Purpose

To determine factors that may affect the repeatability of the foveal avascular zone (FAZ) measurement obtained using optical coherence tomography angiography (OCTA) including instrument type, image segmentation, image quality, and fundus pathology.

Patients and methods

This prospective single-center study enrolled 43 subjects (85 eyes) with retinal vasculopathy, macular edema, optic pathology or normal contralateral eye. The macula was imaged twice using Optovue Angiovue and once using Cirrus Angioplex to obtain 3x3mm OCTA images centered on the fovea. Images were generated by the same operator within 30 mins. The FAZ size for the entire retinal thickness ("overall FAZ") was measured automatically using the OCTA software. The FAZ size of the superficial and deep retinal vascular plexus layers was measured manually using the enface OCTA images of the segmented layers and Image J analysis. Intraclass correlations coefficient (ICC) was calculated to determine repeatability.

Results

For the overall FAZ measurement, repeatability was excellent (ICC 0.953 right eye, 0.938, left eye) using the same machine (intra-instrument) and somewhat lower but still good to excellent (ICC 0.803 right eye, 0.917 left eye) using machines made by different vendors (inter-instrument). For the segmented layers, intra-instrument repeatability of FAZ measurement was excellent (ICC > 0.95) for both plexus layers. Inter-instrument repeatability was good for the superficial plexus layer (ICC 0.86 right eye, 0.88 left eye) but reduced for the deep plexus layer (ICC 0.63 right eye, 0.57 left eye). Suboptimal image quality and presence of retinal vasculopathy and macular edema tended to reduce FAZ repeatability but to a lesser degree.

Conclusion

Inter- and intra-instrument repeatability of the overall FAZ measurement was high using commercial OCTA instruments and only mildly reduced by suboptimal image quality and fundus pathology. For segmented layers, intra-instrument repeatability remained high but inter-instrument repeatability was reduced for the deep plexus layer.