Semi-Automated Glaucoma Prediction based on Retinal Blood Vessel Detection and Cup To Disc Ratio

Glaucoma is a serious eye condition that can result in permanent vision loss. The automated assessment of retinal fundus images stands as a pivotal future tool in the early identification and management of progressive ocular conditions such as Glaucoma. Machine learning models used in Glaucoma prediction are computationally complex. At times, these models demonstrate environmental unsustainability due to their carbon footprint, raising concerns about their eco-friendliness in the context of Glaucoma prediction. In this study, we have developed a fast and robust table-driven approach for novel semi-automated Glaucoma prediction pipeline using retinal blood vessels detection along with the cup to disc ratio values. Automated retinal blood vessel detection is based on a fusion of image processing and binary tree traversal techniques while CDR values are also considered resulting in an improved accuracy. The algorithm has been evaluated on a total of 74 images: 54 images are used for training (glaucoma- 38, normal-16) and 20 images (glaucoma- 16, normal- 4) are used for prediction. The Accuracy obtained by the proposed novel algorithm is 90 % with Sensitivity of 84.62% and F1 Score of 91.67%.