Rapid test for detecting red-green color vision deficiencies using a neural network-assisted color-naming task

This study presents and evaluates a rapid digital color vision deficiency test using a color-naming task supported by a neural network model. Thirty-three normal trichromats, 12 dichromats, and 11 anomalous trichromats named 182 natural-scene colors using 11 basic terms. A neural network classified individuals for screening, classification, and diagnosis. Sensitivity and specificity reached 97% and 99% for screening analyzing the full and a 20-color subset, respectively. For classification and diagnosis, accuracy was slightly lower. Results show that color naming can efficiently detect deficiencies, suggesting that a fast and accurate screening test using only 20 colors can be completed in under 2 min.