Characterizing and quantifying image artifacts in optical coherence tomography angiograms of normal retinal vasculature

Abstract

Purpose : Optical coherence tomography angiography (OCTA) is highly prone to image artifacts which can hinder the interpretation of OCTA images for clinical and research applications. The aim of this study is to develop a feasible method of characterizing and quantifying OCTA artifacts for objective analysis of image quality.

Methods : This was a retrospective study of 3×3 mm² OCTA scans (CIRRUS HD-OCT 5000) from healthy adult subjects without any major eye diseases. The en face angiogram and structural OCT image from each scan at the superficial retinal layer (SRL) was segmented for analysis. Four types of artifact were defined (decentration, motion artifact, defocus, and shadowing) and manually graded on a scale of 0 (not present), 1 (mild), and 2 (severe) as shown in Table 1 and Figure 1. A previously described and validated semiautomated algorithm was used to calculate vessel diameter index (VDI), vessel area density (VAD), vessel skeleton density (VSD), vessel complexity index (VCI), and flux. A linear mixed-effects model was used to analyze the effect of artifact severity on OCTA parameters.

Results : 152 scans from 35 eyes of 19 participants (age 69.4±6.7 yrs, 43.4% male) were included. Severe defocus (Ps<0.001), severe shadowing (Ps<0.001), and mild defocus (Ps<0.01) were negatively correlated with VDI, VAD, VSD, and flux, (range r=[-3.9×10^-3]-[-1.6×10^-1]). Severe decentration (Ps<0.05) was also negatively correlated with VDI, VAD, and VSD, but not flux, with similar magnitudes. Severe motion artifact was positively correlated with VCI (Ps<0.05, r=1.8×10⁷). VCI was not correlated with any other artifacts. Decentration, motion artifact, and shadowing of mild severity were not correlated with any of the vessel parameters.

Conclusions : The presence of severe decentration, defocus, and shadowing on OCTA scans was found to be quantitatively reflected in the vessel parameters of VDI, VAD, VSD, and flux, while motion artifact was uniquely correlated with VCI. Mild artifacts may not significantly impact most OCTA parameters.

This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.

 

Table 1: Grading rubric for type and severity of OCTA artifacts.

 

Figure 1: Examples of decentration (A), motion artifact (B), defocus (C), and shadowing (D) categorized as severe. Panels A, B, and C show angiograms of the SRL. Panel D shows a structural OCT image. Arrows highlight areas of artifact.