Regularized Nonnegative Matrix Factorization for autofluorescence removal in fluorescence optical imaging

Fluorescence imaging in diffusive media locates cancers thanks to injected fluorescent markers specific to the tumors. The region of interest is illuminated with red light and the emitted back fluorescence is analyzed to locate the fluorescence sources. To detect accurately the markers signal and be able to explore thick media (breast, prostate), autofluores-cence emitted by biological tissues has to be removed. We propose a spectroscopic approach, based on Non-negative Matrix Factorization (NMF) method, and present interest of regularized NMF algorithms on unmixing results. In vivo autofluorescence removal and tumor detection enhancement results on mice will be presented.