M. Ranasinghe, H. K. Yashodhara
Abstract
Diabetes has become a global problem due to changing lifestyles, daily eating habits, level of stress encountered by people etc. According to statistics of the World Health Organization (WHO) in 2016, 8.5% of the adult population of the world is suffering from diabetes. Therefore, early detection of diabetes has become a global challenge. The iris of human eye depicts a picture of the health condition of the bearer. Iridology is a technique conceived decades back which focuses on study of iris patterns like color, texture & structure for diagnosis of various diseases. By analyzing the images of human iris, a medical imaging method was explored with computer vision for the identification of diabetes. Iris analysis of the human eye is conducted based on the pancreas, kidney and the spleen of the human body where the local datasets were collected using Digital Single Lens Reflex (DSLR) camera. A low-cost diabetes detection system was created focusing on the localization, segmentation, normalization and the system predict the severity of diabetes with 85% accuracy.
Keywords: Convolutional Neural Network, Diabetic, Feature extraction, Irido-diagnosis, Iris, Localization, Region of Interest, Segmentation
**View Full Paper in the Springer Source