Abstract
Purpose :
To evaluate the ability of a novel Macular Bayesian Deviation Map of macular ganglion cell inner plexiform thickness (mGCIPL) thickness to discriminate between healthy and glaucomatous eyes
Methods :
GCIPL thickness was calculated from Spectralis hotizontal posterior pole scans (61 b-scnas). The standard deviation map approach identifies pixels that are outside normal limits compared to healthy eyes. The Macular Bayesian Deviation Map uses information about the GCIPL thickness pattern in healthy and glaucoma eyes to generate statistical topographic maps to estimate the likelihood that a specific location shows glaucomatous mGCIPL thinning. A training group (29 healthy and 96 glaucoma eyes) was used to estimate the Bayesian Deviation Map parameters and an independent group (37 healthy and 285 glaucoma eyes) both from the Diagnostic Innovations in Glaucoma Study (DIGS) were used to evaluate the approach. Areas under the receiver operating characteristic (AUROC) curves, adjusted for age, axial length and both eyes in a subject, were calculated.
Results :
Compared to the group of healthy eyes, the glaucoma independent test group was significantly older (mean (95% CI) (70.4 (68.8, 72.1) and 59 (50.3, 67.5) years, P < 0.001), had worse mean deviation visual field (-6.3 (-7.2, -5.5) dB and -0.53 (-0.8, -0.2) dB P < 0.001) and similar axial length (24.1 (23.6, 24.6) mm vs 24.3 (24.1, 24.4) mm, P =0.22). The Bayesian Deviation Map performed significantly better (AUROC (95% CI) = 0.93 (0.89, 0.96), P=0.001) than the Standard Deviation Map approach (AUROC= 0.73 (0.56, 0.85)) after adjusting for age and axial length and both eyes in the model. At 95% specificity, sensitivity of the Macula Bayesian Deviation Map was 72%, compared to only 37% for the Macular Standard Deviation Map. When only eyes with early glaucoma (69 eyes) were included (MD > -2dB), the AUROCs significantly better using the Macula Bayesian compared to the Standard Deviation Map (0.85 (0.76, 0.93) vs 0.54 (0.33, 0.74), respectively, P=0.0007).
Conclusions :
The Macula Bayesian Deviation Map approach that utilizes information about the pattern and location of GCIPL thickness improves our ability to discriminate between healthy and glaucomatous eyes over the Standard Deviation Map approach.
This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.