Images With Less Than One Photon: EFLIM Speeds Up Visualization of Temporal Signals Inside Cells

Fluorescence microscopy can record not only where light is emitted, but also how long it takes for that light to fade, revealing the chemical environment and molecular interactions within cells. Accurately estimating this “fluorescence lifetime,” measured in mere nanoseconds, usually requires the accumulation of large numbers of photons. Once living tissue moves, the signal can easily become distorted, while prolonged illumination may also cause photobleaching and phototoxicity.

A research team comprising Tsinghua University and other institutions published an event-based first-photon fluorescence lifetime imaging method, EFLIM, in *Nature Biotechnology*. Conventional analysis accumulates the arrival times of large numbers of photons into a histogram. EFLIM instead treats each laser excitation as an event: the detector either receives no photon or records the arrival time of the first photon detected. This representation preserves the temporal structure of extremely sparse signals, after which a self-supervised denoising model estimates fluorescence lifetimes without requiring high-photon-count images as ground-truth training targets.

The paper shows that EFLIM reduces the number of photons required by more than two orders of magnitude compared with existing algorithms, while still producing stable lifetime images under conditions averaging less than one photon per pixel. The research team also observed neuronal calcium signals approximately 250 micrometers deep in the brains of awake mice. According to information released by the university, imaging used about 0.2 to 0.8 photons per pixel, and the method reduced artifacts caused by animal movement and fluctuations in fluorescence intensity.

Low photon levels provide more than gentler illumination; they also make fast biological processes easier to capture. The researchers tracked transient intracellular calcium changes in living HeLa cells and used different fluorescence lifetimes within a single spectral channel to distinguish germinal center B cells from follicular helper T cells. Three-dimensional imaging over two hours showed the movement and direct contact of the two types of lymphocytes, as well as intercellular signaling that the research team determined might be vesicle-mediated. “Might” remains the key qualification: the images alone are not sufficient to establish the origin and function of the vesicles.

The method’s pathological application was demonstrated using human glioma tissue. According to the university’s information, the researchers imaged 169 fields of view, with each field taking 0.33 seconds—about one-tenth the time required by conventional methods. Without adding exogenous labels, the images distinguished differences in fluorescence lifetime among tumor cells, blood vessels, stroma, and necrotic regions, with comparisons made against histological sections. The result suggests that endogenous fluorescence may help rapidly map tumor heterogeneity, but it cannot yet be equated with clinical validation of diagnostic accuracy or surgical benefit.

The team has made publicly available a Python implementation of EFLIM, MATLAB simulations of photon arrivals, representative raw data, and workflows for processing Becker & Hickl SPC and PicoQuant PTU data. Publicly available code helps other laboratories assess reproducibility and lowers the barrier to adopting the method. However, independent testing is still needed to determine whether equivalent performance can be maintained across different microscopes, detectors, tissue types, and dyes.

EFLIM is currently, first and foremost, a computational imaging method rather than a clinically approved pathology tool. Self-supervised models may transform noise into apparently smooth structures, so future work must not only compare image quality but also quantify bias, failure scenarios, and stability across devices. Only if these issues are clarified can the speed and safety margin gained by reducing photon requirements truly expand its range of applications in in vivo neuroscience, immunology, and rapid tissue assessment.