Automatic detection of glaucoma based on refined complete local binary pattern and random forest classification method

Abstract

Glaucoma is considered to be one of the main root causes of blindness. As it shows no symptoms, if not properly identified at the correct time would result in the loss of vision. This paper proposes a method for the Automatic Detection of Glaucoma based on Refined Complete Local Binary Pattern and Random Forest Classification Method(RCLBP-RFC), which identifies the presence or the absence of glaucoma in a patient at an early stage. The first step is use to convert a color image into gray scale image and the second step we use Neighborhood Fuzzy K Means Clustering to segment Optic Disc(OD) and Optic Cup(OC). In the third step Statistical Optimized and Restoration model is use to extract the enhanced images using the restoration technique. In the Fourth step we exploit Refined Complete Local Binary Patterns Extraction to extract the most relevant features and finally, Random Forest Classification methods are involved to classify the features as normal, abnormal or early detected glaucoma. The experiments show that our RCLBP-RFC method achieves state-of-the-art OD and OC segmentation result on DRIONS dataset. Experimental results indicates that the proposed method identifies the presence or absence of glaucoma more precisely than other existing methods in terms of computational time and complexity, and accuracy.