Fluorescence lifetime imaging (FLI) provides unique quantitative information in biomedical and molecular biology studies but relies on complex data-fitting techniques to derive the quantities of interest. Herein, we propose a fit-free approach in FLI image formation that is based on deep learning (DL) to quantify fluorescence decays simultaneously over a whole image and at fast speeds. We report on a deep neural network (DNN) architecture, named fluorescence lifetime imaging network (FLI-Net) that is designed and trained for different classes of experiments, including visible FLI and near-infrared (NIR) FLI microscopy (FLIM) and NIR gated macroscopy FLI (MFLI). FLI-Net outputs quantitatively the spatially resolved lifetime-based parameters that are typically employed in the field. We validate the utility of the FLI-Net framework by performing quantitative microscopic and preclinical lifetime-based studies across the visible and NIR spectra, as well as across the 2 main data acquisition technologies. These results demonstrate that FLI-Net is well suited to accurately quantify complex fluorescence lifetimes in cells and, in real time, in intact animals without any parameter settings. Hence, FLI-Net paves the way to reproducible and quantitative lifetime studies at unprecedented speeds, for improved dissemination and impact of FLI in many important biomedical applications ranging from fundamental discoveries in molecular and cellular biology to clinical translation.

Over the last two decades, functional near-infrared spectroscopy (fNIRS) has become a well-established neuroimaging modality to monitor brain activity.¹ The ability of fNIRS to quantify cortical tissue hemodynamics over a long time, with relative high-spatial sampling and temporal resolution, has enabled its adoption in numerous clinical settings.^2,3 fNIRS offers the unique advantage to be employed in freely mobile subjects with less restrictions than electroencephalography (EEG) or functional magnetic resonance imaging (fMRI). This permits the deployment of fNIRS in naturalistic scenarios and in patient populations that are typically not considered suitable for EEG or fMRI imaging.⁴ Still, fNIRS faces numerous challenges for increased clinical adoption due to experimental settings,⁵ variations in statistical results,⁶ etc. Of importance, current trends in fNIRS aim to improve spatial resolution via increased spatial sampling, improve cortical sensitivity using data processing to remove unwanted physiological noise, improve quantification by anatomical coregistration, and increase robustness via artifact identification and removal.⁷ Current algorithmic implementations, however, require a high level of expertise to set up parameters that can be system- and/or application-specific but also greatly impact the interpretability of the processed data. Moreover, the computational cost of these methods does not lend itself to bedside implementations. Following a ubiquitous trend in the field of data processing and analysis, new approaches leveraging developments in deep learning (DL) have been recently proposed to help overcome these caveats to a large extent.

The ability to probe deep tissues with light for biomedical applications has been recognized since the scientific reports published as early as the late 1800s for monitoring brain hemorrhage¹ and early 1900s for imaging breast cancer² and performing tissue oximetry.^3,4 Since then, diffuse optical imaging techniques have greatly benefited numerous biomedical fields. Mainly, the goal of these optical techniques is to characterize optical properties of tissue using light from the ultraviolet to the infrared spectral region. The specific light–tissue interaction enables monitoring of a vast array of tissue characteristics, including structural, physiological, metabolic, and molecular properties.