Abstract
Purpose: :
To determine the accuracy of an automatic method for identifying the border and center of the optic disc in peripapillary images taken by the Cirrus optical coherence tomography instrument (Carl Zeiss Meditec, Dublin, CA). Furthermore, to determine the improvement in repeatability achieved using this technique.
Methods: :
Seventeen subjects with no known retinal pathology were imaged in-house. Three OCT volumes (200x200 axial scans) were acquired per eye per session, and a total of three sessions were performed on consecutive days. For each image, the border and center of the optic nerve were identified using an automated graph-based method. This identification determined the center point for a TSNIT analysis. The offset between two scans was determined by comparing the centers. To evaluate accuracy of the center-finding technique, the offset was compared to the offset determined from a blood vessel registration technique. Analysis of variance (ANOVA) was used to determine repeatability standard deviation (SD) of the average RNFL thickness measured in a circle centered on the optic disc. Repeatability SD was compared with and without registration between images.
Results: :
Center finding using the graph-based technique was accurate compared to blood vessel registration to within ± 200 µm. Repeatability SD with no registration technique (operator centering only) was 1.6 µm. Registering scans by lining up the centers of their optic discs improved repeatability SD to 1.4 µm.
Conclusions: :
A graph-based method of determining the boundary of the optic nerve head provides accurate identification of the center of the optic nerve head. Repeatability of Cirrus HD-OCT RNFLT measurements is excellent, and improves with registration of images.
Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • imaging/image analysis: clinical • image processing