Advancing Complex Fluorescence Lifetime Image Reconstruction via Deep 3D-FCNs

Fluorescence lifetime imaging (FLI) is a mainstay imaging technique with the unique benefit of enabling sensitive quantification of the biological micro-environment, such as with metabolic status and reactive oxygen species. Moreover, FLI has an increasing role in preclinical and clinical settings thanks to its ability to reveal tissue composition, to enable biomarker multiplex-imaging and to provide enhanced data sets for tomographic applications. Further, FLI is the most robust approach to perform Förster Resonance Energy Transfer (FRET) which recently has been demonstrated to enable the quantification of target-receptor engagement in live subject. However, the technique is somewhat limited due in large part to its reliance on time-consuming inverse solvers. Herein, we demonstrate a novel fit-free approach for FLI reconstruction based on a 3D-Fully Convolutional Network (3D-FCN) trained entirely in silico capable of high-performance quantification of complex fluorescence decays simultaneously over a whole image in quasi real-time. Our microscopic (FLIM) and macroscopic (MFLI) studies establish our analytic framework as a robust tool for FLI studies over a large range of lifetimes (visible-NIR), photon count and technologies employed.