Retinal blood oxygen saturation mapping by multispectral imaging and morphological angiography

We report on an experiment for mapping two-dimensional blood oxygen saturation distribution by measuring multispectral images in the wavelength range from 500 to 650 nm with the resolution of 7 nm. The multispectral images of the retina are acquired with the originally designed imaging system equipped with the tunable spectral filter. To separate retinal blood vessels from other tissue area, morphological image processing is adopted. The small flick motion is also compensated. After prepossessing, the partial least squares regression model for the oxygen saturation is built by sampling typical spectra reflected from artery and vein. Applying the regression model for all the point extracted by the morphological processing allows the two-dimensional oxygen saturation map.