Vis-OCT is an emerging technique in ophthalmology, providing metabolic information of the retina, specifically oxygen saturation ( \({sO}_{2}\) ), based on the modified Beer-Lambert law (mBLL). However, the conventional mBLL-based model simplifies the scattering attenuation with an exponential term accounting for the tissue absorption. This simplification reduces accuracy because multiple scattering is non-negligible. Additionally, the OCT spectral signal for a blood vessel is usually extracted from a region of interest (ROI) that is either located on the posterior wall or encompasses the entire vessel, thereby compromising spatial resolution. To overcome such limitations, we proposed a diffusion correction of the Beer-Lambert law (dcBLL) from the perspective of the radiative transfer equation (RTE), solved by the diffusion approximation, which takes both single and multiple scatterings into account. Pixel-wise \({sO}_{2}\) estimation was validated through Monte Carlo simulations (MCS) and in vivo human retinal imaging. The dcBLL model demonstrated clear superiority in MCS, with RMSE values less than 10% of those obtained with the mBLL model. In human retinal oximetry, dcBLL further enables pixel-wise \({sO}_{2}\) mapping, in contrast to the conventional approach that averages values over large vessels.
Graphical Abstract
