Abstract
Purpose :
To evaluate the focal loss of nerve fiber layer (NFL) reflectance for glaucoma diagnosis.
Methods :
Participants were scanned with 4.5x4.5-mm optic disc volumetric scan using a spectral-domain optical coherence tomography (OCT). In each axial line, the OCT reflectivity was summed in NFL, normalized by the average reflectivity in the photoreceptor and pigment epithelium complex, and converted to a logarithm scale. Areas occupied by large retinal vessels were excluded. The normalized NFL reflectance map was processed by an angular frequency filter, to remove high frequency noise, and to correct the directional reflectance bias due to the tilt of the retinal plane. The peripapillay area of the map was divided into 160 superpixels. Low-reflectance superpixels were identified as those with the normalized reflectance below the 5 percentile cutoff in the group of normal subjects (Figure 1). Focal reflectance loss was summed over the low-reflectance superpixels. The diagnostic accuracy of the NFL reflectance parameter was compared to overall NFL thickness using the area under received operative characteristic curve (AROC) and sensitivity at 99% specificity. The normative references for NFL thickness and normalized NFL reflectance were both adjusted for age and axial length.
Results :
Thirty-five normal, 65 open-angle glaucoma (30 pre-perimetirc, PPG, and 35 perimetric, PG) participants were enrolled in this study. The focal reflectance loss had significantly higher (p<0.025) diagnostic accuracy (AROC=0.925) than overall average NFL thickness (0.859). Using the 99% specificity cutoff, focal reflectance loss detected 97% of PG eyes and 53% of PPG eyes, which was significantly higher than overall NFL thickness (71% and 23%).
Conclusions :
Focal reflectance loss is a novel diagnostic parameter that improved glaucoma diagnostic accuracy compared to NFL thickness.
This is a 2020 ARVO Annual Meeting abstract.