Thoracic Disease Classification

This study introduces a segmentation-based deep fusion network (SDFN) to improve the automatic diagnosis of thoracic diseases using chest x-ray (CXR) images. Unlike existing methods that rely on entire CXR images, SDFN focuses on higher-resolution lung regions to reduce noise and improve disease detection. The approach demonstrated more accurate disease classification and better lesion localization compared to traditional methods, as evaluated on the Chest X-ray 14 Dataset.