Diagnostic Efficacy of RNFL Thickness Sectors for Glaucoma Detection

Abstract

 

Purpose

 

To compare the diagnostic ability of zones typically used in structure-function mapping against various standard sectors of peripapillary retinal nerve fiber layer (RNFL) measurements in glaucoma.

 

Methods

 

SD-OCT data from 140 Glaucoma (MD Mean±SD: -8.4±7.4 dB; 74 mild, 30 moderate, 36 severe by Hodapp-Anderson-Parrish criteria) and 191 age-matched Normal eyes (331 patients total) combined from three prior studies were analyzed post-hoc to generate peripapillary RNFL thickness maps. RNFL TSNIT profiles were extracted and used to generate various sectors, including Average RNFL thickness, quadrants, and clock hours. In addition, sectors proposed by Garway-Heath (GH) for mapping structure to function [1], equivalent sectors used in scanning laser polarimetry (GDx), and mean RNFL thicknesses within all possible combinations of Superior (1-180°) and Inferior (181-360°) sectors were computed. Area under ROC curve (AUC) values and sensitivities at fixed 95% specificity were then calculated from each sector. Pairwise comparisons between AUCs were computed using the standard method of Delong et al (significance at P < 0.05). For comparison purposes, a set of Optimal Superior and Inferior sectors were defined by averaging the endpoints of sectors with AUCs greater than 99.95% of the maximum AUC.

 

Results

 

The 99.95% Optimal Superior (41-149°) and Inferior (262-306°) sectors had AUC values of 0.924 and 0.941, respectively, and were significantly better than any other individual sectors in the corresponding hemisphere, except for Average RNFL thickness (0.929), Superior Quadrant (0.922), Inferior Quadrant (0.933), Temporal-Inferior GH sector (0.936), Superior GDx sector (0.919), and Inferior GDx sector (0.929). The sensitivities of those sectors at 95% specificity ranged from 80.0% (Average RNFL thickness) to 75.7% (Temporal-Inferior GH). The AUCs of the four superior/inferior GH sectors (Zones 1-4) were also not significantly worse than any of the 6 standard superior/inferior clock hours (11-1, 5-7 o’clock).

 

Conclusions

 

Existing Cirrus sector definitions and Garway-Heath zone-derived RNFL thickness sectors both show good diagnostic ability for glaucoma detection. One GH sector, Temporal-Inferior (Zone 2), is comparable to the best possible RNFL sector (Optimal Inferior). [1] Garway-Heath, DF et al., 2000. Ophthalmology, 107(10).

 

 

Table 1: AUCs and sensitivities.

 

 

Fig 1: RNFL sectors (right eye orientation) .

 

Keywords: 610 nerve fiber layer • 550 imaging/image analysis: clinical • 627 optic disc