Purchase this article with an account.
O. Tan, V. Chopra, R. Varma, D. Huang; Automated Disc Boundary Detection on Optical Coherence Tomography Optic Nerve Head Scans. Invest. Ophthalmol. Vis. Sci. 2009;50(13):356.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Accurate optic disc size determination is important in evaluation of glaucomatous optic neuropathy since neuroretinal rim and cupping are related to disc size. Our goal was to evaluate the repeatability of automated disc boundary detection using Fourier-domain optical coherence tomography (FD-OCT).
Prospectively enrolled participants in the Advanced Imaging for Glaucoma (www.AIGStudy.net) Study at the Doheny Eye Institute were scanned using the RTVue FD-OCT system (Optovue, Inc. Fremont, CA), which performs 26,000 axial scans per second and has 5-micron depth resolution. The disc was scanned using a 4-mm diameter Optic Nerve Head (ONH) pattern which comprises 12 radial scans and 6 circular scans. One randomly selected eye of each subject was analyzed and scans with low signal strength index (SSI<=38) were rejected. The OCT intensity images were exported with RTvue software (ver 3.5). Then they were processed by an automated computer image processing algorithm which we developed to detect the disc boundary based on an active contour model. The coefficient variance (CV) of 3 optic disc parameters: disc area (DA), horizontal diameter (HD), and vertical diameter (VD) were calculated to evaluate the reproducibility.
The database contained 148 subjects and 855 scans. Adequate SSI was found in 706 scan. 35 scans were removed because the active contour model did not convergence. Eyes with at least 2 valid scans was selected for CV calculation. Finally, 126 eyes (35 normal, 54 glaucoma suspect and pre-perimetric glaucoma, 37 perimetric glaucoma) were analyzed. The disc area was 2.11+/-0.59 mm2 (mean+/-standard deviation), HD was 1.62+/-0.24 mm and VD was 1.67+/- 0.24mm. The CV’s of all 3 parameters were < 5%.
This PDF is available to Subscribers Only