Hybrid Learning Approach for COVID-19 Lung Infection Segmentation using CT Imaging

Abstract

The global spread of COVID-19 put significant pressure on healthcare systems, making early diagnosis essential to save lives and limit further transmission. Chest CT scans are effective for detecting lung infections, but manual analysis is time- consuming and depends on the radiologist’s experience, which can delay treatment or lead to inconsistent interpretations. To address this, we propose a deep learning–based system that can automatically detect and segment COVID-19–infected regions in CT scans using a single end-to-end model. Our method is robust to varying lesion sizes, shapes, and image quality, including noise and low contrast. Experiments on public CT datasets demon- strate high accuracy for both segmentation and classification tasks. This system can support radiologists by reducing manual effort, speeding up diagnosis, and helping patients receive timely treatment, especially during pandemic peaks.