CORN-LCs

CORN-LCs comprises 4,224 corneal confocal microscopy (CCM) images with a resolution of 384×384 pixels, all captured using Heidelberg HRT-III systems. This dataset is used to generate an optimal grading model. These images were collected from 332 dry eye disease patients across six clinical centers in two countries. The image severity levels (ranging from 0 to 3) were annotated independently by two corneal specialists, each with over 15 years of experience. The grading was based on the established correlation between disease severity and Langerhans cell activation, combined with an assessment of ten key morphological parameters of Langerhans cells and the specialists' clinical expertise. The final ground truth labels were determined by consensus between the two specialists. The dataset distribution across the severity levels is as follows: Level 0 (1,064 images), Level 1 (1,797 images), Level 2 (885 images), and Level 3 (478 images).