FLIM data analysis based on Laguerre polynomial decomposition and machine-learning.

Area of Science:

• Biophysics
• Microscopy
• Data Science

Background:

• Fluorescence lifetime imaging microscopy (FLIM) is a valuable technique in biological studies, but its data processing relies on estimating lifetimes and abundances from decay traces.
• Conventional methods like curve fitting are computationally intensive and struggle with noisy FLIM data, while graphical analysis requires calibration samples.

Purpose of the Study:

• To develop and validate a machine learning (ML)-based approach for directly extracting fluorescence lifetimes and abundances from FLIM decay traces.
• To compare the performance of the ML method against a conventional commercial software (SPCImage).

Main Methods:

• A novel ML algorithm was developed to directly estimate lifetimes and abundances from fluorescence decay traces.
• The ML approach was validated using simulated data with known parameters and subsequently tested on biological sample datasets acquired via time-correlated single photon counting.
• Performance was evaluated by reconstructing decay traces and calculating the root-mean-squared-error (RMSE) compared to the commercial software SPCImage.

Main Results:

• The ML-based method demonstrated a lower RMSE, indicating higher accuracy in reconstructing decay traces compared to SPCImage.
• The ML approach proved effective in handling complex scenarios, successfully analyzing data with three lifetime components, showcasing its flexibility.

Conclusions:

• The proposed ML-based approach exhibits excellent performance for FLIM data analysis.
• This method offers a more precise and potentially faster alternative for extracting critical parameters from FLIM measurements, especially for complex biological systems.