Blind deconvolution estimation by multi-exponential models and alternated least squares approximations: Free-form and...

Area of Science:

• Microscopy and Imaging
• Biophysics
• Data Analysis

Background:

• Quantitative evaluation of Fluorescence Lifetime Imaging Microscopy (FLIM) relies on accurate deconvolution.
• Blind Deconvolution Estimation (BDE) jointly extracts sample fluorescence impulse responses (FluoIRs) and instrument response (InstR) with minimal prior information.
• Existing BDE methods face challenges in modeling complex FLIM data.

Purpose of the Study:

• Introduce two novel BDE algorithms for FLIM data analysis.
• Model FluoIRs using linear combinations of multi-exponential functions.
• Investigate both local and global perspectives for parameter estimation.

Main Methods:

• Developed two BDE algorithms using multi-exponential functions for FluoIRs and free-form/sparse InstR.
• Employed alternating least squares (ALS) for iterative, non-negative, and constrained optimization.
• Validated algorithms using synthetic datasets with varying noise and experimental FLIM data.

Main Results:

• Local and global BDE perspectives demonstrated consistency with standard deconvolution techniques.
• Achieved faster convergence rates compared to existing BDE algorithms.
• Showcased a superior balance between FluoIRs and InstR estimation errors.

Conclusions:

• The proposed BDE algorithms offer an efficient and accurate approach for FLIM quantitative analysis.
• The local and global perspectives provide robust parameter estimation for FLIM data.
• These methods represent a significant advancement in BDE for FLIM applications.