Benchmarked pterygium images for human and machine graders

In the absence of ground truth, scores from many graders are required to obtain good representation of a clinical grading. The internet enables quick feedback from the experts at the comfort of their home of office. In this study, we demonstrated the use of online form as a tool to get quick feedback from clinicians on clinical grading of pterygium images with various severities. The scores were analyzed using quartile analysis and the median was used to construct the benchmark scores for the images. This dataset was tested on assessing human grader and was later fitted with neural network to measure the performance of the machine learning algorithm.