基于舌象参数与多指标特征联合的2型糖尿病风险预测模型

目的:将舌象参数与基本生理指标相结合,运用机器学习算法建立糖尿病风险预测模型.方法:应用TDA-1型数字舌象仪和舌诊分析系统获取舌象参数,分析糖尿病前期组与糖尿病组基本生理指标、实验室检查与舌象参数统计学差异,借助4种经典算法建立糖尿病风险预测模型.结果:生理指标与舌象参数联合特征支持向量机糖尿病前期预测模型性能最佳,...

A Risk Prediction Model for Type 2 Diabetes Based on The Combination of Tongue Imaging Parameters and Multi-index Features

Objective:To combine tongue parameters with basic physiological indicators and use machine learning algorithms to build a diabetes risk prediction model.Methods:The TDA-1 digital tongue imager and tongue diagnosis analysis system were used to obtain tongue parameters,and the basic physiological indicators,laboratory tests and statistical differences of tongue parameters in the pre-diabetes group and diabetes group were analyzed.Establish a diabetes risk prediction model with 4 classic algorithms.Results:Physiological indexes and tongue-like parameters combined with the support vector machine pre-diabetes model had the best performance,F1 was 0.81,Precision was 0.71,and Recall was 0.94.Physiological indicators and tongue-like parameters combined with neural network type 2 diabetes prediction model AUC was 0.83 and Precision was 0.73.Conclusion:Combining basic physiological indicators with features of tongue parameters can improve the classification effect of SVM pre-diabetes prediction model and neural network type 2 diabetes risk prediction model,which is in line with the clinical needs for risk prediction of type 2 diabetes.