Study On The Method Of Arteriosclerosis Detection Based On CNN With Multiple Physiological Signals

In recent years,cardiovascular disease has become the world’s leading cause of death.Studies have shown that the main cause of cardiovascular events is the continuous development of arteriosclerosis,early diagnosis,timely control and diagnosis and treatment of arteriosclerosis can effectively prevent and control the development of cardiovascular disease.Therefore,it is of great practical significance to complete the screening of early arteriosclerosis in large populations.However,due to the shortage of medical resources in China,the arteriosclerosis detector based on the life information detection method suitable for detecting arteriosclerosis only exists in large public hospitals,and it is difficult to popularize the public.Therefore,it is urgent to propose a new efficient diagnostic algorithm for assisting small and medium-sized vital information monitoring systems to complete arteriosclerosis screening.At present,there are a large number of automatic predictions of cardiovascular diseases based on physiological signals,and many of the automatic research algorithms based on deep learning have achieved good results in the corresponding prediction tasks.However,there is no similar study to apply the deep learning method to the classification and recognition tasks of peripheral arteriosclerosis.Therefore,this paper proposes two models based on Convolution Neural Network(CNN)for classification and identification of peripheral arteriosclerosis.The specific research contents are as follows:(1)Analytical judgment can be used to complete the multiple physiological signals of automatic classification and identification of arteriosclerosis,and build a multi-physiological signal acquisition system.The hardware components of the acquisition system mainly include the main control board module,the ECG module,four pulse modules and four Blood pressure module;the main functions of the software components of the acquisition system include user management,acquisition and storage,and result display.(2)The clinical data was collected by using a multi-physiological signal acquisition system,and the data was calibrated by experts,and then the data of pretreatment was used to construct a database of peripheral arteriosclerosis.(3)A classification model of arteriosclerosis based on multi-physiological signals of one-dimensional CNN is proposed.Because there are few mature frameworks for predicting one-dimensional multi-physiological signals,it is necessary to design good performance according to the characteristics of multiple physiological signals.model.In the one-dimensional CNN model,the multi-task convolutional neural network classification is also introduced,and the factors affecting the classification accuracy rate are predicted together as prediction tasks.The experimental results show that the method can effectively improve the performance of the network.(4)A classification model of arteriosclerosis based on two-dimensional CNN for multiple physiological signals is proposed.The model is based on the VGG16 model for light weight improvement.The input part is the spectrum matrix of the original sample after short-time Fourier transform.By comparing the design with the VGG16 model,it is verified that the model can guarantee network prediction.At the same time of performance,the network’s trainable parameters and computing resources are greatly reduced.