Automatic grading of ocular hyperaemia using image processing techniques

Abstract

The human eye is affected by a number of high-prevalence pathologies, such as Dry Eye Syndrome or allergic conjunctivitis. One of the symptoms that these health problems have in common is the occurrence of hyperaemia in the bulbar conjunctiva, as a consequence of blood vessels getting clogged. The blood is trapped in the affected area and some visible signs, such an increase in the redness of the area, appear. This work proposes an automatic methodology for bulbar hyperaemia grading based on image processing and machine learning techniques. The methodology receives a video as input, chooses the best frame of the sequence, isolates the conjunctiva, computes several image features and, finally, transforms these features to the ranges that optometrists use to evaluate the parameter. Moreover, several tests have been conducted in order to analyse how the methodology reacts to unfavourable situations. The goal was to cover some common issues that assisted diagnosis methodologies have to face in real-world environments. The proposed methodology achieves a significant reduction of the time that the specialists have to invest in the evaluation. Thus, it has a direct repercussion on reaching a fast diagnosis. Moreover, it removes the inherent subjectivity of the manual process and ensures its repeatability. As a consequence, the experts can gain insight in the parameters that influence hyperaemia evaluation.