Identifying Optimal Visual Field Locations to Detect Glaucoma Using Divergence Tests (306439)*Monica Ahrens, The University of Iowa
Keywords: statistical distances, bootstrap
According to the World Health organization, glaucoma is the second leading cause of blindness in the world; but with early diagnosis, blindness due to glaucoma could be prevented. Currently, ophthalmologists use perimetry as a screening tool for glaucoma. In this study perimetry data gathered on both peripheral and central visions are used to determine locations that may show early signs of glaucoma in the visual field. To measure differences between glaucoma and normal controls’ perimetry measurements, three statistical distances are used: Kullback-Liebler divergence, Hellinger distance and total variation distance. Each location’s distribution function is estimated using smoothing, while the integrals involved in the computation of these distances are estimated using finite elements discretization. Bootstrap resampling is used to test whether the distances between the normal and glaucoma are nonzero. The areas of greatest contrasts between the two groups, as determined by these distances, are the nasal and the temporal regions. The recommendation is that these areas may be given special attention during screening as they may hold key information in early glaucoma detection.