*4.1.2 Operator/machine-related motion artifacts*

There are several operator/machine related artifacts that affect the accuracy and precision of OCTA measurements. Poorly focused images cause defocused artifacts (8A) that are caused by a decrease in reflective intensity resulting in reduced definition of retinal microvasculature Projection artifacts (8B) occur when the shadow of vessels in the superficial layer imprint onto the deeper layer. Segmentation errors (8C) are produced when the automated segmentation algorithm inaccurately identifies the layers of the retina or optic nerve. Decentration artifact (8D) occurs if the fovea or ONH is translocated to the periphery of the scan, preventing the algorithm from correctly identifying these areas. Cropping (8E) refers to loss in image data caused by B-scan not being fully within the capture window. Lastly, vitreous floaters may result in shadow artifacts (8F) because of decreased intensity of retinal layers in focused areas and may prevent clear visualization of structures in the affected area.

#### *4.1.3 Key factors that produce optical coherence tomography angiography artifacts*

Since OCTA artifacts may affect the accuracy of glaucoma diagnosis it is important to understand when and why these alterations are likely to occur. Older patients often have poor-quality scans due to a higher frequency of media opacities such as cataracts and age-related retinal pathologies. Recently Kamalipour et al. showed that 40% of OCTA images in the glaucoma group had poorer quality compared to healthy controls [47]. This was confirmed by Cheng et al. who found that the presence of glaucoma significantly increased the odds for the presence of any artifacts [48]. Similarly, patients with glaucoma were found to have two times the odds of incorrect automatic segmentation of the retinal superficial layer even after adjustment for age and VF loss [48]. Image quality was also found to decrease as glaucoma progressed. Declining retinal thickness reduces the signal intensity of reflected waves causing increased prevalence of artifacts.

#### *4.1.4 Prevalence of optical coherence tomography angiography artifacts*

In a well-diversified population of North Texas consisting of 292 patients, Mekala et al. found that most patients (99.3%) had at least one artifact [49]. The most common artifacts were due to eye movements (66.1%), defocus (64.7%) and shadows (40%). The most severe (affecting more than 10% of the scan area) were seen in patients with POAG (31.4%) and myopia (30.75%). A similar study by Weijing et al. found that 88.34% of images had at least one artifact, the most common of which was projection (100%), followed by motion (75.22%), blur (24.78%) and decentration (21.28%) [50]. Another study by Kamalipour et al. excluded 33.9% of scans due to poor image quality and found that 13.6% had one artifact and 9.8% had two or more. In this study the most common artifact was eye movement (10.6%), followed by defocus (9.6%), correctable segmentation errors (7.6%), and uncorrectable segmentation errors (5.4%) [48]. Given that multiple studies have confirmed the presence of artifacts in a significant percentage of images, it is crucial to account for artifacts when evaluating OCTA in a clinical setting.

#### *4.1.5 Current methods for reducing optical coherence tomography angiography artifacts*

Eye tracking technology was first introduced in 2004 as a way to compensate for eye movement and improve the quality of OCT scans [51]. Since then, this technology has shown to decrease the odds of acquiring poor-quality images by half [47]. It is highly effective in reducing eye movement and defocus artifacts but does not currently decrease the occurrence of segmentation artifacts.

Work by Venugopal et al. found that high-density (HD) scans, which increase sampling density from 304 × 304 A-scans to 400 × 400 A-scans, showed significantly greater VD values and fewer poor-quality scans than non-HD scans [52]. Thus, both eye tracking technology and HD scanning are proven methods for reducing the presence of OCTA artifacts. However, additional work is needed to further reduce the prevalence of other artifacts in OCTA.
