Covid-19 Detection Using Deep Transfer Learning Approach Through CT Scan Images

Abstract

The advancement of deep learning techniques has significantly enhanced medical diagnostics by enabling early and accurate disease detection. COVID-19, a highly infectious respiratory illness, has posed severe health and socioeconomic challenges worldwide. Early diagnosis is vital in mitigating its spread. Computed Tomography (CT) imaging has proven useful for non-invasive detection of COVID-19 due to its affordability and accessibility. However, interpreting these scans requires expert radiologists, whose assessments may vary and are subject to errors. Transfer learning offers a promising alternative for developing automated diagnostic tools with high precision. In this study, we introduce a novel 12-layer Convolutional Neural Network (CNN) model and integrate it with three established pre-trained models—VGG16, VGG19, and ResNet50—to improve classification performance. We conduct a detailed comparative analysis to evaluate the effectiveness of our proposed model against these benchmarks. The experimental outcomes demonstrate that our model surpasses the existing approaches, achieving over 98% classification accuracy, an F1 score of 0.96, and precision and recall values of 0.97. The proposed framework is not only accurate but also computationally efficient, making it suitable for rapid COVID-19 detection in clinical environments.