A Study On Coupling Electrochemical Detection Of Glucose And Insulin And Preliminary Typing Of Diabetes Mellitus

Diabetes mellitus is a non-communicable disease that can pose a major threaten to people's health.The rate of mortality caused by its chronic complications has kept increasing,which puts a huge burden on social economy and public health care.Blood glucose and insulin levels in the body are basic criteria for diagnosis of diabetes and the evaluation of islet cell function,and detecting both of concentration levels of these two substances is crucial for diabetes diagnosis and typing and for islet ? cell function evaluation.Therefore,a portable,commercial diagnostic device that can detect blood glucose and insulin levels at the same time can potentially find a broad range of applications.In view of this,this study utilizes the electrochemical detection principle and machine learning classification algorithms to examine the simultaneous detection of blood glucose and insulin levels,and present a theoretical prediction model for the preliminary diagnosis and typing of diabetes based on the measured glucose and insulin concentrations.Our work can be summarized as follows:(1)The principle for the electrochemical detection of glucose and insulin levels was discussed,and an electrochemical detection platform was set up.With screen printing electrodes serving as the sensing components,the working electrode was modified using the nickel hydroxide and electrodeposition technology.The electrode modification performance was characterized using electron microscopy,and electrochemical impedance spectroscopy experiments were performed on the modified electrodes.The results show that the modified electrodes have good electrical activity and can be used for quantitative detection of glucose and insulin levels.(2)A glucose and insulin detection scheme was designed.Based on the biological structure of glucose and insulin and the principle for electrochemical detection,cyclic voltammetry and time current methods were used to detect the concentrations of glucose and insulin solutions.It is experimentally validated that our proposed electrochemical detection system is capable of quantitively detect glucose and insulin solutions at different concentrations.The detection limit of glucose was 45.9 ?M,with a sensitivity of 11.12 ?A/mM;the detection limit of insulin was 138 nM,with a sensitivity of 15.3 ?A/?M.(3)Feature parameters were extracted from a large amount of cyclic voltammetry experimental data to build a feature data set.Based on this data set and by using linear regression and model prediction,the concentrations of glucose and insulin in simple alkaline mixture and complex fetal calf serum mixture were obtained.The repeatability and selectivity of nickel hydroxide modified electrodes in detecting glucose and insulin from mixed solutions were also verified.(4)Based on the feature data set,five methods,i.e.,logistic regression,support vector machine,random forest,Adaboost,and KNN,were used to investigate diabetes typing.The results show that the comprehensive typing accuracy of type 1,type 2 and gestational diabetes can reach up to 98.02%.In addition,four evaluation parameters,i.e.,accuracy,precision,reproducibility and F1-score,were used to compare the typing performance of the five methods.The results show that logistic regression and KNN perform better in diabetes typing than the other methods,delivering an accuracy of up to 97% and 98.02%,respectively.With its unique combination of electrochemical detection method and machine learning,our proposed scheme can detect the concentrations of glucose and insulin in mixed solutions relying on the mutual interference between these two substances.As a brand-new subversive attempt in the detection of multiple substances,this scheme enables coupled detection of glucose and insulin levels and preliminary typing of diabetes based on the detection results.Due to its advantages of low costs,high efficiency,and high portability,this scheme will enjoy a promising prospect of application in preliminary diagnosis and typing of diabetes as well as the monitoring of diabetes in home settings.