Development of machine learning algorithm based models for automatic detection of diabetic retinopathy

Abstract

Digital images are easily acquirable and can be stored, processed and analyzed according to our need. Even in medical field, the digital images like x-ray, CT, MRI play an important role to analyze the patient s problem. Image processing is used in many applications like pattern recognition, natural language processing, character recognition and robotics. In medical field, image processing is used to develop automated systems for finding out the presence of symptoms of a disease and to identify the severity of the disease. The automated systems reduce the cost, save our time and also help doctors in their analysis of diseases. For example we can easily identify the cancerous cells with the help of image processing techniques. The image processing technique can be used to analyze the retinal images for the presence of different retinal diseases. There are few retinal diseases which lead to the loss of vision if not detected at its early stages. Diabetic Retinopathy (DR) is one of the retinal diseases which lead to blindness if not detected at its initial stage. Also the symptoms of DR are visible only in its last stage. The people who have diabetes for more than ten years are more vulnerable to DR and a survey states that more than 50% of diabetic patients will have the chance of being affected by DR. But the number of ophthalmologists is not in a drastic increasing level. By the end of 2030, nearly 300 millions of people will be affected by this disease. So an automated system should be developed to detect the presence of DR in its initial stage. newline