Investigation of severity level of diabetic retinopathy using adaboost classifier algorithm

Abstract

Diabetics is a severe disease which affects the different parts of the human body. Due to variations of sugar level the blood vessels present in the retina get damaged. This disease is known as Diabetic Retinopathy (DR). Early identification of diabetic retinopathy will prevent vision loss. Exudates, Haemorrhages, Micro aneurysms are the primary symptoms of Diabetic Retinopathy. To diagnose the abnormalities, the ophthalmologists use fundus images which are captured after applyingdrugs inside the patient’s eye. These drugsmay causeinconvenience to the patient’s eyes. So we developed an automatedsystem to detect retinal lesions with help of non-dilated retinal fundus images. In our proposed system pre-processed images are segmented by new technology called region growing method. In this approach after pre-processing step, certain features are extracted by using Gabor Wavelet method. Adaptive boosting classifier is proposed to investigate the abnormality level of the disease. This result is compared with various segmentation techniques. This method achieved sensitivity of 100% and specificity of 98.8% with accuracy of 98.4%.