Machine Learning Techniques for Severity Assessment of Pterygium

Pterygium is an external growth which is fibro vascular tissue that grows towards cornea of the eye. It is a challenging for health practitioners to identify and assesse the progression of the pterygium in the early stages. As a result, this paper describes the implementation of comprehensive system that effectively diagnoses, measure, and evaluate the onset and progression of the pterygium. HOG descriptors were tried with dichotomous classifier, SVM for the classification of healthy eye and pterygium. Here two ML algorithms BPNN (Back Propagation Neural Networks) and CNN (Convolution Neural Networks) were tried for grade the severity of the Pterygium in order to empower ophthalmologists to provide high-quality diagnosis and facilitate personalized health care in the near future. Of the two machine learning approaches in this study, CNN achieves 90-96% of accuracy compared to 66-77% accuracy in BPNN and other parameters specificity, sensitivity is also sufficient to evaluate the damaged caused by Pterygium.