Deep Learning-Based Image Classification for COVID-19 Disease: A Comprehensive Review

Abstract

The diversity of biological data being generated today is increasing. Among all the data being generated, the coronavirus disease (COVID-19) outbreak has underscored the importance of rapid, reliable diagnostic techniques. In this context, automatic disease diagnosis using deep learning algorithms with medical image data has become an area of research interest. To identify gaps in the literature, investigate methodological trends, and suggest directions for future research, this review paper critically evaluates the use of deep learning techniques in the categorisation of medical images for the diagnosis of COVID-19. It utilised both a systematic bibliometric review and a systematic review of the literature as a methodology to assess the relevant literature. The literature search yielded a total of 1,138 records; 775 articles were then screened, and 46 studies were included in the qualitative analysis of the literature. The increase in research on medical imaging-based diagnosis has been considerable between the years 2020 and 2025, as evidenced by the reviewed studies included in this article. In general, deep learning methodologies yield superior performance when compared to traditional machine learning methods. The studies reviewed demonstrate that the use of convolutional neural network models, transfer learning, ensemble learning, and hybrid approaches yields a high level of performance for classifying chest X-ray and computed tomography images. The review identifies existing research gaps and also indicates future directions of research to create reliable artificial intelligence-based diagnostic tools for use in medical imaging.