Traditional Chinese Medicine(TCM)mainly realizes constitution identification through "observing,hearing,inquiring and cutting".However,this method relies heavily on doctors’ professional knowledge and clinical experience,and the diagnostic results are not objective,so it is difficult to measure the results by unified standards.With the development of neural network theory,the application of neural network algorithm in medical field has been more and more extensive,and has achieved good results.Aiming at TCM tongue diagnosis,firstly,TCM tongue image should be segmented,and then TCM constitution recognition should be realized based on segmented tongue image recognition.Aiming at the TCM consultation scenario,the key is to realize the TCM constitution identification based on the TCM constitution questionnaire data.Therefore,this paper focuses on the theme of "TCM constitution identification" and does the following three aspects based on neural network:(1)For the TCM tongue diagnosis scenario,a TCM tongue image segmentation algorithm based on the attention mechanism and Res Ne Xt is proposed to realize the segmentation of the tongue part in the tongue image.The proposed algorithm was compared with Grab Cut,FCN and UNet algorithms,and four evaluation parameters such as accuracy,Dice coefficient,m IOU coefficient and misclassification error were used to evaluate the performance of the algorithm.The experimental results show that the proposed algorithm can achieve tongue image segmentation more accurately.(2)For the TCM tongue diagnosis scenario,two TCM constitution identification algorithms based on convolutional neural network were proposed.One is the TCM constitution identification algorithm based on multi-scale filtering convolution neural network;the other is the TCM constitution identification algorithm based on parallel dense convolution neural network.The tongue image data set is used to train the neural network and generate the TCM constitution identification model.The two proposed algorithms were compared with Res Net,Dense Net,Rex Ne Xt,Vi T and T2T-VIT algorithms,and three evaluation parameters of accuracy,accuracy and recall were used to evaluate the algorithm performance.The experimental results show that when the second algorithm proposed in this work is applied to the problem of TCM constitution identification,it has better classification effect.(3)For the TCM consultation scenarios,two TCM constitution identification algorithms based on artificial neural network were proposed.One is the TCM constitution identification algorithm based on ANN and GBDT feature transformation;the other is the TCM constitution identification algorithm based on residual DNN and multinetwork fusion.Questionnaire data sets are used to train the neural network and generate the TCM constitution identification model.The experimental results show that the second algorithm proposed in this work has better classification effect when applied to TCM constitution identification problems.er combination in less times,so as to realize the TCM constitution identification more quickly and accurately. |