| BackgroundHypertension is the most common chronic disease in clinic.Studies have shown that hypertension has many complications and great harm,this greatly affects the quality of patients’s life,and has become an important public health problem in the world.It is reported that the prevalence of hypertension in Chinese adults is 27.9%.It is estimated that there are about 300 million people suffering from hypertension in China,and the number of new patients with hypertension can reach 10 million every year.Although the awareness rate,treatment rate and control rate of hypertension in China are higher than before,the overall level is still at a low level,so the prevention and treatment of hypertension is still facing great pressure.Hypertension is the name of modern medicine,which has not been recorded in ancient Chinese medicine books,but it can be classified as vertigo and headache according to the clinical manifestations of patients.Clinical research shows that traditional Chinese medicine has significant advantages in the treatment of hypertension,in addition to significantly improving symptoms,improving the quality of life,less toxic and side effects,but also has the characteristics of stable blood pressure,long duration and effective reduction of complications.As the main means of TCM syndrome differentiation diagnosis,traditional Chinese medicine four diagnosis is limited by the experience and ability of doctors in clinical application,and it has strong subjectivity and can not guarantee the accuracy of syndrome differentiation,which leads to the uncertainty of TCM curative effect.Therefore,how to improve the accuracy of TCM syndrome differentiation is of great significance to the prevention and treatment of hypertension.Standardizing syndrome diagnosis from an objective point of view is an important direction to improve the accuracy of TCM syndrome differentiation.Taking modern science and technology as a bridge of inheritance and innovation and choosing appropriate scientific and technological means as a breakthrough is the only way to realize the modernization of traditional Chinese medicine as soon as possible.In recent years,with the progress of science and technology,artificial intelligence is constantly developing,and its application in various fields is also increasing.Artificial intelligence deep learning can imitate the way of human thinking,classify and summarize the disordered data step by step,find out the general rules from the disordered massive data,summarize and refine in the repeated learning process,and then draw a conclusion.As the most important algorithm in artificial intelligence computer vision,convolutional neural network has been widely used in image classification,face recognition,automatic driving and other related fields.Therefore,from the perspective of objectification of facial inspection,this study takes the facial image information of the research object as the sample data,constructs the facial image diagnosis model of hyperactivity of liver fire syndrome of hypertension through convolution neural network algorithm,so as to carry out the auxiliary diagnosis of hyperactivity of liver fire syndrome of hypertension,and discusses the feasibility of convolution neural network algorithm in TCM syndrome diagnosis of hypertension.ObjectiveIn order to solve the problem of low level of TCM syndrome differentiation of grassroots medical staff in the community,which affects the clinical efficacy of traditional Chinese medicine in the treatment of hypertension,this study starts from the objectification of facial observation,constructs the facial image diagnosis model of hyperactivity of liver fire syndrome of hypertension,assists the clinical diagnosis of hyperactivity of liver fire syndrome of hypertension,and discusses the application of convolution neural network algorithm in TCM syndrome diagnosis of hypertension,and provide reference and basis for the current research methods of TCM disease and syndrome diagnosis.Method1.The basic information of the subjects was collected from the self-made case report form,including general information(name,sex,age,occupation,etc.)and physical examination information(height,weight,chest circumference,abdominal circumference,etc.).The blood pressure information of the subjects was collected by using the upper arm electronic sphygmomanometer.2.The tongue unit of DS01-A tongue pulse information acquisition system is used to collect the facial image information of the research object.3.According to the related records in Lingshu wuse,combined with the localization map of facial viscera in diagnostics of traditional Chinese medicine,the liver part was selected,and the liver point was determined as the midpoint of the connection between the two pupils and the tip of the nose by facial markers.The face detection algorithm is used to determine the study area.The size of the part varies from person to person.The two pupils and the tip of the nose constitute 1/4 of the rectangular area.4.The subjects were selected from the community residents who had established health records in Qiaozi Town,Huairou District,Beijing.According to the inclusion and exclusion criteria,three groups of subjects were selected:hypertension with hyperactivity of liver fire syndrome group,hypertension without hyperactivity of liver fire syndrome group and healthy group.The information was collected by the case report form and the tongue surface pulse information collection system.The collection time was from August 2018 to January 2019,and the collection place was Qiaozi town community health center,Huairou District,Beijing.5.The image database is established by data enhancement and data preprocessing.Three groups of data in the image database were randomly divided into training set and test set according to the ratio of 8:2.The deep residual network ResNet101 network structure was used as the basic framework of the diagnosis model.The training set image was used to train the model,and the test set image was used to test the model.Result1.A total of 369 subjects were included in this study,including 129 cases of hypertension with hyperactivity of liver fire,accounting for 34.96%;128 cases of hypertension without hyperactivity of liver fire,accounting for 34.69%;112 cases of healthy people,accounting for 30.35%.2.Through the analysis of the basic information of the three groups of research objects,it is found that there are differences in age,chest circumference,abdominal circumference and waist circumference among the three groups of research objects.There is no difference between the hyperactivity of liver fire syndrome group of hypertension and the non hyperactivity of liver fire syndrome group of hypertension.The age of the healthy group is significantly lower than that of the hypertension group;the chest circumference and waist circumference of the hyperactivity of liver fire syndrome group of hypertension are significantly higher than those of the healthy group,and the abdominal circumference is higher than that of the healthy group There were significant differences in the degree of hyperactivity of liver fire and hyperactivity of brain fire between hypertension group and non hypertension group.In terms of blood pressure information,there were significant differences among the three groups in systolic blood pressure,diastolic blood pressure and pulse pressure difference,and there was no difference between the hypertension group with hyperactivity of liver fire and the hypertension group without hyperactivity of liver fire.Health group was significantly lower than hypertension group,in line with the grouping situation.3.The accuracy of the facial image diagnosis model of hypertension with hyperactivity of liver fire was 84.82%.The accuracy of hypertension with hyperactivity of liver fire was 80.35%,hypertension without hyperactivity of liver fire was 89.73%,and the accuracy of healthy people was 84.37%.This model training has 100 iterations,that is,all 2450 training samples have been trained 100 times,and the batch size of training is 32,that is,the number of training images is 32 each time;the learning rate of the first 20 iterations is set to 0.001,and then the learning rate is reduced by 1e-1 every 20 iterations(the learning rate is to control the progress of model training,which has a significant impact on the performance of the model).In order to select the optimal model structure,we use the same database to train ResNet34,ResNet152,Xception and MobileNet network structures,and compare them with ResNet101.The results show that ResNet101 structure has the best classification effect in this study.Conclusion1.The accuracy of facial image diagnosis model of high blood pressure liver hyperthyroidism syndrome constructed by convolution neural network algorithm is high,which is helpful to the clinical diagnosis of high blood pressure liver fire syndrome.At the same time,the diagnosis model constructed in this study can realize the objective quantitative diagnosis of hyperactivity of liver and fire syndrome in hypertension.It is easy to operate in clinical application,and belongs to noninvasive detection method,and has high clinical practicability.2.It is feasible to apply convolution neural network algorithm to the research of objectification of TCM facial inspection.Combining with artificial intelligence method to assist TCM syndrome diagnosis is one of the important directions of TCM syndrome research. |
PDF Full Text Request