Detection of Changes in Retinal Vessel Blood Oxygen Saturation using Hyperspectral Imaging

Abstract

Purpose: To describe the use of a prototype hyperspectral retinal camera (HRC) for detecting changes in the blood oxygen saturation of retinal arterioles and venules using manual retinal reflectance intensity measurements.

Methods: Eleven (8 male and 3 female) healthy, non-smoking, volunteers participated in the study. Following pupillary dilation, three repeated sets of retinal images were captured at 10 nm intervals between 500 and 700 nm, and for the wavelength pair 586nm and 610nm, using a newly developed HRC (Photon HR Camera, Optina^TM, QC, Canada). Images were taken at baseline end-tidal oxygen partial pressure (P_ETO₂) and hypoxia (P_ETO₂, 50mmHg). The P_ETO₂ was regulated using a computer controlled end-tidal targeting device (RespirAct^TM, Thornhill Inc., Toronto, ON, Canada) connected to a sequential re-breathing circuit. Throughout the procedure, the arterial oxygenation was monitored by means of finger pulse oximetry (SPO₂). ImageJ^TM was then used to generate reflectance profiles of a first-degree retinal arteriole and a venule to determine: the minimum within vessel reflectance intensity, minimum within vessel reflectance intensity ratio of arterioles and venules, and the optical density (OD) of the vessel.

Results: As the mean SPO ₂ decreased from 98% +/- 1% (mean +/-SD i.e. normoxia) to 86% +/- 3% (i.e. hypoxia), mean arteriolar retinal oxygen saturation (SO₂) values, based on the wavelength pair 586nm and 610nm, also decreased significantly (96% +/- 12.6% to 85% +/- 13% p=0.003). The mean SO₂ across retinal venules also decreased from baseline to hypoxia; however, this difference was not statistically significant (56% +/- 12.4% to 50% +/- 15.9% ; p=0.493). When comparing hyperspectral retinal images at 500nm to 700nm between the two provocations, repeated-measures ANOVA showed statistically significant differences for: OD (p < 0.001), minimum within vessel reflectance intensity (p < 0.001), and minimum within vessel reflectance intensity ratio (p < 0.001), for both retinal arterioles and venules.

Conclusions: Retinal oximetry using this novel prototype HRC has the potential to detect changes in retinal vessel blood oxygen saturation. Automation of hyperspectral image analyses is likely to further improve its clinical applicability.

Keywords: 688 retina • 550 imaging/image analysis: clinical • 635 oxygen