Assessing patterns for compressive fluorescence lifetime imaging.

Area of Science:

• Optics and Photonics
• Biomedical Imaging
• Spectroscopy

Background:

• Hyperspectral imaging offers rich spectral information but generates large datasets.
• Compressive sensing (CS) reduces data acquisition requirements by exploiting signal sparsity.
• Optimizing CS basis patterns is crucial for efficient hyperspectral data reconstruction and analysis.

Purpose of the Study:

• To evaluate and compare various compressive basis patterns for hyperspectral imaging applications.
• To determine the optimal basis patterns for intensity imaging, fluorescence lifetime imaging (FLI), and Förster Resonance Energy Transfer (FRET) quantification.
• To assess the performance of selected patterns under varying experimental conditions and compression ratios.

Main Methods:

• Development and implementation of a novel hyperspectral single-pixel system.
• Experimental comparison of six popular basis patterns using a two-fluorescent-dye phantom.
• Application of best-performing patterns for FRET quantification in well-plate samples with varied acceptor-donor ratios.
• Analysis of image quality and quantification accuracy at a 50% compression ratio.

Main Results:

• The ABS-WP approach utilizing Haar patterns demonstrated superior performance.
• Compressive sensing with Hadamard Ranked patterns also exhibited excellent overall performance.
• Both Haar and Hadamard patterns excelled in FLI and FRET quantification tasks.
• Optimal pattern selection significantly improved data reconstruction and quantitative accuracy.

Conclusions:

• Haar and Hadamard Ranked patterns are highly effective compressive basis patterns for hyperspectral single-pixel imaging.
• These patterns enable accurate FLI and FRET quantification at significant data compression ratios.
• The findings provide a foundation for developing more efficient and robust hyperspectral imaging systems.