Certain investigation on medical image processing using hybrid algorithm

Abstract

Diabetic patients are tremendously increasing worldwide and newlineDiabetic Mellitus (DM) is a chronic disease that may cause complications in newlineEye, Heart and Kidneys. Diabetic Retinopathy (DR) plays a vital role that newlinecauses vision loss in diabetic patients, if not treated in earlier stage. Normally, newlineclinicians will detect the different signs of DR from the retinal images taken newlinethrough fundus photography. Among the various stages of DR, newlineMicroaneurysms (MAs), Haemorrhages and Exudates are the stage newlinecharacterized by spots on the retina, also called as lesions. newlineNowadays, ophthalmologists face lot of challenges during the newlinediagnosis and classification of Diabetic retinopathy disease stages using newlinemachine learning methods. Also, machine learning methods using single newlinefeature extraction and selection technique leads to error and more time newlineconsuming task as stages differ only by the dimension of the blood spots and newlinetheir numbers which makes less impressive. Hence an enhanced detection and newlineclassification system using machine learning model becomes essential to help newlinethe ophthalmologist in detecting and classifying the stage of the diabetic newlineretinopathy patients. newlineIn this research, a novel based enhanced automated system is proposed newlineand implemented using image processing techniques. This system includes newlinethe following stages like pre-processing, segmentation, hybrid feature newlineextraction, feature selection and classification. The pre-processing stage newlineincludes normalization, Denoising and Contrast Enhancement. The pre newlineprocessed image is further segmented to separate the Optic Disc (OD) using newlineregion based segmentation and ABC segmentation technique. newline