A spectral-domain model-based method for simultaneous oxygen saturation quantification and contrast agent identification

Unmixing multispectral photoacoustic (PA) images is difficult because the excitation spectra in deep tissue are contaminated by absorption and scattering of the surrounding tissue in a highly unpredictable manner. In this work, we found a close relationship between the covariance matrix of a multispectral photoacoustic image and its average tissue oxygenation level. Based on the photon diffusion process, a spectral-domain model of multispectral photoacoustic imaging is established. Combined with the above two findings, accurate estimation of blood oxygen saturation (median error 2.7%) and accurate probe identification (detection rate 86%, false alarm rate 0.035%) were realized in realistic simulation test.