An improved ensemble machine learning approach for diagnosis of diabetic retinopathy in retinal fundus images

Abstract

The main cause of growing numbers of blindness in diabetic patients is proven to be Diabetic Retinopathy (DR). The problem in disease is it does not give symptoms till it reaches to the damage stage. Detection of diabetic retinopathy disease in its early stages can prevent human from blindness. An identification of landmark features present in the fundus images has to accurately find the features from the optic disc of fundus images. The timely identification of DR grazes like hemorrhages, optic disc, exudates, cotton wools and microaneurysms supresses the rate of growth of this disease, and supports doctors in treating it. But, observing, and identifying DR Graces are exhausting, recurring and may lead to wrong diagnosis. newline newlineInformation about level of seriousness of DR in patients who are supposed to have timely checks in a recurring manner and those who are not in advanced stages is to be given to the doctors for controlling and treating DR. But detection of such a tiny blood vessel is very complex, time consuming and may be error prone which need expert in ophthalmologist. Sadly, situation in economically poor societies affording an expert with required infrastructure may not be an easy task, and also, the recommended tests of the eye fundus of a diabetic patient is too risky or not easy. There is a growing demand for alternatives to be explored thru medical and technological research to address this gap and make this process an affordable one for even a remote society in an economically backward country. newline newlineEarly identification of DR is a common discussion among the research fraternity, existing researches used various Artificial Intelligence (AI) techniques for screening and diagnosing the DR earlier to prevent the diabetic patients from blindness which was determined based on the level of DR severity. However, the existing models were inefficient as the consumed time and the premature convergence was resulted with the drawback for the real world approaches.