Purchase this article with an account.
Homayoun Bagherinia, Luis De Sisternes, Thomas Callan, Patricia Sha, Maximilian Pfau, Laura Tracewell, Susan Su, Roger Goldberg, Mary K Durbin; A method for automated Bruch’s membrane segmentation in optical coherence tomography. Invest. Ophthalmol. Vis. Sci. 2020;61(7):489.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Accurate Bruch’s membrane (BM) segmentation is essential to characterize possible choriocapillaris loss as well as elevations and dysfunctions of the retinal pigment epithelium, which are important diagnostic indicators of retinal diseases. This abstract proposes a BM segmentation method in optical coherence tomography (OCT) volumes.
The BM segmentation algorithm enhances the BM layer by using both structural (Vs) and flow (Va) OCT volumes. The enhanced OCT volume (Ve) is calculated by subtracting a proportion of mixture of structural and flow data from the structural data as Ve=Vs-α(wsVs+waVa ). ws and wa are weights and set to 0.5. α=Cov(wsVs+waVa,Vs)/Var(wsVs+waVa) assuming the similarity (squared normalized cross correlation) between Ve and the mixture (wsVs+waVa) is minimized. The segmentation algorithm is based on a multiresolution approach and a graph search algorithm. The segmentation baseline of each resolution level is used as a starting segmentation for the segmentation of the next higher resolution. The number of resolution levels is set to two for faster processing. Performance of the algorithm is evaluated by comparison to manual edits from two readers using 120 B-scans extracted from 40 OCTA cube scans of prototype 3x3 mm, 6x6 mm, 9x9 mm, 12x12 mm, and 15x9 mm acquired using 200kHz PLEX® Elite 9000 (ZEISS, Dublin, CA). All scans were mix of disease cases such as DR and AMD.
Fig 1 shows two examples of enhanced OCT for BM visualization using structural and flow data, and three examples for BM segmentation with corresponding choriocapillaris vasculature maps. Fig 2 shows the mean absolute difference (including 95% confidence interval) and R2 between two readers and the readers and BM segmentation. The mean absolute difference for each scan pattern demonstrates strong correlation and great agreement between the readers and BM segmentation.
We proposed a new BM segmentation algorithm. Overall the automated and manual segmentations have a strong correlation and great agreement. Automated segmentation may be a valuable diagnostic tool for retinal diseases.
This is a 2020 ARVO Annual Meeting abstract.
Fig 1: First two rows show examples of OCT enhanced B-scans with corresponding OCT and OCT flow B-scans. Last two rows show the segmentation results with corresponding choriocapillaris slabs by using OCT flow data between BM and BM+20 µm.
Fig 2: Mean absolute difference and R2 between two readers and BM segmentation.
This PDF is available to Subscribers Only