The Research On Medical Image Classification Method Of Coronavirus Disease 2019 Based On Deep Learning

With the development of science and technology and the continuous progress of medical equipment,medical imaging technology has gradually become an important way of disease detection.Diagnosing diseases through medical images requires radiologists to manually analyze and compare their features,which is a task that tests the level and energy of radiologists.In recent years,artificial intelligence technology has developed rapidly,and the application scope of deep learning methods has become wider and wider.As a representative algorithm of deep learning,convolutional neural networks have achieved great success in the field of computer vision.Processing medical images through convolutional neural networks to assist doctors in diagnosis has gradually become a popular research direction.Coronavirus disease 2019(COVID-19)is an acute infectious pneumonia with a long incubation period and strong infectivity during the incubation period.Patients need to be detected,isolated and treated in time.Therefore,it is of great significance to automatically analyze medical images through deep learning methods to assist doctors in diagnosing Covid-19.In this regard,we proposed two convolutional neural network models:1.This thesis proposes a COVID-19 chest X-ray image classification method based on the deep learning model CFW-Net.Aiming at the characteristics of high inter-class similarity and low intra-class variance of chest X-ray images,a channel feature weight extraction module(CFWE)is designed,and on this basis,a new convolutional neural network CFW-Net is proposed for classifying COVID-19 chest X-ray images.The method is tested based on two open-source datasets.The experimental results show that the overall accuracy of the CFWNet56-GFC model reaches 94.35%,the accuracy of the COVID-19 category reaches 94.35%,and the sensitivity of the COVID-19 category is 100%.Compared with other methods,the CFW-Net model has better recognition performance on chest X-ray images.2.This thesis proposes a COVID-19 chest CT image classification method based on the deep learning model ECS-Net.According to the characteristics of chest CT images,a parallel multi-branch attention mechanism feature extraction module ECS module is designed.The ECS module improves the feature extraction ability of the network by introducing multiple attention mechanisms.In this thesis,a new deep convolutional neural network ECS-Net is designed based on the ECS module for auxiliary recognition and detection of COVID-19 chest CT images.The method is experimented on the open-source dataset COVID-CT.The experimental results show that the overall accuracy of the ECS-Net131 model reaches 91.81%,and the accuracy,specificity and F1-score reach the highest 93.44%,91.62%and 92.68%,respectively.Compared with other network models,the ECS-Net model has better classification results.