Pretrained-ResNet-based COVID-19 CT image classification

Abstract. The COVID-19 pandemic has significantly affected global public health and the economy. Clinically, compared to nucleic acid testing, CT imaging offers a more intuitive display of disease progression, particularly in cases where patients exhibit atypical symptoms or when nucleic acid test results are inconclusive. In such scenarios, CT imaging serves as a valuable supplementary diagnostic tool. However, the task of classifying COVID-19 CT images presents numerous challenges. Firstly, the imaging features associated with COVID-19 exhibit significant heterogeneity, often overlapping with those of other pulmonary diseases such as pneumonia or tuberculosis, which complicates the classification process. Additionally, CT images typically contain noise and artifacts that can interfere with The model's capability to accurately differentiate between various conditions. This paper proposes a ResNet-based method for the classification of COVID-19 CT images. Experimental results demonstrate that the ResNet-based approach offers significant advantages, delivering high accuracy and strong generalization capabilities in classifying images related to COVID-19.