Application Of Machine Learning In Teaching Tongue Diagnosis In Chinese Medicine

Chinese medicine is the treasure of ancient Chinese science,and is the crystallization of the medical art of the Chinese nation for thousands of years.Tongue diagnosis is one of the most characteristic diagnostic methods in Chinese medicine,which has become a familiar way of treatment for thousands of years of Chinese people.Traditional Chinese medicine students learn tongue diagnosis on the basis of observation,without the aid of precise instruments,only through visual perception to determine disease information.The focus of "looking and knowing is said to be divine" is the observation of the tongue.The blood vessels,muscles,and fluids of the tongue are intuitive,and the fact that tongue changes are closely linked to the health of the body’s five internal organs makes tongue diagnosis one of the most proven diagnostic methods with clear evidence of diagnosis and educational heritage.However,the traditional way of learning tongue diagnosis also has certain disadvantages.The results of tongue diagnosis depend on the subjective diagnosis of students and are influenced by their experience reserve,which hinders the development of tongue diagnosis and diagnosis.Therefore,it is the trend to study the automation of tongue diagnosis and implement the tongue diagnosis learning system,to bring into play its high clinical teaching value,to assist TCM scholars in their daily diagnostic study,to help scholars quickly accumulate tongue diagnosis experience,to promote the academic of TCM tongue diagnosis,to give full play to the role of TCM tongue diagnosis in health development,and to help TCM students improve their tongue diagnosis in the cycle of theory and practice with the help of the tongue diagnosis learning platform.The two major concerns of medical students learning tongue diagnosis are: tongue coating and tongue texture.The strength and weakness of the essence of the five organs of the body is reflected in the color,shape and texture of the tongue,while the depth of the disease and the nature of the evil are reflected in the state of the tongue coating.Combined with the background of today’s rapidly changing science,this paper will focus on the objectification and systematization of tongue diagnosis,mainly involving tongue body localization,separation of tongue texture and tongue coating,and automation of tongue dialectic,and develop a deep learning-based Chinese medicine tongue diagnosis learning aid system for scientific diagnosis of tongue information uploaded by students to help Chinese medicine students gain experience quickly.The main work of this paper is as follows:1.Realize tongue body localization.Labeling the tongue body parts and proposing a RA-Unet model based on convolutional attention mechanism and recurrent neural network to realize the function of tongue body localization,which separates the tongue body,an important research object,from the complex human body background.2.Realize the separation of moss and texture.With the help of professional herbalists,the data set is calibrated for tongue texture and tongue moss according to the TCM tongue diagnosis and dialectic,and the segmentation of tongue texture and tongue moss is realized with the help of RA-Unet model based on convolutional attention mechanism and recurrent neural network.3.Construction of classification models for tongue images.Using Dense Net network structure,a TCM tongue model based on the features related to tongue coating and tongue texture is trained.4.Build a Chinese medicine tongue diagnosis learning system.The trained segmentation and classification model is applied to give corresponding health condition analysis and related daily regimen and dietary precautions to the images uploaded by TCM scholars.