Design Of Electronic Nose System For Detecting Diabetes Gas Markers

At present,the diagnosis of diabetes is usually carried out in the traditional way of blood test.It is an invasive diagnosis,which takes a long time and is not convenient to use.Patients can't find diseases in time and delay the best time of treatment.Previous studies have shown that the content of acetone in exhalation of diabetic patients is significantly higher than that in normal people.Therefore,the content of acetone in human breath is used as the basis for detecting diabetes,and it is possible to realize the diagnosis of noninvasive diabetes.The purpose of this project is to design a set of "artificial olfactory" electronic nose system through the combination of metal oxide semiconductor gas sensor array and pattern recognition algorithm.The accurate identification of trace acetone and the automation and intellectualization of the whole detection process were completed,so as to provide technical reference for subsequent rapid noninvasive diabetes detection.The main tasks completed are as follows:(1)Design the electronic nose hardware system.The system mainly includes sensor selection,array circuit,dynamic test,data acquisition and so on.Choosing the type and quantity of sensor based on the composition and concentration of the exhalation of diabetes mellitus,the electronic nose array is formed.The electrical parameters of the sensor are analyzed and the array circuit is designed.Three kinds of gas samples are prepared through three gas flow controllers,and dynamic tests are completed.According to the number and accuracy of sensors in the array,data acquisition card is selected.(2)Data processing and sensor array optimization.The original data are smoothed to complete data preprocessing using smoothing filtering method.In the process of response,the data that each sensor can express the feature of the sample is extracted,and the data feature is extracted.According to the data collected from experiments,the sensitivity,selectivity,correlation and repeatability of the sensors were analyzed,which eliminates the poor performance of the sensor and completes the sensor array optimization.(3)Recognition algorithm research.According to the requirements of electronic nose system identification,three algorithms are selected: BP neural network,BP neural network based on principal component analysis,support vector machine,which constitute three identification models respectively,and perform performance comparison and analysis to determine the best recognition model.The experimental results show that the BP neural network model based on the principal component analysis is similar to the BP neural network model,and the recognition accuracy is similar.However,the former reduces the complexity of the network and reduces the training time of the model.The recognition model designed by SVM for three kinds of samples has higher recognition accuracy than the BP neural network model optimized by principal component analysis.The recognition accuracy is close to 100%,at the same time,the training time is short and the robustness is good.(4)Software design of electronic nose system.It mainly includes data acquisition,pattern recognition and patient information management human-computer interaction interface.The data acquisition man-machine interface is developed based on the C# platform,which completes the operation of the sensor's response data,drawing the dynamic response curve,data processing and so on.The pattern recognition human-machine interaction interface is based on C# and MATLAB mixed programming.It completes the fast recognition and detection of gas samples.The patient information management human-computer interaction interface is based on the ADO.Net technology in C# to record and query the patient's basic information and diagnosis results.