Abstract
Purpose :
To test 3 methods of applying the mathematical model proposed by Jansonius et. al. [1] to optical coherence tomography (OCT) retinal nerve fiber layer bundle (RNFL) probability/deviation maps from eyes with glaucomatous damage.
Methods :
Jansonius et al.’s mathematical model describes the trajectories of RNF bundles as a function of the origin (angular position) around the disc.[1] Individual variance in the degree of trajectory curvature for a given angular position around the disc is explained by the parameter beta in the model. A custom program was used to compute and superimpose the model’s trajectories on RNFL probability and thickness maps for 18 eyes with well-defined arcuate damage as seen in the abnormal region of the RNFL probability map (Fig. 1).[2] For Method 1, trajectories were fitted to this region by manually varying beta and the associated region (AR) (red and blue vertical lines in Fig. 1) on the circle scan (Method 1, Fig.1A). For Method 2, a linear regression was performed to determine a relationship between beta and fovea-to-disc distance. Then, beta was automatically determined for each eye by the regression relationship, with AR manually adjusted for best fit. For Method 3, the average beta value [1] was used for all eyes, with AR manually adjusted for best fit. Trajectory fit was then examined across all eyes for average beta with AR set as the portion of the circle scan with thickness less than 5% of normal (Fig. 2).
Results :
Using regression-based beta values (Method 2) did not out-perform using the average beta value (Method 3). While manually adjusting the beta (Method 1) gave the best agreement between the model and the data (as expected), the differences between its performance and Method 3 were relatively small (Fig. 1). Further, testing Method 3 by fixing the AR based upon the circumpapillary RNFL thickness (Fig. 2), produced reasonably good fits.
Conclusions :
While manual determination of a separate beta for each eye performs best, implementing this method clinically is not practical. Fortunately, simply using the same beta value for all eyes did well enough for clinical use. Application of this model should improve both qualitative and quantitative evaluation of glaucomatous damage on OCT RNFL probability plots.[2] 1. Jansonius et al Exp Eye Res 2009, 2012; 2. Hood et al, JoG 2020.
This is a 2021 ARVO Annual Meeting abstract.