Machine Learning based Segmentation and Classification Algorithms for Glaucoma Detection

Glaucoma is one of the leading causes of blindness due to neuro-degeneration. It is extremely difficult to detect at early stages using conventional methods. Moreover, these methods rely on the expertise of ophthalmologists that consume a lot of time and are prone to human errors. Since glaucoma damages retina by affecting the optic nerve head with the increased intraocular pressure that can result in permanent vision loss, it is required to perform segmentation of optic disc while detecting glaucoma which is one of the challenges due to its extremely small boundary as well as blockage of blood vessels. There is a need to develop an automated system for performing tasks of segmentation of optic disc along with classification of healthy images from glaucomatous ones to overcome the existing challenges. A systematic analysis has been conducted in this regard from various publications. It is found that techniques such as random walk, contour optimization, histogram equalization etc. have been utilized for detection of glaucoma although authors achieved efficient results in terms of highest accuracy of 99.5% and 99.93% respectively via deploying super pixel based red channel as well as selection of hough peak value for segmenting optic disc while diagnosing glaucoma at an early stage. Moreover, support vector classifier is found to be best suited for classification which gave excellent accuracy of 98.12% while focusing on removal of blood vessels. Therefore, results acquired from these techniques help in identifying challenges ranging from segmentation to classification and contribute in deceleration of glaucoma.