Learning to predict diabetes from iris image analysis

Abstract

Iris image analysis for clinical diagnosis is one of the most efficient non invasive diagnosis methods for determining organs health status. Iridodiagnosis is an alternative branch of medical science which can be used for diagnostic purposes. To begin with we created database of eye images with clinical history of subject’s emphasis on diabetic subject (Type II) in pathological laboratory/Hospital. The entire process involves various modules such as image quality assessment, segmentation of iris, iris normalisation and clinical feature classification for clinical diagnosis. The artificial neural network is used for training and classification purpose. The entire process shows classification accuracy of 72–75% between diabetic and non-diabetic subjects.