Purpose: To compare the capability of the deviation map algorithm and peripapillary retinal nerve fiber layer (pRNFL) thickness measurements, both determined using Cirrus spectral domain optical coherence tomography (Carl Zeiss Meditec Inc., Dublin, CA), in discrimination between healthy and glaucomatous eyes.
Methods: This cross-sectional comparative prospective study included 54 patients with glaucoma with localized visual field (VF) defects confined to 1 hemifield and 54 normal controls. Mean pRNFL thickness corresponding to localized VF defect was calculated. A deviation Score (DS) was assigned by customized Image J software (Http://www.rsb.info.hih.gov/ij/18) calculating the number of abnormal superpixels in the deviation maps corresponding to localized VF defect. Correlations between DS and mean pRNFL thickness, and VF parameters were obtained. The areas under receiver operating characteristic curves of the DS and mean pRNFL thickness used for discriminating between controls and glaucomatous eyes were compared.
Results: DS and the average pRNFL thickness measurements showed a strong correlation (R, -0.844, P<0.0001). There were also significant correlations between DS, mean deviation (MD), and pattern standard deviation (R, -0.678, 0.699, respectively). The areas under receiver operating characteristic curves of the mean pRNFL thickness (0.892) and DS (0.958) differed significantly (P=0.0051). The sensitivities at 80%, 90%, and 95% specificities were 94.4%, 90.6%, and 88.6%, respectively, for DS and 74.5%, 58.8%, and 47.3%, for mean pRNFL thickness.
Conclusions: The deviation map algorithm showed a superior capability for detection of localized glaucomatous VF defects when compared with traditional pRNFL thickness measurements. There were statistically significant correlations between the deviation map algorithm and VF indices.