A MultiResUNet-based Deep Learning Architecture for Accurate Lung Segmentation in Chest X-Rays

Accurate chest X-ray segmentation (CXR) is vital for disease diagnosis and monitoring, such as Covid-19. This study proposes an advanced deep learning-based framework for enhanced CXR segmentation incorporating a multi-resolution methodology to improve lung segmentation accuracy. We experimented on a dataset of 3616 COVID-19 images and their corresponding segmentation masks from the COVID-19 Radiography Database, evaluating different state-of-the-art models. The results demonstrated that the proposed MultiResUNet outperforms existing methods with faster convergence and an accuracy of 99.52%. The results show that the proposed MultiResUNet model improves the accuracy and precision of CXR segmentation, especially for detecting COVID-19, and enables the development of more robust AI-driven tools in medical image analysis.