Purchase this article with an account.
Min Gao, Tristan T. Hormel, Jie Wang, Yukun Guo, Steven Bailey, Thomas S Hwang, Yali Jia; Reconstruction of high-resolution OCT angiograms of retinal intermediate and deep capillary plexuses using deep learning. Invest. Ophthalmol. Vis. Sci. 2021;62(8):1032.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
We propose to use deep learning to reconstruct flow signal from under-sampled 6×6-mm optical coherence tomographic angiography (OCTA) images of the intermediate capillary plexus (ICP) and deep capillary plexus (DCP).
6×6-mm macular scans with a 400×400 A-line sampling density and 3×3-mm scans with a 304×304 A-line sampling density were acquired on one or both eyes of 180 participants (including 230 eyes with diabetic retinopathy (DR) and 44 healthy controls) using a 70-kHz commercial OCT system (RTVue-XR; Optovue, Inc.). Projection-resolved OCTA algorithm was applied to remove projection artifacts in voxel. ICP and DCP angiograms were generated by maximum projection of the OCTA signal within the relevant plexus. We proposed a deep-learning-based method, dubbed “deep capillary angiograms reconstruction network” (DCARnet), to reconstruct 6×6-mm high-resolution ICP and DCP en face OCTA images from sparsely-sampled, low-resolution scans of the same area. DCARnet takes registered 3×3-mm ICP and DCP angiograms with proper sampling density as the ground truth reference. Same network can also be applied on 3×3-mm angiograms. We evaluated the reconstructed 3×3- and 6×6-mm angiograms based on vessel connectivity, false flow signal (flow signal erroneously generated from background), and the noise intensity in the foveal avascular zone (FAZ).
Compared to the originals, the angiograms reconstructed by DCARnet significantly reduced noise intensity (ICP, 7.38±25.22, p<0.001; DCP, 11.20±22.52, p<0.001), improved vascular connectivity (ICP, 0.95±0.01, p<0.001; DCP, 0.96±0.01, p<0.001), and did not generate false flow signal at the level of noise intensity in normal FAZ. DCARnet not only enhanced the image quality of 6×6-mm ICP and DCP angiograms, but also reduced noise and improved connectivity in 3×3-mm ICP and DCP angiograms. Furthermore, DCARnet preserves the appearance of the dilated vessels in the reconstructed angiograms.
DCARnet can reconstruct high-resolution ICP and DCP angiograms from low-definition 6×6-mm en face OCTA images. The enhanced angiograms may improve characterization of biomarkers such as non-perfusion area and vessel density.
This is a 2021 ARVO Annual Meeting abstract.
This PDF is available to Subscribers Only