Purchase this article with an account.
Ashley Sun, Emmanouil Tsamis, Melvi Del Valle Eguia, Jeffrey M Liebmann, Dana Blumberg, Lama A Al-Aswad, George A Cioffi, C Gustavo De Moraes, Donald C Hood; An evaluation of common methods for detecting progression of early glaucoma with optical coherence tomography. Invest. Ophthalmol. Vis. Sci. 2020;61(7):3928.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To evaluate optical coherence tomography (OCT) methods for detecting progression of early glaucoma using a clinically relevant reference standard.
Widefield swept-source OCT scans (Topcon) from 151 eyes with an average of 3.5 [range 2-5] scans within 6 months formed a short-term group. 49 were healthy controls (H), and 102 eyes were glaucomatous or glaucoma suspects. Of these 151 eyes, 104 (76 patients, 28 H eyes) had another OCT scan at least 1 year after their baseline scan (mean: 24.9±8.7 months); these scans formed the long-term group. Circumpapillary retinal nerve fiber layer (cRNFL) thicknesses from a derived 3.45-mm circle centered on the optic disc and retinal ganglion cell (RGC) thicknesses from a ±8° region around the fovea were obtained from every scan. Global cRNFL (G) and global RGC (Gmac) average thicknesses were calculated. Quantile regression was applied on the G and Gmac metrics, with the independent variable being the baseline values and dependent variable all the follow-up values in the short-term group. Thus, we determined the lower 95% limit of each metric. The long-term group values were then compared to those limits. Eyes that fell below the 95% limit were classified as “statistical progressors.” For a progression reference standard (P-RS), 4 experts evaluated all OCT and 24-2 and 10-2 VF information, including OCT reports with probability maps (Fig. 1).
24 (G) and 26 (Gmac) of the 76 patient eyes, and 2 (G) and 6 (Gmac) of the 28 H eyes, were classified as “statistical progressors” (Fig. 2). For the P-RS, 28 of the 76 eyes were identified as progressors, while none of the H eyes were progressors. Only 10 of these 28 eyes were identified as statistical progressors on both G and Gmac, while 4 of the 28 were missed by both metrics (Fig. 2), although the OCT maps showed clear progression (Fig. 1, arrows). Further, based on the P-RS, G and/or Gmac metrics falsely identified 13 eyes as statistical progressors. An analysis of the OCTs for these false positives (FP) showed segmentation errors and/or reduced contrast as the cause, rather than progression.
For detecting progression, the common metrics G and Gmac can lead to FPs due to scanning artifacts. Most importantly, G or Gmac metrics can miss eyes with clear signs of progression when assessed by a clinically relevant P-RS that includes RGC and RNFL probability maps.
This is a 2020 ARVO Annual Meeting abstract.
This PDF is available to Subscribers Only