Research On Early Diagnosis And Risk Prediction Of Disease Based On Deep Learning And Medical Data

With the wide application of computer technology,electronic healthcare records(EHR)has also increased rapidly,and data driven medical data analysis methods have emerged as the times require.Through reasonable analysis of medical data,early diagnosis of disease can be achieved in the early stage of illness,or we the health status of patients can be analyzed based on their medical records and the risk of some diseases in the future can be predicted.However,because medical data have many different formats,some records are incomplete and have a lot of noise,it is difficult to analyze medical data accurately.In this paper,the latest model of deep learning is used to design different neural network structures for the most common medical data formats,medical examination images and electronic medical text records,and early diagnosis and risk prediction of diseases are realized.For medical examination image data,this paper takes brain functional magnetic resonance images(fMRI)as an example,transforms the image data to matrix at first,and calculates the functional correlation coefficients between different brain regions.After that,an Autoencoder structure which is the most compatible with this data is designed,this model can automatically extract the features of brain network,analyze brain health status,and make early diagnosis of Alzheimer's disease.For electronic medical text record data,a combination model of 3D convolutional neural network and space pyramid pooling is first built in this paper.Through the 3D convolution structure,the internal features of records and the time series features between different records can be extracted,through the spatial pyramid pooling structure,the data with arbitrary length can be processed by the model,and risk prediction of heart failure and diabetes can be achieved.The experimental results of the two models show that deep learning has great advantages in medical data processing and the effect of early diagnosis and risk prediction is far better than other traditional machine learning algorithms.In addition,the most influential factors for diseases can be analyzed based on the experimental results,which could help doctors to analyze the pathogenesis of diseases.Moreover,this method can be easily extended to other kinds of medical data analysis or disease diagnosis and prediction tasks,and it has great significance on effective utilization of medical data,improving early detection and cure rate of diseases,promoting the level of medical healthcare.