Area and volume analysis of the superior and inferior retinal nerve fiber layer (rNFL) bundles from SD-OCT thickness maps: correlation with thickness of the corresponding rNFL and ganglion cell-inner plexiform layer (GCIPL) sectors

Abstract

Purpose: Peripapillary rNFL and macular GCIPL thickness analyses are powerful diagnostic tools to evaluate early glaucoma. We have proposed a novel method to detect rNFL damage based on area and volume of superior and inferior rNFL bundles. Here we show the correlation between standard sector analyses and new rNFL bundle analysis.

Methods: Previously published SD-OCT data of 55 glaucoma patients (26 early, 11 moderate, & 14 severe) were provided by Budenz and colleagues (PMID: 21917932). Raw thickness maps of subject optic disc OCT scans were provided by Carl Zeiss Meditec. The area and volume of superior (S), inferior (I), and summed superior-inferior (S+I) rNFL bundles were measured as described separately (Burke et al, ARVO 2015). These values were compared to corresponding rNFL and GCIPL thickness (i.e. respective quadrant for rNFL and hemi-field for GCIPL). The correlation between bundle area and volume and the corresponding rNFL or GCIPL thickness was calculated using the Spearman (r) coefficient. Paired correlation coefficients were used to determine if bundle analysis correlates better with rNFL vs. GCIPL thickness.

Results: The S, I, and S+I bundles positively correlate with corresponding rNFL sectors for area (r = 0.879, 0.811, and 0.855) and volume (r = 0.883, 0.799, and 0.858) (all p<0.0001). The S, I, and S+I bundles also positively correlate with the corresponding GCIPL sectors for area (r = 0.606, 0.386, and 0.546) and volume (r = 0.612, 0.379, and 0.544) (p<0.01 I bundle, otherwise p<0.0001). Bundle area and volume had greater correlation with corresponding rNFL thickness than GCIPL thickness (p<0.0001). S bundle volume and S quadrant thickness had the best correlation (r=0.883). Nevertheless, there is significant variability in bundle volume for a given quadrant thickness and vice-versa. For example, for a S quadrant thickness of 75µm, S bundle volume ranged between 40-230 x 10^-3 mm³.

Conclusions: rNFL bundle area and volume are highly correlated with corresponding rNFL and GCIPL thickness. Despite high correlation with current rNFL analysis, there is significant variability of bundle area or volume for a given quadrant thickness value, and vice-versa. Further studies are required to determine the variability and reproducibility of rNFL bundle analysis.