Towards an automated approach to the detection of retinal abnormalities

Detection of Diabetic Retinopathy and Age related Macular Degeneration is a challenge for the ophthalmologists as the abnormalities are merely visible at the early stage. Early detection of these diseases can prevent permanent vision loss. Handling a large amount of retinal images and detection of abnormalities due to these diseases is laborious as well as time consuming. In this research work, an algorithm is developed for identifying the abnormal objects in retina, if any with a machine learning technique using Naïve Bayes classification is proposed. A training set is generated with features of abnormalities present in retinal image and the type of disease the retina is suffering from. The Naïve Bayes classifier helps to predict the disease for each retinal image after gathering the knowledge from training set. The correctness of prediction is calculated to measure the efficiency of the classifier. The system achieves 97.014% of accuracy on an average.