Grading of retinal nerve fiber layer with a photographic reference set

Abstract

Purpose: We developed a new quantitative approach for the evaluation of diffuse atrophy of the retinal nerve fiber layer.

Methods: A simultaneous, visually supported grading system was designed, which consisted of a set of 25 reference photographs, numbered from 25 (broad, clearly striated nerve fiber bundles) to 1 (no nerve fibers visible). We prospectively evaluated this method by matching 60 retinal nerve fiber layer photographs of patients with glaucoma or ocular hypertension and normal subjects to the reference photographs twice by three observers with varying experience levels in evaluating retinal nerve fiber layer

Results: Intraobserver and interobserver reliability, evaluated by the intraclass correlation coefficient statistics, was excellent (intraclass correlation coefficient > 0.8). Intraclass correlation coefficients within observers were from 0.89 to 0.98 (lower limits 905% confidence interval, 0.84 to 0.97), and intraclass correlation coefficients between observers were from 0.81 to 0.91 (lower limits 95% confidence interval, 0.50 to 0.87). Clinical validity, in which the scores for upper and lower halves of photographs were compared with the mean deviation of the Humphrey 30-2 visual field program by using the Spearman correlation coefficient, was substantial (r = -.68, P < .01 for upper photo score and lower field mean deviation, r = -.53, P < .01 for lower photo score and upper field mean deviation).

Conclusions: By using a reference set of of photographs of the retinal nerve fiber layer, we defined a method to derive a quantitative measurement of retinal nerve fiber layer with good reliability and to extend evaluation of retinal nerve fiber layer photographs to nonspecialists.