Eye Gaze Based Optic Dis

by Nilima Kulkarni

Eye movement includes the voluntary or involuntary movement of the eyes. Theeye tracker device is used for measuring the eye movements. Eye gaze dataanalytics can bring out information about observer's age, sex, education, experience, visual processing, cognitive processing, and many more.The thesis attempts to use the fact that the person's experience and expertise havean impact on their eye gaze pattern. The experts' eye gaze pattern while viewingthe medical images depicts the attentional behaviour of the individual, which hasbeen captured and further utilised for detecting the target region. The researchwork strives to automate and propose a complete eye gaze-based system that usesattentional theories inferred from eye gaze pattern.The designed system is evaluated for detecting optic disc in fundus retinal image.The chosen topic has been interesting as optic disc detection is a fundamental taskfor retinal image processing for classifying other fundus structures and is crucialfor the identification of eye-related diseases.Owing to the fact that the human visual perception has been less studied for opticdisc detection and an attempt to derive eye gaze based data analytics, the thesisfirst discovers how human perception works for optic disc detection using bottomupvisual attention model. The inference/ conclusions derived, paved way topropose an eye gaze-based optic disc detection (EGODD) system to detect opticdisc in fundus retinal images. The eye gaze data while the user is performingsimple target search task were collected from different users groups comprising ofexpert and nonexpert groups. Extensive data analysis has been carried to extracteye gaze features like fixation and using machine approach label regions in fundusretinal image. The segregated labelled data has been used to build a top downknowledge to bias the search map towards the target region.