CT Image Classification and Detection of COVID-19 Based on Convolutional Neural Network

As Corona Virus Disease (COVID-19) emerged at the end of 2019, traditional detection is mainly carried out using four methods: coronavirus screening detection strips, COVID-19 antibody detection kits, COVID-19 nucleic acid detection and CT detection, and the problem of low efficiency exists. In order to solve the problem of using neural network to detection a large number of data, slow speed, low efficiency, high cost, complex algorithm structure and low accuracy of detection of large data sets at present. In this paper, by collecting known public COVID-19 CT image data sets, a convolutional neural network algorithm based on residual network is proposed to reduce parameter complexity, modify weights and biases associated with neurons, and simplify the overall network structure. This algorithm is used to improve the accuracy of COVID-19 case classification detection and the convergence speed of the model. Through model verification, the accuracy of the proposed algorithm model is 0.985, the precision is 0.805, the area under the curve (AUC) of the ROC curve is found to be 0.852, and the recall rate is 0.897. The results show that the classification detection algorithm model proposed in this paper has higher accuracy than the general image classification model, is more concise in the network model, reduces the complexity, and can be more effectively applied to the detection of COVID-19. The combination of traditional medical imaging diagnosis and deep learning technology helps medical personnel to make more rapid, accurate and effective diagnosis.