Unmixing of multispectral photoacoustic images

Photoacoustic imaging is an hybrid modality to image biological tissues. Using multispectral optical excitation, photoacoustic imaging allows to obtain functional images due to the fact that a tissue has a specific optical absorption depending on the used wavelengths. Quantification of multispectral photoacoustic images can be of great interest to different applications by quantifying oxygenated (HbO2) and deoxygenated (Hb) blood in tissue. Quantification can be done by examining the abundance maps resulting from unmixing methods. Two hyperspectral unmixing methods, namely Group Lasso with Unit sum and Positivity constraints (GLUP) and Fully Constrained Least-Square (FCLS), are used to quantify multispectral photoacoustic images. Experiments using two synthetic and one experimental dataset show that unmixing methods provide good quantification performances, which is of great interest for various medical applications.