Purchase this article with an account.
J. Xu, H. Ishikawa, G. Wollstein, K. Sung, L. Kagemann, Jr., R. A. Bilonick, J. S. Kim, M. L. Gabriele, K. A. Townsend, J. S. Schuman; Automated Extraction of Optic Nerve Head Parameters From Stereoscopic Optic Nerve Head Photographs. Invest. Ophthalmol. Vis. Sci. 2007;48(13):3312.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To develop an automated system to define optic nerve head (ONH) and cup margins on stereo disc photographs and to evaluate its performance in comparison with human expert assessment.
Stereo disc photos were taken using Nidek 3Dx camera. Fully automated detection of the ONH margin was performed using a software program of our own design. A 3-dimensional (3D) model of the ONH was generated by detecting corresponding points on a pair of stereo images. A deformable model technique was applied on the 3D model to automatically extract the cup margin. Three glaucoma specialists manually defined the ONH and cup margins in a randomized masked fashion. The majority opinion of the expert defined margins was compared with the automatically generated ONH measurements. Disc area, cup area, cup-to-disc (C/D) area ratio and C/D vertical ratio were computed based on both machine and expert defined margins. Areas under the receiver operating characteristic curve (AROC) were calculated using only healthy and glaucomatous eyes to test the glaucoma discriminating ability of each parameter.
1 eye of each of 63 consecutive patients were enrolled (24 healthy, 19 glaucoma suspect, and 20 glaucomatous eyes). All ONH measurements based on machine defined margins showed a high correlation with the expert defined margins (R2 0.78 for disc area, 0.52 for cup area, 0.21 for C/D area ratio, 0.29 for vertical C/D ratio, all p≤0.0001, Pearson correlation). No significant difference was found in glaucoma discriminating ability between machine and human defined margins (AROC 0.87 vs 0.85 for C/D area ratio, AROC 0.89 vs 0.86 for vertical C/D ratio, respectively, both p>0.6 for the comparison of AROCs between methods).
Fully automated ONH assessment of stereo disc photos performed as well as human experts in glaucoma discriminating performance and quantification of disc and cup area. The automated method provides an objective and quantitative option for ONH evaluation using widely available disc photographs.
This PDF is available to Subscribers Only