Purchase this article with an account.
Kimberly E Stepien, Sri Meghana Konda, Tyler Etheridge, Ian Holmen, Laura Kopplin, Jeong W Pak, Amitha Domalpally; Impact of artifacts in optical coherence tomography angiography image analysis. Invest. Ophthalmol. Vis. Sci. 2020;61(7):4818.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Artifacts in optical coherence tomography angiography (OCTA) are known and include defocus, shadowing, movement, segmentation error, tilted scans, and projection. Recent advances in the OCTA algorithms have improved projection and motion artifacts. However, many artifacts such as tilt, defocus, and shadowing remain, which can significantly change the quantitative outputs of OCTA. We performed a prospective study to assess the quantitative impact of OCTA artifacts in normal and retinal disease.
Subjects were recruited prospectively for OCTA imaging using RTVue XR Avanti OCTA (Optovue Inc, Freemont, CA) using software version 2018.0.0.18. After consent was acquired, a reading center certified photographer acquired at least 6 OCTA images per eye; 3x3mm and 6x6mm scans of a good quality image, a scan where tilt was induced by patient and machine positioning, and a scan where movement was induced. De-identified images were then graded by the Fundus Photograph Reading Center, University of Wisconsin for presence and severity of artifacts defined as technical (decentration, segmentation, tilt, defocus, z-offset) and/or patient related (shadow, blink, eye movement). Each image was assigned a confidence score (CS). Scan quality metrics, fovea avascular zone area (FAZ) and vessel density (VD) were also recorded.
Of a total of 221 scans from 31 eyes included (9 normal, 22 retinal disease), 68% (151/221) were graded unreliable for VD due to artifacts with 71% (80/113) from 3mm scans and 66% (71/108) from 6mm scans. Images from subjects with retinal disease were more likely to be graded unreliable (74% retinal disease vs 54.7% normal). More than one artifact was present in 72% of images with unreliable VD. Reduced signal strength index was more common in unreliable images. Both technical error and patient attributed artifacts were present in 54% of all images. Twenty-two pairs of a good quality (CS1/CS2) and poor quality images (CS3) from the same patient were identified. Mean difference in FAZ was 0.01mm2, 95% CI [0.08, -0.06]. Mean difference in whole image VD of retina was 4.2, 95% CI [15.4, -7] and whole image VD of superficial retina was 3, 95% CI [26.9, -20.8].
Artifacts that affect quantitative measurements are highly prevalent in OCTA images and more common with retina pathology. Awareness of these artifacts when assessing OCTA is an important part of the interpretation of these images.
This is a 2020 ARVO Annual Meeting abstract.
This PDF is available to Subscribers Only