Abstract
Purpose :
Averaging multiple optical coherence tomography angiography (OCTA) structural and flow volumes is a powerful technique to improve quality but it can be memory and time consuming. In this work we introduce a method to average multiple OCTA acquisitions in a fast manner.
Methods :
Repeated retinal volumes of a 6x6mm field of view were collected using a PLEX® Elite 9000 with AngioPlex OCT Angiography (ZEISS, Dublin, CA). Quality scores were automatically generated for each repetition and the one of highest quality was chosen as reference. A set of volumetric registration functions are then obtained using both structural and flow information by registering each of the remaining repetitions to the reference both topographically, using common keypoints observed in flow data projections, and axially, using retinal layer information in the structural data. The average structure and flow volumes are then generated by combining the registrations of each repetition, producing higher-quality datasets that can be visualized and processed in the same manner as an individual OCTA acquisition. The method has some benefits over other common approaches using 2-dimensional averaging of flow slabs, such as the possibility of visualizing averaged B-scans of higher quality and employing the averaged volumes in segmentation algorithms.
Results :
We processed 240 OCTA volumes from 23 healthy and 22 disease eyes, with at least 3 repeated acquisitions per eye (average of 5.33). The individual acquisitions of highest quality form each repeated set and averaging results were graded by visual inspection of both B-scan and en face views in a scale of 1-5, with higher score indicating higher quality, and scores were compared (see figure). The averaging results for both the healthy and disease eyes resulted of significantly higher visual quality than observed in the individual acquisitions (t-test of scores p-value<0.01), both for visualizing retinal layers in a B-scan (3.24 and 4.64 average score for individual and averaged data, respectively) and vessels in flow slabs (3.31 and 4.15). The averaging results also had higher continuity in the observed vessels, mitigating possible tracking artifacts.
Conclusions :
The presented method can average repeated OCTA volumes producing results of enhanced quality in a relatively fast manner.
This abstract was presented at the 2019 ARVO Imaging in the Eye Conference, held in Vancouver, Canada, April 26-27, 2019.