Deep-Learning Based 3D Lung Segmentation

Medical imaging plays a crucial role in diagnosing and treating lung conditions, where accurate lung segmentation in Computed Tomography (CT) scans is a fundamental step in various applications, from disease detection to surgical planning. This study presents a deep learning-based approach for lung segmentation using an artificial neural network trained on a dataset derived from the AAPM Thoracic Auto-segmentation Challenge. By integrating robust preprocessing, dynamic advanced data augmentation, and an optimized U-Net-based architecture, we enhance segmentation accuracy while mitigating overfitting. Our results demonstrate the effectiveness of this approach, achieving a Dice Similarity Coefficient of 0.9468 and an Intersection over Union of 0.8990 on unseen test data, confirming improved generalization and high performance on key evaluation metrics.