Abstract
Medical imaging is crucial in detecting pneumothorax, a potentially life-threatening condition characterized by air accumulation in the pleural cavity. Chest X-Rays (CXR) and Computed Tomography (CT) scans are vital tools for diagnosing pneumothorax, offering detailed insights into the chest cavity. The accurate delineation of lung regions and the detection of lung diseases pose significant challenges. Variations in patient anatomy, limited training data, and the complexity of lung diseases contribute to the difficulty of the task. Recent approaches, such as advanced U-Net architectures with residual connections and 3D U-Net variants for lung nodule segmentation, are enhancing the precision and efficiency of lung image analysis. This research presents a novel approach that combines Deep Learning (DL) based segmentation and disease detection techniques to enhance accuracy in CXR image analysis. The proposed methodology comprises two core components: A U-Net-inspired segmentation model with residual connections for precise lung region extraction and a Convolutional Neural Network (CNN)for disease detection.