Classification of COVID-19 Utilizing CT Scan Images Employing the EfficientNet Model

In the realm of medical imaging, particularly for the critical task of distinguishing between COVID-19 and non-COVID cases, the analysis of CT scans has emerged as a formidable challenge. The primary concern is achieving high accuracy and low loss in predictive models to support diagnostic decisions. To tackle this, various deep learning architectures, including ResNet50, Inception, and VGG19, were employed, yet none matched the proficiency of the Proposed EfficientNet model. Utilizing a dataset enriched through augmentation techniques to yield 17,099 images, the models were trained and tested on a uniform framework of 15 epochs, a consistent learning rate of 0.00001, and an image resolution of 224 × 224. In this standardized environment, the proposed EfficientNet model, with a strategic batch size of 64 and 100 steps per epoch, showcased a substantial performance improvement, attaining a near-perfect training accuracy of 99.54% and a test accuracy of 98.24%, with impressively low training and test losses of 0.015 and 0.145, respectively. Compared to other models, this translates to an accuracy improvement of over 4% on training and approximately 14% on testing. Conclusively, the Proposed EfficientNet model marked an improvement of roughly 5% over the closest competitor, presenting a significant advancement in the field of medical image classification for the detection of COVID-19.