FFA images from laser CNV in WT and Vldrl−/− mouse models were manually segmented to delineate vascular lesions using the NIS-Elements manual segmentation tool. The NIS.ai suite within the NIS-Elements software was trained on a dataset of manually segmented FFA images identifying vascular lesions, which served as the ground truth. To account for interoperator variability, seven of these images were manually segmented independently by three different operators. Each of these operator-marked segmentations was treated as a unique training image, resulting in 21 images for training from these seven original images. Additionally, two other images were manually segmented independently by two operators, yielding four images for training from this subset. Finally, 55 unique images, each manually segmented by a single operator, were included in the training dataset. Overall, this segmentation strategy resulted in a final training set of 70 images, incorporating interoperator variability into the ground truth to enhance the ability of the model to generalize. Objects smaller than 60 µm2 were not included to eliminate artifacts. Across these 70 images, 1796 vascular lesions were identified and manually segmented. In each FFA image, the number of vascular lesions, lesion area, and mean fluorescence intensity were computed.