Event-Based Denoising Enables High-Fidelity Fast Fluorescence Lifetime Imaging

Fluorescence lifetime imaging microscopy (FLIM) can reveal how molecules behave in their native environments, including protein conformations, intermolecular interactions and micro-scale biochemical changes. Yet traditional FLIM is held back by a basic bottleneck: accurate lifetime estimation typically requires large photon counts, which slows acquisition and limits use in fast, deep-tissue imaging where light is scarce.

A team led by Zhou, Xiao, Zhou and colleagues now reports a strategy designed specifically for ultralow-light conditions. In their approach, called event-based first-photon FLIM (EFLIM), the microscope records each excitation as an event—capturing the first photon arrival—rather than building conventional photon histograms. This shift converts the data from binned intensity accumulations into a binary, event-centric representation.

EFLIM is paired with a self-supervised denoising framework that learns to infer fluorescence lifetime from these sparse event streams. Because the method does not rely on high-photon statistics, it reduces the photon requirement by more than two orders of magnitude relative to leading FLIM algorithms. The authors highlight that lifetime estimates can remain reliable at extremely low signals, even approaching an “apparent mean lifetime” measurement below one photon per pixel.

Importantly, the system is described as robust to intensity artifacts, a frequent failure mode when illumination fluctuates or when measured intensities drift across tissue depth. By focusing on first-photon timing and by denoising without external labels, EFLIM aims to preserve lifetime fidelity while minimizing sensitivity to non-ideal acquisition conditions.

To validate the technique, the researchers tracked transient intracellular dynamics linked to ligand-dependent molecular states. They also used multiplexed imaging to capture putative vesicle-mediated contacts between different lymphocytes within a single spectral channel, leveraging the speed advantage to observe biologically relevant interactions.

Beyond cell biology, EFLIM enabled rapid, label-free visualization of tumor heterogeneity in human glioma tissue. The authors report that the combination of speed and low photon demand makes it feasible to probe structural and compositional differences without prolonged illumination.

Together, these results position event-based denoising as a practical path toward high-fidelity FLIM under light-limited conditions. The authors argue the method could accelerate applications in neuroscience, immunology and pathology, where dynamic molecular processes in vivo are often out of reach for conventional lifetime imaging.