About the Methodology

The paper details our approach using a ResNet18 convolutional neural network. Subjects were screened using standardized airway positioning and images were processed through a custom augmentation pipeline to improve model robustness across diverse patient populations.