Extraction And Analysis Of TCM Tongue And Pulse Information And Design Of Auxiliary Diagnosis System

The diagnosis of traditional Chinese medicine(TCM)is a special way of medical diagnosis in China.TCM has a long history of development,it is also the crystallization of the wisdom of the Chinese nation for thousands of years.Tongue and pulse information is an important basis in the diagnosis of TCM,but the tongue and pulse information in the clinical diagnosis process will be affected by the environment,doctors' experience and many other aspects,so at this stage,there is a greater subjectivity,uncertainty and non repeatability,which is not conducive to the development of the objectification and modernization of the diagnosis of TCM.With the development of information and technology,it provides rich means for the collection and analysis of tongue and pulse information.Standardized collection and analysis of tongue and pulse information can better serve the clinical diagnosis of TCM.The main purpose of this thesis is to establish a data set by collecting the tongue and pulse information of the corresponding population,and accurately recognize the physiological information reflected by the tongue and pulse information of human body by modern computer technology,provide objective diagnosis basis for TCM clinic,and promote the objective and modern development of TCM diagnosis.The main work of this thesis is divided into two parts.The first part is the clinical collection of pulse data and the construction of machine learning model for Pregnancy Discrimination.The first step is to design a complete clinical collection scheme of pulse signal,then collect the corresponding pulse information through pulse acquisition instrument in hospital and school,and filter the pulse data with zero phase low-pass filter.Then we extract 21 dimensional feature data in the time domain and frequency domain of pulse signal at the same time,and use Random forest algorithm to analyze the correlation between each feature,and screen the important features to build a tagged pulse data set.Then,Ada Boost algorithm is used to build the machine learning model of pregnancy discrimination,and the model is trained and predicted to achieve the function of clinical auxiliary diagnosis,the accuracy of cross validation of pregnancy recognition model was 84.4%.The second part is the clinical collection and classification of tongue image.Firstly,the standard tongue image is collected through the tongue image acquisition instrument,and the collected tongue image data are annotated by five TCM experts combined with clinical experience.Next,we use the multi-color channel fusion OSTU threshold segmentation method to extract the tongue body part of the collected image,which can improve the classification accuracy by removing the useless part of the tongue image.Then we use the method of data augmentation to increase the amount of data,build the tongue image data set with labels,and build the convolutional neural network(CNN)model to train and test the data set,the accuracy of tongue image 4-classification reaches 91.2%.Finally,a visual operation platform is built based on the trained model to realize an intelligent tongue image aided diagnosis system of TCM.The system can automatically classify the tongue image input by the user,and provide the results of tongue image recognition,the result of tongue image aided diagnosis,the body condition that may be accompanied and the suggested conditioning scheme,so as to better realize its auxiliary diagnosis function.