Blind end-member and abundance extraction for multispectral fluorescence lifetime imaging microscopy data

Area of Science:

• Chemometrics
• Microscopy
• Biomedical Imaging

Background:

• Multispectral fluorescence lifetime imaging microscopy (m-FLIM) generates complex data requiring advanced analysis.
• Extracting underlying spectral signatures (end-members) and their proportions (abundances) is crucial for m-FLIM data interpretation.
• Existing methods like MCR-ALS and MVC-NMF have limitations in accuracy and robustness.

Purpose of the Study:

• To introduce a novel blind end-member and abundance extraction (BEAE) method for m-FLIM data.
• To enhance the accuracy and robustness of end-member and abundance estimation in m-FLIM.
• To compare the performance of BEAE against established chemometrical techniques.

Main Methods:

• Developed a BEAE method based on a linear mixture model with positivity and sum-to-one constraints.
• Employed an iterative estimation process utilizing quadratic optimization solved via alternating least-squares.
• Validated the method using synthetic m-FLIM datasets and real-world data from atherosclerotic plaques.

Main Results:

• BEAE demonstrated superior accuracy in estimating end-members with a minimum relative error of 0.32% and a worst-case scenario of 13.82%.
• BEAE achieved lower absolute errors for abundance estimation (0.05 minimum, 0.12 worst-case) compared to MVC-NMF and MCR-ALS.
• The method showed improved convergence and robustness against signal variability, with a computation time faster than MCR-ALS.

Conclusions:

• The proposed BEAE method provides more accurate and robust end-member and abundance extraction for m-FLIM data.
• BEAE offers a significant advancement over MVC-NMF and MCR-ALS for analyzing complex fluorescence decay mixtures.
• This technique holds promise for detailed analysis of biological samples using m-FLIM.