Abstract
Purpose:
A new imaging technique for the assessment of retinal vascular oximetry has been introduced recently. The technique utilizes two-wavelength illumination of the fundus for the estimation of oxygen saturation in the retina. Utilizing a well-calibrated, three-dimensional stochastic model of light transport, we have examined the error associated with the two-wavelength technique (570 nm, 600 nm). We analyzed the error produced when the analyzing algorithm calibration assumptions are incorrect, vessel diameter varies, choroidal melanin concentration varies, and when there is vascular crosstalk from the choroid.
Methods:
The voxel-based Monte Carlo code used is capable of handling heterogeneous tissue structures, such as a vessel embedded in a stratified medium with different optical properties. A simplified tissue geometry comprised of four layers was used: retina, RPE, choroid, and sclera. A vessel embedded within the retina was also implemented. Two hundred million photons were launched in each simulation. Photons back-reflected by the layers were used to create images of fundus vasculature and then analyzed using the standard two wavelength algorithm approach. This included the calibration mechanism where a representative artery and a vein were selected and their values were fixed to 96% and 54 % respectively.
Results:
Under ideal conditions, two-wavelength retinal oximetry estimates compare well with true oxygen saturation levels of the vessel of interest. When the assumed values of calibrating arteries or veins do not correspond to true values, errors as high as 20% in veins and 10% in arteries may occur. When vessel size is increased, the optical density ratio (ODR) decreases and the assessed oxygen saturation error can be as high as 15% of its true value. Similar trends were found when analyzing the effect of melanin in the choroid and choroidal vessel crosstalk.
Conclusions:
Two-wavelength retinal oximetry is a popular technique because of its simplicity, however, it may be prone to significant error. An advanced light-tissue interaction modeling approach was effective in quantifying and elucidating key sources of error. Although some have attempted to correct oxygen saturation measurements for differences in pigmentation or vessel size, there are still many confounding variables that may degrade accuracy; this may lead to misdiagnosis of a range of retinal pathologies.