The following datasets are part of a future paper. Kidney OCT dataset Three type of tissues were sampled: cortex, medulla, and pelvis. Image size: 185*210 pixels. The OCT dataset comes from 10 porcine kidneys. For each kidney and tissue, there are 30 volumes. Each volume contain 20 images. Total number of images: 10*3*30*20 = 18,000. The same dataset was partitioned in three different levels: image (folder: split_random), volumen (folder: split_volume), and subject (folder: split_subject). Chest X-ray repository A chest X-ray repository was built using the ChestX-ray8 dataset, the CheXpert dataset, the MIMIC-CXR dataset, and the PadChest dataset from the TorchXRayVision library. A chest X-ray repository was built using the ChestX-ray8 dataset, the CheXpert dataset, the MIMIC-CXR dataset, and the PadChest dataset from the TorchXRayVision library. The chest X-ray repository was partitioned into four folds using three different partitioning levels. In image-level partitioning(folder: split1_random), images were randomly distributed across four folds. In patient-level partitioning (folder: split2_patient), all images from the same patient were assigned to the same fold. Finally, in dataset-level partitioning (folder: split3_dataset), each dataset was exclusively allocated to a separate fold.