Abstract
Purpose: :
To examine the impact of manual versus automated demarcation of retinal pigment epithelium (RPE) on optic nerve head and retinal nerve fiber layer thickness (RNFL) parameters, determined using Fourier-domain optical coherence tomography (FDOCT).
Methods: :
Normal, glaucoma suspect and glaucoma patients prospectively underwent FDOCT (RTVue, Optovue Inc, Fremont, CA) imaging. Inclusion criteria consisted of age between 40-85, refractive error between -8.00D and +4.00D, visual acuity ≥20/40, reliable standard automated perimetry and no prior intraocular surgery except uncomplicated cataract extraction. FDOCT Images obtained during eye movement, or images that were unfocused, poorly centered or had a scan score index of <40 were excluded. The ONH scan algorithm was used to obtain ONH and RNFL parameters. Software that automatically determines the position of RPE edges was used to define the optic disc margins. A single trained operator manually redefined the RPE edges to its best possible position. Analysis of variance, intraclass correlation coefficient (ICC), Lin’s concordance correlation coefficient (CCC) and coefficient of variation (COV) were used for the analyses.
Results: :
99 eyes of 50 patients (31 female, 19 male, age 68±10 years, 17 glaucoma, 4 normal, 29 glaucoma suspect) were enrolled. Automated and manual RPE demarcation made no significant (p>0.05) difference in measurements of disc area, rim volume, and RNFL in average and inferior-temporal (IT), superior-nasal (SN) and superior-temporal (ST) quadrants. There was a significant difference (p<0.05) between automated and manual RPE demarcation for the measurements of cup area, rim area, average C/D ratio and RNFL in the inferior-nasal (IN) quadrant. The ICC, CCC and COV values for ONH parameters were as follows: cup area 0.96, 0.88, 17.6%; cup volume 0.97, 0.91, 7.4%; average C/D ratio 0.92, 0.78, 18.1%; rim area 0.88, 0.69, 25.3%; and rim volume 0.83, 0.71, 42.6% respectively. The ICC, CCC and COV values for RNFL parameters were as follows: average 0.99, 0.98, 2.1%; IT quadrant 0.88, 0.79, 8.1%; IN quadrant 0.83, 0.67, 12.6%; SN quadrant 0.85, 0.83, 7.8%; and ST quadrant 0.93, 0.88, 7.8% respectively.
Conclusions: :
There was a high agreement, and a moderate to substantial concordance between automated and manual determination of RPE edges for ONH and RNFL parameters. More variation was observed in rim measurements and IN quadrant compared with other parameters. Automated RPE demarcation of FDOCT can be used reliably in clinical practice.