Beyond Pixels: Tackling COVID-19 with CNN, Transfer Learning, and Traditional ML Classifiers

This study addresses the critical urgency of early Covid-19 detection, a vital aspect in preventing potential fatalities due to delayed treatment. Deep Learning (DL) techniques offer a promising avenue for early screening using Chest X-ray images to facilitate medical diagnosis. In pursuit of this objective, Convolutional Neural Networks (CNNs), namely VGG16, ResNet50, and EfficientNet, were harnessed to detect the presence of Covid-19 in X-ray images, utilizing a dataset sourced from KAGGLE. These CNN models underwent rigorous training on preprocessed images representing both Covid-19 and Pneumonia cases, with a focus on enhancing edge detection specific to Covid-19. The architectural design of these networks was meticulously optimized for the classification task, incorporating layers and learning parameters. The study's findings underscore the remarkable similarity in outcomes produced by VGG16, ResNet50, and EfficientNet, taking into account key factors such as training time, learning rate, and overall accuracy. This suggests that these DL-based models offer comparable performance in the domain of Covid-19 detection, reaffirming their robustness and efficacy in medical imaging analysis, potentially contributing to improved patient care. Furthermore, the study emphasizes the significance of fine-tuning model parameters and considerations regarding computational resources when implementing DL solutions for medical image analysis, thereby presenting important insights for the efficient utilization of resources in clinical settings.