Research on the Diagnosis of Corona Virus Disease Based on Deep Learning

Abstract

The outbreak of Corona Virus Disease (COVID-19) poses a great threat to the safety of human life and property worldwide. A computer-aided diagnosis method based on deep learning can effectively improve the efficiency of COVID-19 diagnosis, avoid the risk of exposure to infection, and alleviate the problem of strained medical resources. In order to achieve a lightweight deep learning network and meet the practical needs of clinical new coronary pneumonia detection, a COVID-19 CT identification algorithm based on the lightweight Transformer (LWT) model is proposed in this paper. First, a hierarchical Transformer is constructed to reduce the number of CheXNet parameters. Then, the attention mechanism is incorporated into the hierarchical Transformer to further extract the global and local information of COVID-19 and improve the interaction ability of intra-group and inter-group features. Finally, tests are conducted on the publicly available COVID-19 CT dataset, and the results show that the number of parameters of this model is greatly reduced compared with the traditional CT recognition model, and the performance is better than the comparison method in several evaluation indexes, which has positive implications for COVID-19 computer-aided diagnosis.