Deep Learning Based Covid-19 Diagnosis and Severity Assessment Using Computed Tomography Imaging

The COVID-19 pandemic has highlighted the need for advanced diagnostic tools for early detection and management. This study presents a Deep Learning (DL) approach using CT scans to detect COVID-19 and assess lung involvement. A Convolutional Neural Network (CNN) achieves 94% accuracy in distinguishing COVID-positive from negative cases. Image preprocessing, including Gaussian blurring and Otsu's thresholding, enhances lung region clarity. Lung segmentation isolates critical areas, enabling severity classification (mild, moderate, severe) based on opacity-to-lung area ratio. Using the SARS-CoV-2 CT dataset, the CNN model with Transfer Learning (TL) via ResNet reaches 96.74 % accuracy. This method provides a valuable diagnostic tool for real-time healthcare decision-making during pandemics.