Segmentation of Hard Exudates in Retinal Fundus Images using Deep Convolutional Neural Network

  • Manoj Kumar Behera, Rutuparnna Mishra, Anshit Ransingh, S. Chakravarty

Abstract

Diabetes has become a global menace for the cause of several problems in the patient like kidney failure, cardiovascular problems and many more. Blind ness is one of them, which occurs due to uncontrolled blood sugar level in the body for a longer period of time and starts swelling in the blood vessels of retina. This eventually damages the retina and leads diabetic retinopathy (DR) for the patients those are suffering from diabetes mellitus for a long period of time (20 years). The ratio of blindness due to DR is very high in all over the world. Best way to deal with DR is, detecting and curing this at its earlier state and this only can be possible by conducting regular interval check-ups for the patients those are suffering from diabetes for a long period of time. If DR is detected at its earlier state it can be treated and cured. But due to lack of adequate number of Ophthalmologists and ignorance it generally leads to blindness. In this paper, to perform the test of DR in a more convenient and cost effective way, An Automatic model is developed by using convolutional neural network(CNN) and image processing technique, which can predict a patient is suffering from DR or not without any interference of doctors or Ophthalmologists . This model will analyse the input retinal image (captured by fundus camera) of the patient and find out the hard exudates present in it. Hard exudates are tiny white lipid deposits on the outer layer of the retina with a distinctive margin. The recommended system is very beneficial in the field of ophthalmology for detection of diabetic retinopathy, which archives a Prediction accuracy of 96.7%.

Published
2020-05-01
How to Cite
Manoj Kumar Behera, Rutuparnna Mishra, Anshit Ransingh, S. Chakravarty. (2020). Segmentation of Hard Exudates in Retinal Fundus Images using Deep Convolutional Neural Network. International Journal of Advanced Science and Technology, 29(06), 5192 - 5199. Retrieved from https://sersc.org/journals/index.php/IJAST/article/view/19577