Retinal Vessel Oxygen Saturation Measurement Protocols and Their Agreement

To assess agreement between different image sizes and analysis protocols for determination of retinal vessel oxygen saturation in the peripapillary retina of healthy individuals.

Retinal oximetry measurements were acquired from 87 healthy volunteers using the IMEDOS Systems oxygen module. The peripapillary retinal vessels were assessed in a concentric annulus around the optic nerve head. Single and average vessel comparisons were made at different image field sizes of 30° and 50°. Comparisons between images obtained at 30° and 50° were made in a subset of 47 of the 87 individuals.

All subjects were normotensive and had normal intraocular pressures (9-16 mm Hg). Analyses of agreement between single vessel, averaged vessel, and between different size images were sought by Bland-Altman analyses, of which all yielded a low bias (<1% oxygen saturation). However, agreement between single vessels of consecutive images showed increased limits of agreement compared with saturation values calculated by averaging all or just the four major arcades of one image. Agreement between 30° and 50° images showed a similar bias as when comparing data obtained with the same camera angle setting but exhibited larger confidence intervals (arteries: bias = 0.21% [9.04/-8.62]%; veins: bias = 0.71% [14.82/-13.40]%).

Averaging methods yielded the best agreement; there was little difference in average arterial and venous oxygen saturation between protocols, which analyze all vessels versus the four largest vessels. The least agreement was found for single vessel measurements and comparisons between different camera angles.

Standardization of image capture protocols (same image size and undertaking a vessel averaging approach for oxygenation analysis) will enhance the detection of smaller physiological changes in eye disease.