This study has revealed that high myopes had better fixation stability during OCTA measurement when soft contact lenses were worn (TRR: 10.2 pixels, CV: 65%), compared with using the built-in auto-focus system (TRR: 12.5 pixels, CV: 72%). However, low myopes had better fixation stability than high myopes regardless of correction method used (Table 2). Regarding fixation deviation, it was similar regardless of correction method in each group. Clinically, it is common to acquire OCTA images using the built-in auto-focus system to compensate for patients’ refractive errors. Fixation deviation was found to be significantly greater in high myopes when compared with low myopes using the built-in auto-focus system. Post-hoc power analysis, using G*Power (version 18.104.22.168, Dusseldorf, Germany), determined that the sample size had 86% power (α = 0.05, one-tailed) to detect an effect that exists.
The implications of fixation deviation and test-retest repeatability are different. A poor test-retest repeatability means that the scanned areas varied a lot in sequential measurements. A great fixation deviation means that the fovea is not at the center of the grid. For the latter, one or two ETDRS sectors could have smaller scanned areas compared with the others. However, it could still have good test-retest repeatability from consecutive measurements. Test-retest repeatability could be more important. Two examples can be used to illustrate this. Figure 4 is OCTA maps of the left eye from three consecutive measurements of a high myope, of which Figs. 4a) to 4c) were results using the auto-focus mode. Fixation deviation from each measurement varied differently (from 3.0 pixels to 19.2 pixels, mean of 9.1 pixels). The test-retest repeatability was 8.8 pixels. Figure 4d) to 4f) were results using habitual contact lens correction, in which fixation deviation in each measurement was less (from 2.2 pixels to 9.2 pixels, mean of 5.2 pixels) and the test-retest repeatability was smaller, 3.6 pixels. Both the mean fixation deviation and TRR were better using contact lens correction. Figure 5 shows OCTA maps of the left eye from three consecutive measurements of another high myope. Although mean fixation deviations were similar in the two correction modes (6.0 and 7.1 pixels), fixation deviations from three consecutive measurements were all towards the same superior temporal region, which resulted in a small TRR (3.3 and 2.7 pixels).
From our knowledge, this is the first study to investigate the effect of fixation deviation and stability during OCTA measurements. Clinically most practitioners acquire just one OCTA measurement because there is no averaging function from any proprietary software. It is important to have the same measurement area in follow up visits for accurate comparison and diagnosis. Taking Fig. 4 as an example, if an examiner relies on Fig. 4a), the outer ring at the temporal sector was cropped. The cropped areas were outer rings of the nasal and inferior sectors in Fig. 4b). To tackle this problem, a practitioner can put the ETDRS grid to the center of the 6x6mm scan area. This is not normally done because the fovea may not be at the center of the 6x6mm scan area.
A 6x6 mm scan has 350x350 pixels. The transverse resolution is 17.2µm. A 3x3mm scan has a higher resolution of 12.2µm, but the scan area is smaller. The test-retest repeatability in high myopes was 12.5 pixels when using the auto-focus system, which was improved to 10.2 pixels when corrected with soft contact lenses. This is equivalent to a variation of 215µm and 175µm, respectively. This order of variation would result in significant measurement errors in OCT.14,15
Signal strength was found better in low myopes (Table 2). Previous studies have demonstrated impact of low signal strength on OCTA measurements.20 The signal strength was set at 7 as the criterion and the averaged signal strength was above 8 in all four conditions. Lim et al.16 compared effect of signal strength on OCTA metrics using the same Cirrus OCTA device. VD and PD increased with a higher signal strength. There was no significant difference in VD and PD between signal strength of 9 and 10. Lee et al.21 also used a Cirrus OCTA device and found that repeatability (in terms of coefficient of variation) improved even when signal strength was slightly higher from 9 to 10. OCTA metrics were found increased when signal strength was 9 compared with signal strength of 8. Each OCTA device has its own proprietary algorithms. Yu et al.10 found that vessel density from AngioPlex was more influenced by signal strength compared with AngioVue.
It is important to have good fixation stability to ensure similar ETDRS subfields are measured in consecutive OCTA measurements. To achieve this, myopes with good CL-VA can continue wearing their contact lenses during OCTA measurements. This is more important in subjects with refractive astigmatism corrected with toric contact lenses (Table 2). Berkenstock et al.11 corrected their patients with soft contact lenses based on spherical equivalent even for astigmatic refractive errors. Optimal refractive correction is important in various ophthalmic procedures, such as kinetic perimetry,22 OCT,8,23 and OCTA.9 Youm et al.24 compared retinal nerve fiber layer (RNFL) thickness measurement with and without soft contact lens wear. Although they found that RNFL was thicker without soft contact lenses, they used old version time-domain OCT with poor resolution.
In general, little difference in OCTA metrics was observed via two different correction methods. This could be due to the protocol using an average from three OCTA measurements. Previous studies found that averaging OCTA images could improve vessel clarity that might enhance accurate acquisition of OCTA metrics.13,25 Uji et al.26 reported that averaging three OCTA frames resulted in significant improvement, while averaging five frames could result in almost identical findings. Clinically, automated averaging of several OCTA metrics for an overall result is not available from any proprietary software. In addition, it may not be feasible to acquire too many OCTA images. Our recent study found that OCTA metrics from averaging three OCTA measurements could reach similar VD and PD compared with averaging five measurements.19 Considering that OCTA is more commonly performed in patients with a compromised retina and, hence, with poor vision, taking several OCTA measurements may be advisable.
There are different OCTA devices which are not interchangeable.27 Therefore, patients must be reviewed using the same machine for accurate longitudinal monitoring. Our institute has both the Cirrus (Cirrus 5000 HD-OCT; Carl Zeiss Meditec, Inc, Dublin, California, USA) and Spectralis (Spectralis; Heidelberg Engineering, Heidelberg, Germany) systems, which use a split-spectrum amplitude-decorrelation algorithm and a probabilistic OCTA algorithm, respectively. The Cirrus system was found to be superior to the Spectralis system in terms of fewer artefacts and better repeatability.28,29 The operation manual from Spectralis advised examiners to use glasses or contact lenses for patients with high astigmatism in order to enhance image quality. From the current findings, contact lens correction for OCTA measurements is recommended, especially for high myopes.