Research On Noninvasive Blood Glucose Measurement Model By Near-infrared Spectroscopy

Diabetes is a serious threat to human health.Currently,there is no radical cure for diabetes,and the blood glucose concentration in diabetics is maintained at normal level with frequent detection and drug treatment.Clinically,the blood glucose concentration is detected with a finger-prick test based on the electrochemical or photochemical principles,which not only brought pain to patients,but also increased the risk of infection.Near-infrared spectroscopy technology is a kind of noninvasive blood glucose measurement technology with great application prospect,the accuracy and robustness of which is key to its clinical application.Based on this,the paper studies the near-infrared noninvasive blood glucose detection model to improve its accuracy and robustness.(1)According to the research status of noninvasive blood glucose detection technology by near-infrared spectrum,the research direction was determined.The measurement wavelength,measurement method and detection location of near-infrared spectrum were determined by synthesizing various factors.In order to ensure the reliability of subsequent experiments,the validity and repeatability of the near-infrared spectrum were tested.(2)Considering the nonlinear relationship between the blood glucose concentration and near-infrared absorption spectra,the paper presents the near-infrared noninvasive blood glucose detection model based on PSO-2ANN to overcome individual differences in the detection process of blood glucose concentration.The PSO-2ANN model considered two artificial neural networks(ANN)with certain structure and arguments as the submodules,and was built up after optimizing the weight coefficients of the two networks by particle swarm optimization(PSO).According to the results of the experiments in vivo,the robustness of PSO-2ANN model is improved and the individual differences resulting from environmental and physiological factors are effectively overcome by calibrating the PSO-2ANN model everyday.(3)Further studies have found that the detection process of blood glucose concentration is influenced by environmental factor and physiological status,and there is fluctuation rule to the blood glucose concentration.While the PSO-2ANN model does not consider these phenomena.Besides,the near-infrared noninvasive blood glucose detection technology aims to help diabetics to self-management of blood glucose level,and the daily calibration of the blood glucose detection model is detrimental to the household promotion of the noninvasive blood glucose meters.The paper continues to put forward the near-infrared noninvasive blood glucose detection model based on PCA-NARX.The feature vectors are extracted by principal component analysis(PCA)from ambient temperature,ambient humidity,systolic pressure,diastolic pressure,pulse rate,body temperature and near-infrared absorption spectroscopy.The blood glucose noninvasive measurement model is established based on nonlinear autoregressive with exogenous input(NARX)with the feature vectors as independent variables and the measurement of blood glucose concentration as dependent variable.The results of experiment in vivo show that the accuracy and robustness of the PCA-NARX model are superior to the PSO-2ANN model,and there is no need for calibrating the PCA-NARX model in a short time.The paper studies the noninvasive blood glucose detection model based on near-infrared spectrum,which improves the accuracy and robustness of the noninvasive blood glucose model,and promotes the improvement of the noninvasive blood glucose detection technology.