基于无创血糖的糖尿病评估

为了提高糖尿病前期的检出率,在糖耐量受损（IGT）常规诊断方法的基础上,增加糖化血红蛋白作为糖尿病筛查的因素,构建一个IGT检测模型。采集受试者的身高、体质量、腹围、血压、皮脂厚度、空腹血糖和糖化血红蛋白作为模型的特征输入,用K-近邻算法和神经网络对其分类,模型输出包括血糖值正常、IGT和糖尿病。结果显示增加糖化血红蛋白作为分类特征后,神经网络和K-近邻算法的分类准确率分别为88.89%和93.09%,明显高于传统方法的分类准确率（83.33%和78.38%）。本研究提出的IGT检测模型对糖尿病的临床诊断有重要意义。

Diabetes evaluation based on noninvasive blood glucose

Abstract: On the basis of routine methods for diagnosing impaired glucose tolerance (IGT), an IGT detection model is constructed by adding glycosylated hemoglobin as the factor of diabetes screening, thereby improving the detection rate of pre-diabetes. The height, body weight, abdominal circumference, blood pressure, thickness of sebum, fasting blood glucose and glycosylated hemoglobin of subjects were collected as the feature inputs of the model, and then the subjects were classified by K-nearest neighbor algorithm and neural network. The output of the model included normal blood glucose, IGT and diabetes. The results show that after adding glycosylated hemoglobin as the factor of diabetes screening, the classification accuracies of neural network and K-nearest neighbor algorithm are 88.89% and 93.09%, respectively, which are significantly higher than 83.33% and 78.38% of traditional methods. The proposed IGT detection model is of great significance for the clinical diagnosis of diabetes.