Purchase this article with an account.
Giovanni Gregori, Jianqing Li, Mengxi Shen, Yingying Shi, Hao Zhou, Qinqin Zhang, Yuxuan Cheng, Rita Laiginhas, Jeremy Liu, Ruikang K Wang, Philip J Rosenfeld; The Importance of Bruch’s Membrane Segmentation in Choriocapillaris Measurements Underlying Drusen in Eyes with Age-Related Macular Degeneration. Invest. Ophthalmol. Vis. Sci. 2022;63(7):2893 – F0046.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To investigate how the properties of choriocapillaris (CC) flow deficits (FDs) measurements in regions underlying drusen are affected by Bruch’s membrane (BM) segmentation.
Swept-source optical coherence tomography angiography (SS-OCTA) imaging (PLEX® Elite 9000; Carl Zeiss Meditec, Inc, Dublin, CA) of eyes with age-related macular degeneration (AMD) were acquired using a 6x6 mm scan pattern. For a given OCTA scan, a reference surface corresponding to the location of BM was generated in several different ways, manually as well as using semi-automated and fully automated layer segmentation algorithms on B-scans. A completely automated algorithm used the OCTA scan and an associated BM surface as inputs to generate compensated CC slabs and binarized FDs maps, as described in several prior studies. Maps of local agreement between different FD maps of the same scan were generated by computing the absolute difference between the FD% within a small window centered at each pixel (Fig. 1C).
Two graders manually segmented the full BM in 4 eyes with large confluent drusen. The mean absolute distance was 5.4µm between these surfaces over the regions where drusen were present. The FDs maps produced by the graders were qualitatively and quantitatively different, see Fig.1. On an additional set of 5 eyes, the graders manually corrected the BM in the regions under drusen. The resulting mean CC FD% under the drusen were significantly different between the graders. However, a well-trained, experienced grader can generate consistent measurements on different images of the same eye. When such a grader manually corrected the BM in the central region (1mm diameter) of 28 eyes with drusen, each imaged three separate times, the mean standard deviation of the CC FD% measurements was 2.3%.
CC FD maps can be strongly dependent on the exact shape and position of the BM surface used to segment them. In regions where the retinal pigment epithelium (RPE) geometry and its relation to BM are affected by an underlying pathology, as when drusen are present, the variability associated the BM surface segmentation and the resulting CC FDs maps can be substantial. In these cases, great care needs to be exercised in both the measurement process and the interpretation of results.
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.
This PDF is available to Subscribers Only