Diagnostic accuracy of inner retinal layer thickness and reflectance intensity measurements assessed in posterior pole OCT scans of healthy and glaucomatous eyes

Abstract

Purpose : To determine whether quantifying reflectance intensity in addition to layer thickness improves the ability to discriminate glaucomatous (GL) from healthy control (HC) eyes.

Methods : Posterior pole OCT scans (Spectralis, Heidelberg Engineering GmbH) were obtained in 188 GL and 360 HC eyes (61 B-scans, 768 A-lines each, spanning a 30°× 25° area). GL eyes were defined as having 24-2 visual field (VF) pattern standard deviation or glaucoma hemifield test outside normal limits on 2 consecutive tests. Automated segmentations within each B-scan were manually corrected when necessary. Mean thickness and intensity values for each A-line were measured for nerve fiber layer (NFL), ganglion cell layer (GCL), and inner plexiform layer (IPL). These intensity values were also normalized using the average value of the 7-pixels corresponding to the retinal pigmented epithelial (RPE) layer to generate axially (vertically) normalized (AN) intensity. Values were age-adjusted using the HC eyes. The average thickness, intensity and AN intensity values were calculated and used to create receiver operating characteristic (ROC) curves. Logistic regression models were used to assess whether intensity added additional diagnostic power.

Results : The median VF mean deviation in the GL eyes was -4.37 dB. The highest univariate areas under the ROC curve (AUROC) were for NFL thickness (NFLT: 0.916) and intensity (0.903, Table 1). The AUROC for GCL thickness (GCLT) was significantly smaller (0.882) than for NFLT (p=0.04), comparable to NFL intensity (p=0.24), but higher than all other parameters (all p<0.001). NFL intensity (p<0.001), NFLT (p<0.001) and GCLT (p=0.01) were significant predictors of the probability of having glaucoma. Addition of NFL intensity to thickness significantly improved this prediction (p<0.001). A model also including GCLT had higher AUROC (0.958) than any other combination (all p < 0.02, Figure 1).

Conclusions : In this group of mostly early GL eyes, the best diagnostic predictors were NFLT, NFL intensity, and GCLT. Current axial normalization of intensity values to RPE reflectance did not enhance diagnostic accuracy. Macular NFL reflectance intensity provides diagnostic information beyond thickness values alone and may have pathophysiological significance.

This is a 2021 ARVO Annual Meeting abstract.