Nuclear Cataract Database for Biomedical and Machine Learning Applications

Published: 4 April 2023| Version 2 | DOI: 10.17632/6wv33nbcvv.2
Israel Cruz-Vega, Hans Israel Morales


The dataset collected for multi-class nuclear cataract classification consists of 1437 labeled images. The images were obtained from a digital Topcon DC-3 slit-lamp camera with a resolution of 3264 x 2448 RGB; they were focused with a Topcon SL-D7 slit-lamp, a slit angle of 45º and 135º for each eye, a 6 mm x 0.5 mm aperture and maximum luminous intensity. The number of volunteers for the multi-class dataset was 68 women and 47 men, with at least three images per right and left eye. The expert ophthalmologists from the Conde Valenciana Institute from Mexico classified this dataset according to the LOCS III Protocol.



Instituto Nacional de Astrofisica Optica y Electronica