以"络病学说"构建的2型糖尿病周围神经病变电导法经络测量预警模型及其外部验证OA

Objective To develop a screening model for diabetic peripheral neuropathy(DPN)in patients with type 2 diabetes mellitus(T2DM)based on meridian electrical conductance measurement guided by the traditional Chinese medicine(TCM)Collateral Disease Theory and to evaluate its screening performance and clinical applicability through in-dependent validation.Methods Clinical data from 189 patients with T2DM treated at Guangdong Provincial Hospital of Chinese Medicine were collected,including 53 patients with DPN and 136 without DPN.Demographic and clinical charac-teristics and electrical conductance values at selected acupoints were collected.Because meridian electrical conductance measurements and laboratory tests were not performed on the same day,glucose-related laboratory data obtained within 1 month before the meridian measurement were extracted,including fasting plasma glucose(FPG),2-h postprandial plas-ma glucose(2hPG),and glycated hemoglobin A1c(HbA1c).For patients with multiple FPG or 2hPG measurements during this period,the mean value was used for the primary analysis,whereas the HbA1c measurement closest to the date of meridian measurement was selected.Least absolute shrinkage and selection operator(LASSO)regression was used to i-dentify key electrical conductance features of the acupoints,followed by construction of a multilayer feedforward neural net-work model.Synthetic Minority Over-sampling Technique(SMOTE)was applied during model training to address class imbalance in the training set.Model performance was evaluated using accuracy,sensitivity,specificity,and F1 score and was subsequently validated in a separate cohort of 30 patients.Results No statistically significant differences were ob-served between the DPN and non-DPN groups in age,sex,body mass index,or glucose-related indicators at baseline.LASSO regression identified 11 key electrical conductance variables,including left Lidui(ST45),right Zhongchong(PC9),and left Chongyang(ST42).The optimal neural network model,with four hidden-layer nodes and a cutoff value of 0.4,showed good performance in the training set,with an accuracy of 97.35％,sensitivity of 90.57％,specificity of 100.00％,and F1 score of 95.05％.In the separate validation cohort of 30 patients,the model achieved an accuracy of 90.00％,sensitivity of 87.50％,specificity of 90.91％,and F1 score of 82.35％.Conclusion Guided by the TCM Collateral Disease Theory,the neural network screening model based on meridian electrical conductance measurement dem-onstrated good discriminatory performance and relatively stable results in identifying DPN among patients with T2DM.This noninvasive,convenient,and economical approach may provide a promising TCM-based strategy for early screening and risk assessment of DPN.