Nonnegative matrix factorization: a blind spectra separation method for in vivo fluorescent optical imaging.

Area of Science:

• Medical imaging
• Biophotonics
• Spectroscopy

Background:

• Fluorescence imaging in diffusive media is promising for medical applications, using fluorescent markers to target specific molecules like carcinoma.
• Signal attenuation and autofluorescence in thick tissues limit imaging depth and specificity.
• Separating signals from multiple fluorescent markers or background autofluorescence is crucial for accurate localization.

Purpose of the Study:

• To develop and validate a spectroscopic method for isolating specific fluorescence signals from interfering contributions in thick diffusive media.
• To enhance the accuracy and specificity of fluorescence-guided medical imaging.
• To demonstrate the separation of in vivo fluorescence spectra using a novel algorithm.

Main Methods:

• Exploration of a spectroscopic approach for fluorescence signal separation.
• Application of nonnegative matrix factorization (NMF), a blind positive source separation technique.
• Development and implementation of an original regularized NMF algorithm on experimental data.

Main Results:

• Successful separation of in vivo fluorescence spectra was achieved using the developed regularized NMF algorithm.
• The method effectively distinguished between specific fluorescent markers and autofluorescence.
• Demonstrated the potential to isolate individual fluorescence signals in complex biological environments.

Conclusions:

• The regularized NMF algorithm provides an effective solution for separating mixed fluorescence signals in diffusive media.
• This spectroscopic approach significantly improves the quality and interpretability of fluorescence imaging data.
• The findings support the advancement of fluorescence imaging for precise molecular targeting and diagnostics in medicine.