Purchase this article with an account.
Henriet Springelkamp, Kyungmoo Lee, Wishal D. Ramdas, Johannes R. Vingerling, Caroline C. Klaver, Milan Sonka, Michael D. Abramoff, Nomdo M. Jansonius; Optimizing The Information Yield Of 3D OCT In Glaucoma. Invest. Ophthalmol. Vis. Sci. 2012;53(14):684.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The information yield of 3D OCT in glaucoma can, theoretically, be optimized by a detailed analysis of the entire scanned volume. However, technical limitations may hamper the use of the entire volume and a too detailed analysis may compromise test-retest variability. The aims of this study were to determine (1) which parts of OCT scans can be segmented completely in the majority of subjects and (2) the relationship between detail and test-retest variability.
907 6x6x1.68 mm OCT scans (Topcon) centered around the fovea and 912 scans centered around the optic nerve head (ONH) of the right eye of 924 participants of the Rotterdam Study were used; 10 were scanned twice. Using our standard 3D graph search approach, all volumes were segmented into 11 surfaces (10 layers). For each A-scan we determined, across subjects, how many times out of 907, all 11 surfaces could be segmented, and then determined the largest continuous region of where all surfaces could be segmented in 95% of subjects. For the ONH scans, we determined the radii of two circles centered around the scan center; the ONH had to fit in the smaller circle completely and the larger circle had to fit in the scan in 90% of subjects. The macula scan was divided in 4 (one per quadrant) and 68 (0.6x0.6 mm; following the 10-2 perimetry grid) blocks and we determined the test-retest variability (twice the standard deviation of the differences) of the mean retinal nerve fiber layer (RNFL) and retinal ganglion cell layer (RGCL) thicknesses within the blocks. The area between the circles of the ONH scan was divided in radially oriented segments of 1 and 1.5 clock hours.
96% of the macula scan could be segmented in at least 95% of the subjects; the segmentable part of the ONH scan was smaller (Fig. 1). The radii of the two ONH scan circles were 1.5 and 2.0 mm. Test-retest variability for RNFL (RGCL) ranged from 2 to 12 (3 to 5) and 2 to 12 (4 to 14) um for the 4 and 68 macula blocks, respectively; typical mean thicknesses were 30 um. For the ONH RNFL this ranged from 4 to 38 and from 2 to 32 um for the 1 and 1.5 clock hours segments, respectively. Typical mean thickness was 90 um.
Essentially the entire macula scan can be segmented and the block size can be reduced to 0.6x0.6 mm without sacrificing test-retest variability. Unsegmentable parts are more common around the ONH.
This PDF is available to Subscribers Only