NIH-Chest_Xray_Retrirval

The dataset is sourced from the publicly available NIH Chest X-ray database, which contains 112,120 frontal-view X-ray images from 30,805 unique patients. Each image is labeled with one or more of 14 common thoracic pathologies identified in the associated radiological reports. From this dataset, we selected 51,480 images representing the 13 most frequent pathologies, including Atelectasis, Consolidation, Infiltration, Pneumothorax, Edema, Emphysema, Fibrosis, Effusion, Pneumonia, Pleural thickening, Cardiomegaly, Nodule, and Mass. These images are organized into three distinct sets: a training set with 38,610 images, a gallery set with 10,296 images, and a query set with 2,574 images. All images are stored in `.npy` format. The training set is used during training, while the gallery and query sets are used during inference.