Fundus Image Analysis for Detection of Papilledema

Abstract

Rapid development in the field of ophthalmology has increased the demand of computer aided classifiers for various eye diseases. Papilledema is an eye disease in which the optic disc of the eye is swelled due to increase in intracranial pressure. If it is not detected on time, it can lead to permanent loss of vision. In the past, although there were several case studies on papilledema from medical point of view but no significant work has been done on the automatic detection of papilledema. In this thesis, we have designed a system which will automatically detect papilledema from fundus images. Firstly, the images are pre-processed by going through optic disc detection and vessel segmentation. After pre-processing, a total of 26 different features are extracted which are categorized in 4 groups. The optimal features are selected and combined to form a feature matrix to be used to classify between normal images and images with papilledema using the support vector machine (SVM) classifier. Additionally, a detailed analysis is performed to find out significance of different groups of features based on our classification results. The proposed method is tested on 160 fundus images obtained from two different datasets namely STARE (a publicly available dataset) and our local AFIO dataset giving accuracy of 80.3% for AFIO, 97.8% for STARE and 87.6% for both datasets combined. Since there is not much work done on detecting papilledema using fundus images, this study would act as one of the pioneer works in this field.