Adaptive spectroscopic visible-light optical coherence tomography for clinical retinal oximetry
Abstract Background Retinal oxygen saturation (sO 2 ) provides essential information about the eye’s response to pathological changes that can result in vision loss. Visible-light optical coherence tomography (vis-OCT) is a noninvasive tool that has the potential to measure retinal sO 2 in a clinical setting. However, its reliability is currently limited by unwanted signals referred to as spectral contaminants (SCs), and a comprehensive strategy to isolate true oxygen-dependent signals from SCs in vis-OCT is lacking. Methods We develop an adaptive spectroscopic vis-OCT (ADS-vis-OCT) technique that can adaptively remove SCs and accurately measure sO 2 under the unique conditions of each vessel. We also validate the accuracy of ADS-vis-OCT using ex vivo blood phantoms and assess its repeatability in the retina of healthy volunteers. Results In ex vivo blood phantoms, ADS-vis-OCT agrees with a blood gas machine with only a 1% bias in samples with sO 2 ranging from 0% to 100%. In the human retina, the root mean squared error between sO 2 values in major arteries measured by ADS-vis-OCT and a pulse oximeter is 2.1% across 18 research participants. Additionally, the standard deviations of repeated ADS-vis-OCT measurements of sO 2 values in smaller arteries and veins are 2.5% and 2.3%, respectively. Non-adaptive methods do not achieve comparable repeatabilities from healthy volunteers. Conclusions ADS-vis-OCT effectively removes SCs from human images, yielding accurate and repeatable sO 2 measurements in retinal arteries and veins with varying diameters. This work could have important implications for the clinical use of vis-OCT to manage eye diseases.