Variation in Signal Quality as an Indicator of Retinal Nerve Fiber Layer (RNFL) Segmentation Error Using Optical Coherence Tomography (OCT)

Abstract

Purpose: : Commercial OCT systems use global signal quality indices to quantify scan quality, although signal quality can vary at different A-scan locations, which may contribute to RNFL segmentation errors (SE). The purpose of this study was to examine if signal quality variability within a scan coincides with SE.

Methods: : Peripapillary circular OCT scans (Fast RNFL scan protocol; Stratus OCT; Carl Zeiss Meditec, Dublin, CA) with global signal strength (SS) ≥ 6 were obtained from 11 healthy, 35 glaucoma suspect, and 74 glaucoma eyes. Scans were grouped based on SE, where SE was defined as obvious deviation of the segmented inner and/or outer RNFL borders from the subjectively perceived borders for consecutively 15% or cumulatively 20% of a given image. Quality Index (QI) values were computed for each A-scan, using software of our own design. Logistic mixed effects regression modeling was applied to evaluate SS, global mean and standard deviation (SD) of QI, and the probability of SE. Models were evaluated using odds ratios (OR) and the Akaike Information Criterion (AIC), where lower AIC meant smaller amount of information loss.

Results: : The average subject age was 63. The difference between mean of local QI in the SE region and no-SE region was -5.08 (95% Confidence Interval [-6.33--3.84]) (p<0.001). Using global mean QI and QI SD and the interaction between the two terms resulted in the model of best fit (AIC=197.3) for predicting SE. The model using SS, QI SD and their interaction had an AIC=212.2 and SS alone had an AIC =247.2. Categorical OR analysis showed that for low mean QI (≤16.4), the odds of SE increase rapidly as QI SD increases. For mid mean QI (16.4< mean QI ≤19.6), the odds of SE increase with mean QI, but not as rapidly. High mean QI (>19.6) shows little chance for SE, even with high QI SD.

Conclusions: : When combined with mean signal quality, the variation of signal quality between A-scans (QI SD) provides significant information about the quality of an OCT scan and can be used as a potential predictor of segmentation error.

Keywords: imaging/image analysis: clinical • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • nerve fiber layer