Abstract
Purpose: :
To assess the relationship between error and the number of delineated radial B-scans in the manual delineation of RNFLT within SD-OCT images.
Methods: :
Ten normal non-human primate (NHP) eyes were imaged by SD-OCT (SpectralisTM, Heidelberg Engineering, GmbH). In each, a 15° radial scan containing either 40 or 48 B-scans (756 A-scans/B-scan, each B-scan the average of n=9 sweeps) was obtained, centered on the disk. RNFL boundaries within each B-scan were manually delineated using custom software, then RNFLT estimates were extracted at a peripapillary eccentricity of ~1120 µm (est. for the NHP eye based on a 1400 µm Spectralis designation for human eyes). The accuracy of RNFLT estimates was assessed as a function of delineated B-scan number both overall and for 8 individual regions (Figure) by computing an error term equal to the absolute difference between the estimate derived from the complete set of B-scans and the estimate derived from the averaged subset (6,8,12,etc).
Results: :
The relationship between percent error and number of delineated B-scans is depicted in the Figure. For overall thickness, delineating as few as 6 sections yielded a maximum of 6% error, which decreased to 3% for 12 delineated B-scans. Regional RNFLT estimates, however, required at least 36 total B-scans to be delineated to achieve less than 5% error.
Conclusions: :
Delineating twelve radial B-scan sections produces acceptable global estimates of peripapillary RNFLT in normal NHP eyes. However, within these data, delineation of at least 36 radial B-scans was required to precisely estimate regional RNFLT.
Keywords: nerve fiber layer • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • optic disc