Computational analysis of longitudinal changes to optic nerve head structure in normal and glaucoma individuals

Abstract

Purpose : To characterize longitudinal changes in optic nerve head (ONH) structure and identify features associated with the development of primary open angle glaucoma (POAG).

Methods : Clinical measurements and stereo fundus photos were collected at regular intervals from a large cohort of participants (n = 1636) as part of the Ocular Hypertension Treatment Study. A subset of this data consisting of participant demographics (age, sex, ethnicity), standard clinical measurements (intraocular pressure, cup-to-disc ratio), stereo images, and disease outcomes (conversion to POAG) collected over a ten-year period was identified. ONH structural features were computed for each participant at each stereo imaging time point using our previously published methods for quantifying ONH structure from stereo. Longitudinal changes in the clinical and structural features were characterized for both normal and POAG participants to identify changes associated with the disease. Predictive models were also constructed using only data collected prior to conversion to POAG to evaluate the ability of these features to perform early disease detection.

Results : The structural measurements computed from stereo were significantly associated (p < 0.05) with disease status at the time of imaging. The features were also significantly (p < 0.05) associated with standard clinical measurements. Models predicting POAG achieved an area under ROC of 0.752.

Conclusions : Computational analysis of longitudinal stereo fundus images collected over the course of ten years revealed disease associated changes to the ONH. Augmenting standard clinical measurements with image-derived structural measurements also aided predicting conversion to POAG. Future work will incorporate available genotype data to characterize the influence of genetics as well as disease on longitudinal changes to the ONH.

This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.