Abstract
Purpose:
The glaucoma analysis algorithm on the Heidelberg Spectralis spectral-domain optical coherence tomograph (SD-OCT) identifies macular retinal thickness asymmetry based on a grid of blocks centered on the fovea. We hypothesized that creating thickness bins following an anatomic layout, such as the arcuate retinal nerve fiber layer (RNFL) distribution, would have a greater diagnostic utility than the current system. The purpose of this pilot prospective, observational clinical study was to test the efficacy of two anatomically derived retinal thickness maps in detecting glaucoma.
Methods:
We included 40 eyes from 22 participants, of which 10 (18 eyes) had a clinical diagnosis of glaucoma and 12 (22 eyes) had healthy eyes. SD-OCT macular volume scans were obtained from each participant. Glaucoma detection using an arcuate pattern map developed by the authors (Samuels et al. 2014) was compared to that of the Garway-Heath map (Garway-Heath et al, IOVS, 2002; 43:2213-20). Each superior and inferior hemifield map was segmented into sequentially smaller arcuate bins ranging in size from 2-8 bins. Mean thickness in each segment was calculated. The same Spectralis SD-OCT images were used to calculate mean thickness in 5 segments of the Garway-Heath map (excluding the nasal sector). Mean thickness of each segment was calculated for all normal eyes in both anatomical mapping patterns. The 5th percentile for each segment was defined as the lower limit of normal. An eye was considered glaucomatous if one bin or sector was lower than the 95% confidence intervals (CI). The sensitivity and specificity of detecting glaucomatous vs normal subjects using the arcuate bin pattern was compared to the Garway-Heath map.
Results:
Using the arcuate bin, the pattern with 5 superior and inferior segments showed the highest specificity (86.63%) and had a sensitivity of 72.2%. The Garway-Heath map exhibited a specificity of 86.63% with a sensitivity of 83.4%.
Conclusions:
Specificity was similar for both maps and the arcuate bin map had a lower sensitivity. In future studies, we plan to individualize the size and locations of the bins which may result in increased sensitivity. Refining the resolution and precision of anatomically derived maps may result in improved diagnostic accuracy and provides proof of concept for developing a more robust normative data set for arcuate bin analysis.