Purpose
Spectral domain optical coherence tomography (OCT) allows in vivo measurements retinal layers. However, the utilization of OCT for tracking retinal degeneration in mice has been challenged because of the absence of data regarding test-retest reliability for the parameters measured. As opposed to histology, OCT offers the opportunity to perform longitudinal measures. We aimed to identify the most suitable scan protocols for the quantitative analysis of the retinal layers in cross-sectional and longitudinal studies in mice.
Methods
Retinal images from 10 C57BL/6 mice were obtained using spectral domain (SD)-OCT (Heidelberg Spectralis). We performed raster based volume, peripapillary ring, and horizontal and vertical line scans with semi-automated segmentation of retinal layers to analyze the retest reliability of consecutive measurements. Independent investigators performed semi-automated segmentation to assess inter-rater reliability.
Results
We obtained the best reproducibility with retinal layer thickness measurements from consecutive volume scans (interclass correlation coefficient (ICC), 95% confidence interval (CI), total retinal thickness: 0.96, 0.83-0.99; retinal nerve fiber layer: 0.81, 0.04-0.96; ganglion cell layer plus inner plexiform layer: 0.94, 0.75-0.99; inner nuclear layer: 0.81, 0.16-0.96; outer plexiform layer: 0.61, -0.33-0.91; outer nuclear layer: 0.96, 0.83-0.99; photoreceptor layer: 0.93, 0.64-0.99). Peripapillary ring scans showed the poorest reliability and reproducibility. The inner retinal layers showed more inter-rater reliability (0.995, 0.989-0.998) and reproducibility (0.94, 0.73-0.99) than the outer retinal layers.
Conclusions
Our data suggest that among the different scan protocols, volume scans are most suitable to assess the thickness of the different retinal layers. The measurement of inner retinal layers thickness showed better reliability and reproducibility than the deeper retinal layers. These data indicate that OCT reliability is dependent on the scan protocols selected and that particular scan protocols in mice show promise for monitoring longitudinally.