Identifying retinal nerve fiber layer defects using attenuation coefficients

Abstract

Purpose : Attenuation coefficient (AC), an index of retinal nerve fiber layer (RNFL) reflectance, is an optical property found to be lower in areas of RNFL damage. This exploratory study assesses the potential of using AC generated from spectral domain optical coherence tomography (SD-OCT) scans to detect RNFL abnormalities in patients with glaucoma (PWG).

Methods : Thirty PWG (67 years ± 8) with preexisting RNFL defects and 29 age-similar (68 years ± 9) gender-matched (17 males, 12 females) controls were selected for this study. One eye from each patient was analyzed and laterality was matched in controls. Using the internal limiting membrane (ILM) as a reference, AC maps were generated from 30^o H x 15^o W SD-OCT scans (Heidelberg Spectralis™) centered at the optic nerve, consisting of 145 vertical B-scans (30 µm spacing, ART 9). Mean AC values were computed for circular regions of interest (ROI) with a radius of 15 pixels. Fourteen ROIs were placed between the ONH and the temporal edge of the image in a vertical column to ensure that similar B-scans were included. The ROIs were carefully placed to avoid large blood vessels. Custom Matlab programs were developed for quantifying AC. Four slabs were generated, averaging between 0-52µm, 24-52µm, 24-36 µm, and 36-60 µm below the ILM. Three conditions of analysis were carried out with each slab; one using all 14 locations, and two others excluding either two or four of the central ROIs involving the papillomacular bundle. To normalize AC for each subject, the difference for each ROI from the mean of all ROIs was calculated. A 99^th percentile and maximum difference calculated from the control group were used as the criteria for normal range.

Results : The most effective combination excluded the central four ROIs and used the maximum difference criteria. With the 36-60 µm slab, zero controls and 22 patients were flagged. The other slabs flagged 5-16 patients. Assessing individual depths using this combination, 44 µm performed equally as the slab, while other depths flagged 2-16 patients. Using the 99^th percentile, 1-9 controls and 1-24 patients were flagged between the 12 combinations.

Conclusions : Attenuation coefficients can be used to flag areas of RNFL damage, however individual differences can be a challenge. A single depth can be as effective as a slab at identifying RNFL abnormality. These findings will need to be confirmed on a different sample.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.