Research On Diagnosis Algorithm Of Diabetes And Hypertension Based On Pulse Wave

As a major cause of death worldwide,cardiovascular disease has attracted widespread attention both at home and abroad.Hypertension and diabetes,as important predisposing factors of cardiovascular diseases,can seriously affect cardiovascular health,and timely and accurate detection is an important means of accurate prevention and control.Clinical invasive testing methods are commonly used,which are cumbersome and time-consuming,and may cause discomfort to subjects.Pulse wave signals carry abundant pathological information,which can fully reflect the disease information of diabetes and hypertension.Daily detection of blood pressure and blood glucose values still has serious limitations and is more suitable for clinical application.Therefore,this thesis studies the classification tasks of daily hypertension and diabetes patients on the basis of self-measured multi-modal data sets and publicly available data sets.The main studies are as follows:In view of the problem that single scale convolution in traditional convolutional neural networks cannot fully extract the internal features of pulse waves,a multi-scale pulse wave diagnosis algorithm based on attention mechanism is proposed for classification diagnosis of different diseases.Firstly,the signal granulation information at different scales is extracted through coarse-grained operations,and the multi-scale spatial features of pulse waves were further extracted by using convolutional kernels of different sizes.Secondly,the internal temporal features of pulse wave signals were extracted by using causal dilation convolution to learn more pathological information.Finally,combined with the attention mechanism,the weights of different scales are adaptively adjusted,so as to further improve the effect of disease diagnosis.Conduct experiments on multiple datasets to demonstrate the effectiveness and reliability of the proposed model.Using the function of Python Graphical User Interface,the real-time acquisition and analysis system software based on pulse wave is developed.The software can realize real-time acquisition of human pulse wave signals,and make pulse wave diagnosis more objective and portable through computer technology.A multimodal pulse wave diagnosis algorithm based on graph convolutional neural network is proposed to address the issue that single mode sensors cannot be able to obtain correlation information between pulse wave signals.Firstly,one-dimensional pulse wave data is transformed and fused into three-channel image by Gramian Angular Field,Recurrence Plot and Markov Transition Field.Then,with the idea of transfer learning,three channel image features are extracted through Res Net.Finally,combined with Pearson correlation coefficient and the topological structure of multimode pulse wave,the adjacency matrix is constructed and input into the Graph Convolutional Network to automatically diagnose the disease.The experiment proved that the proposed model achieved 99.6% accuracy in the classification task of diabetes and hypertension,which verifies the effectiveness of the model and fully demonstrates the necessity and superiority of the fusion of pressure pulse wave and photoelectric volume pulse wave.