Enhancing Covid-19 Detection through Deep Learning- Based Segmentation of Chest X-ray Images

Accurate diagnosis of Covid-19 is pivotal for timely patient care and public health interventions. In this study, we propose an innovative approach to enhance diagnostic precision through deep learning-based image segmentation of chest X-ray images. We systematically assess three leading segmentation architectures: U-Net, ResNet18-U-Net, and MobileNet-U-Net, utilizing a comprehensive dataset comprising 11,956 Covid-19 chest radiographic images, each matched with accurate lung segmentation masks. The dataset is thoughtfully partitioned into a 70:30 training and testing sets. Rigorous quantitative analysis employing critical performance metrics, including Dice Coefficient, Recall, Precision, Binary Accuracy, and Intersection over Union (IOU), unveils MobileNet-U-Net as the standout performer, consistently yielding a Dice Coefficient of 95.25%, Recall of 93.98%, Precision of 97.62%, Binary Accuracy of 98.19%, and IOU of 90.06%. These quantifiable outcomes underlines the potential of our approach to significantly elevate the accuracy of Covid-19 diagnosis. By providing clinicians with a potent tool for precise decision-making in patient care, our research contributes to the ongoing efforts to combat the pandemic.