A hemidisc comparison of OCT minimum rim width (MRW) and circumpapillary retinal nerve fiber layer thickness (cRNFL) measures in eyes with or suspected glaucoma.

Abstract

Purpose : A hemidisc analysis was performed to better understand the differences between two optical coherence tomography (OCT) measures of glaucomatous damage, Bruch’s Membrane opening-minimum rim width (MRW) and circumpapillary retinal nerve fiber layer (cRNFL) thickness.

Methods : OCT circle scans were obtained for 2 groups of eyes diagnosed as glaucoma or glaucoma suspects by referring specialist: an early glaucoma group (EG) of 88 eyes with 24-2 MD better than –6.0 dB,[1] and a broadly inclusive group (BG) of 188 eyes with 24-2 MD from -0.15 to -27.0 dB.[2] Based upon the commercial summary metric (Fig. 1, arrows), each superior (inferior) hemidisc was classified as abnormal (Abn) if either of the two superior (inferior) sectors, TS & NS (TI & NI) was yellow or red, and as normal (Nor) if both were green. Fig. 1 shows an example of agreement and disagreement (black and red rectangles). To determine the false-positives (FP) and false negatives (FN), 3 authors performed a post-hoc analysis of additional structural (RNFL and RCG thickness/probability maps) and functional (24-2/10-2 visual fields) information for the EG group.

Results : The MRW and cRNFL measures agreed in 81.9% (BG) and 73.9% (EG) of the hemidiscs (Table). In both groups an Abn-MRW/Nor-cRNFL disagreement was as common as an Abn-cRNFL/Nor-MRW (EG: 22 vs. 24; BG: 36 vs. 32 hemidiscs). Of the 46 EG hemidisc disagreements, the number of FP plus FN (i.e. “mistakes”) for MRW, 27, was higher than that for cRNFL, 19 (Fisher’s Exact Test: p=.15). Common causes for “mistakes” involved: a significance level on the border of Nor and Abn (FP: 5 cRNFL, 4 MRW; FN: 6 cRNFL, 9 MRW); a local defect that was missed (FN: 7 cRNFL, 9 MRW); an aberrant blood vessel (BV) location (4 cRNFL FP) or BV inclusion in measure (3 MRW FP); thicker than average cRNFL (4 FN) or MRW (3 FN) when originally healthy. [note: a hemidisc could have more than one.]

Conclusions : Both MRW and cRNFL measures can detect damage missed by the other. It should be possible to improve these measures by accounting for BV location; correcting segmentation errors; and combining with information from RGC/RNFL probability maps; and perhaps by comparing actual probability levels, as opposed to set levels (e.g. yellow ≤5%). 1. Hood, Tsamis et al. IOVS, 2019; 2. Wu, Vianna et al. BJO, 2019.

This is a 2020 ARVO Annual Meeting abstract.