndrome groups. Conclusion. The fully connected neural network model study showed that syndrome characteristics are highly correlated with type 2 diabetes comorbidities. Syndrome location is commonly in the heart, spleen, stomach, lower jiao, meridians, etc., while syndrome pattern manifests in states of deficiency, heat, phlegm, and blood stasis. The different combinations of disease location and disease pattern reflect the syndrome characteristics of different comorbidities forming the characteristic syndrome group of each comorbidity. Major comorbidities could be predicted with a high degree of accuracy through TCM syndromes. Findings from this study may have further implementations to assist with the diagnosis, treatment, and prevention of diabetic comorbidities at an early stage.