Highly multiplexed spectral FLIM via physics informed data analysis.

Area of Science:

• Biophotonics and Imaging
• Molecular Spectroscopy
• Computational Biology

Background:

• Spectral fluorescence lifetime imaging (S-FLIM) combines spectral and lifetime data for multi-fluorophore deconvolution.
• Existing S-FLIM analyses struggle with pre-calibration dependency, spectral overlap, and high photon requirements.
• Fluorophore spectra are inherently present in S-FLIM data, negating the need for external calibration.

Purpose of the Study:

• To develop a novel Bayesian S-FLIM framework for simultaneous, photon-by-photon learning of spectra and lifetimes.
• To overcome limitations of current S-FLIM analysis, including spectral overlap and high photon budgets.
• To enhance data efficiency and enable deconvolution with limited photon counts.

Main Methods:

• A Bayesian framework is proposed for S-FLIM data analysis.
• The method performs simultaneous spectral and lifetime learning on a photon-by-photon basis.
• Framework is validated using both synthetic and experimental S-FLIM datasets.

Main Results:

• The Bayesian framework demonstrates high data efficiency, requiring fewer photons per pixel.
• It successfully deconvolves signals from up to 9 spectrally overlapping fluorophore species.
• The method eliminates the need for pre-calibrated spectral information.

Conclusions:

• The proposed Bayesian S-FLIM framework offers a robust and efficient solution for multi-species deconvolution.
• It significantly reduces photon budget requirements and overcomes spectral overlap challenges.
• This approach advances S-FLIM capabilities for analyzing complex biological samples.