Non-negative Matrix Factorization under sparsity constraints to unmix in vivo spectrally resolved acquisitions

Abstract

Fluorescence imaging in diffusive media is an emerging imaging modality for medical applications which uses injected fluorescent markers (several ones may be simultaneously injected) that bind to specific targets, as tumors. The region of interest is illuminated with near infrared light and the emitted back fluorescence is analyzed to localize the fluorescence sources. To investigate thick medium, as the fluorescence signal decreases with the light travel distance, any disturbing signal, such as biological tissues intrinsic fluorescence — called autofluorescence —, is a limiting factor. To remove autofluorescence and isolate each specific fluorescent signal from the others, a spectroscopic approach, based on Non-negative Matrix Factorization, is explored. We ran an NMF algorithm with sparsity constraints on experimental data, and successfully obtained separated in vivo fluorescence spectra.