基于近红外光谱和机器学习的无创血糖浓度回归研究

Research on Non-invasive Blood Glucose Concentration Regression Based on Near-infrared Spectroscopy and Machine Learning

无创血糖浓度预测是目前的热点研究问题,其预测精度容易受到各种不同因素干扰。针对基于近红外光谱的无创血糖浓度回归问题,重点分析了指端相邻位置光谱散射对回归精度的影响。OGTT实验是无创血糖浓度预测问题中的采集数据的经典方法,通过OGTT实验采集了指端区域近红外光谱及对应的血糖浓度数据,并利用支持向量机回归算法预测血糖浓度。总共设计了3组实验,实验结果表明,多区域数据的采集和平均能有效降低指端相邻位置对光谱不同散射带来的回归误差。

Non-invasive blood glucose concentration prediction is a hot research issue at present,and its prediction accuracy is easily interfered by various factors. Aiming at the problem of non-invasive blood glucose concentration regression based on near-infrared spectroscopy,in this paper,the analysis of the influence of spectral scattering is focused on at the adjacent position of the fingertip on the regression accuracy. The OGTT experiment is a classic method of collecting data in the problem of non-invasive blood glucose concentration prediction. Through the OGTT experiment,the near-infrared spectrum of the fingertip area and the corresponding blood glucose concentration data are collected;and the support vector machine regression algorithm is used to predict the blood glucose concentration. A total of 3 sets of experiments are designed.The experimental results show that the collection and averaging of multi-region data can effectively reduce the regression error caused by the different scattering of the spectrum at the neighboring position of the finger.