Abstract

Study Design-Cross sectional, prospective
Purpose-Detection of retinal abnormalities by Computer aided diagnostic system. Classification of Age Related Macular Disease and Diabetic retinopathy (NPDR and PDR and their stages). Comparison between accuracy with Random Forest classification and Naïve Bayes Classifier
Methods-A total of 405 images formed composite database comprising of images from publicly available databases with different symptoms of DR and ARMD. Next retinal input Image of patients pre-processed. K means clustering done. Bright and Dark clusters were extracted. After elimination of Optic Disc and blood vessels tree, appearance-based and region based features were extracted. Lastly, classification by machine learning algorithm(Random Forest Classification) and performance analysis done
Results-Random Forest classifier showed higher accuracy and our purpose was met
Conclusion-Random Forest Classifier results were promising for classifying retinal images of DR and ARMD