Abstract:
Diabetic Retinopathy (DR) is an eye mellitus that may lead to progressive loss of vision due to impairment of retinal blood vessels. The effects of DR gets pronounced due to late detection and treatment. Approximately, 80% of the individuals who had been diabetic for more than 10 years or so are likely to develop and suffer from DR. DR occurs due to high concentration of blood sugar levels known as ‘Glucose’ that damage retinal blood vessels due to swelling, leakage or stopping blood circulation through retinal blood vessels (BVs).
The rapid growth in population of Pakistan coupled with limited health facilities and scarcity of specialists demands proactive approach for detection of DR so that only suspected persons are referred to ophthalmologists for thorough examination. The scarcity of Ophthalmologist and work load on the available ones makes it difficult to screen every person suspected of DR. Our proposed approach will not only reduce work load on the limited available specialists but also ensures timely detection of the DR, hence, saving time and resources as per national needs.
Due to recent advancements in digital image processing (DIP) techniques, well established pattern recognition algorithms, spatial imaging transformations and state of the art artificial neural networks (ANN), digital retinal fundus images are one of the main sources available in Ophthalmology which could be used in detecting most of the eye related diseases including DR.
The proposed system will analyze digital fundus images with the help of pattern recognition algorithms and performs decision making through classifiers to find out the stage and severity of DR. The proposed technique is simple, user friendly and computationally inexpensive that will enable operators (assistants to ophthalmologist) to refer only potential individuals suspected of DR to reduce burden on available specialists.
