Purchase this article with an account.
K. Lee, M. Niemeijer, M. K. Garvin, Y. H. Kwon, M. Sonka, M. D. Abràmoff; Automated Optic Disc Segmentation From 3D SD-OCT of the Optic Nerve Head (ONH). Invest. Ophthalmol. Vis. Sci. 2009;50(13):1102.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To validate an automated, supervised, optic disc cup and rim segmentation method in SD-OCT of the ONH.
24 ONH centered 3D SD-OCT volumes (200×200×1024 voxels, 6×6×2 mm3) were obtained from the 12 glaucoma subjects (mean age of 67±16 years) using a CirrusTM HD-OCT machine (Carl Zeiss Meditec, Inc., Dublin, CA). Two disc and cup manual tracings by two glaucoma specialists were obtained: one as the reference standard and the other as the second observer. The optic disc cup and neuroretinal rim were segmented by a trained classifier, in leave-one-out-fashion, with convex hull based post-processing.
The unsigned segmentation errors were 0.079±0.026 mm for the cup and 0.064±0.025 mm for the rim. Inter-observer variability as indicated by the unsigned border positioning difference between the second observer and the reference standard were 0.078±0.032 mm for the cup and 0.065±0.024 mm for the rim. The unsigned segmentation errors were not significantly different from the inter-observer variability for both the cup (p > 0.8) and rim (p > 0.9).
This preliminary study shows that the algorithm performs comparably to a glaucoma specialist on this small dataset. Even though other glaucoma metrics are potentially obtainable from 3D SD-OCT of the ONH, the advantage of cup and rim segmentation is that the results are immediately familiar to clinicians. If these results can be reproduced in a larger dataset, automated, objective, segmentation of the cup and rim has the potential to improve the management of patients with glaucoma.
Figure 1. Optic disc segmentation results. (a) Color fundus photograph overlapped with the reference standard. The optic disc cup is in red, and the neuroretinal rim is in green. (b) Cross-sectional image of the original OCT scan. (c) OCT projection image overlapped with our segmentation result. (d) Cross-sectional image of the original OCT scan overlapped with our segmentation result.
This PDF is available to Subscribers Only