The National Eye Institute (NEI) Age-Related Eye Disease Study (AREDS) dataset is available at Database of Genotypes and Phenotypes (dbGap) phs000001.v3.p1.
The UK Biobank (UKB) dataset is accessible through the online showcase of UK Biobank resources.
We followed the same preprocessing protocol for DeepSeeNet available in their repository. Each color fundus image is cropped to a square which encompasses the Macula region, and then, the square image was resized to 224 * 224 pixels. Our code for preprocessing all the images in the dataset is available in main_resize_images.py file. As the images in AREDS dataset are from two batches 2010 and 2014, this code assumes that the images for batch 2010 (2014) are in the directories img_2010 (img_2014).
To make reproducing the experiments easier, we added three files with the name format FullVisitsGapXThr10.pkl (X = 4, 6, or 8) containing the images used for each subject in experiments with the pairs with 4 (2), 6 (3), and 8 (4) months (years) time gaps. They are used in the line 176 of the LongitudinalAMDNet/Data/AMDDataAREDS.py file to partition the dataset into the train, validation, and test sets. The names of images used in each experiment can be found by running the code until line 164 of the same file.