Purchase this article with an account.
Mengyu Wang, Tobias Elze, Thomas Peschel, Louis R Pasquale, Lucy Q. Shen, Kerstin Wirkner, Joachim Thiery, Markus Loeffler, Christoph Engel, Toralf Kirsten, Franziska G Rauscher; Improving Circumpapillary Retinal Nerve Fiber Layer Thickness (RNFLT) Norms with Retinal Anatomy. Invest. Ophthalmol. Vis. Sci. 2022;63(7):3357 – F0166.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Current circumpapillary RNFLT norms only adjust for age, gender and ocular magnification. We aim to improve the circumpillary RNFLT norms with retina anatomy.
From the population-based Leipzig Research Center for Civilization Diseases (LIFE) adult study, we include data with: (a) no retinal pathological findings on fundus images; (b) image quality ≥ 20 dB, average number of B-scans ≥ 50, and no more than 5% missing or unreliable RNFLT measurements among the 768 locations of Spectralis optical coherence tomography (OCT) scans. The retinal anatomy was represented by the 2D inner limiting membrane (ILM) contour map and scanning laser ophthalmoscopy (SLO) fundus photo at the optic nerve head. Feature exaction was performed by principal component analysis followed by uniform manifold approximation and projection. Linear regression was applied to associate the ILM and SLO features with pointwise RNFLTs adjusting for age, gender and ocular magnification. One eye per subject was randomly selected for regression analyses. Model selection based on Bayesian information criteria (BIC) was used to remove redundant features from the linear models. Adjusted r-squared and BIC values penalized for linear model complexity were used to measure the linear models’ accuracies.
7,285 eyes from 7,285 subjects (56.1 ± 12.2 years; 53.3% are female) were included. Figure 1 (a) and (b) shows the R2 improvement by using ILM and SLO features on top of age, gender and ocular magnification, respectively. Additionally using ILM and SLO features strongly (BIC improvement ≥ 6) improved RNFLT norms at 85.4% and 96.5% of the 768 locations with average R2 improvement of 0.026 (max: 0.051) and 0.029 (max: 0.065) compared with average R2 of 0.042 (max: 0.082) using age, gender and ocular magnification alone, respectively. Combining ILM and SLO features (Figure 2) provided more accurate personal norms than using SLO and ILM separately, which was supported by RNFLT norm improvement (BIC improvement ≥ 6) at 79.9% and 93.2% of the 768 locations with average R2 improvement of 0.017 (max: 0.032) and 0.020 (max: 0.049).
Adjusting for retinal anatomy represented by ILM and SLO imaging features substantially improved personal circumpapillary RNFLT norms. Our new norms may improve the diagnostic accuracy of OCT for glaucoma patients, although further validation is needed in patient population.
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.
This PDF is available to Subscribers Only