Support Vector Machine Based Method for Automatic Detection of Diabetic Eye Disease using Thermal Images

Abstract

Diabetic eye disease is one of the major problems worldwide. That can cause major impairment to the eyes, including a permanent loss of vision. Early detection of eye diseases increase the survival rate by successful treatment. The proposed methodology is to explore machine learning technique to detect diabetic diseased using thermography images of an eye and to introduce the effect of thermal variation of abnormality in the eye structure as a diagnosis imaging modality which are useful for ophthalmologists to do the clinical diagnosis. Thermal images are pre-processed, and then Gray Level Co-occurrence Matrix (GLCM) based texture features from gray images, statistical features from RGB and HSI images are extracted and classified using classifier with various combination of features. To detect diabetic diseased eye, here Support Vector Machine classifier is used for classification and their performance are compared. A 5-fold cross validation scheme is used to enhance the generalization capability of the proposed method. Experimental results obtained for various feature combinations gives maximum accuracy of 86. 22%, sensitivity of 94. 07% and specificity of 79. 17% using SVM classifier with five-fold validation.