Summary

Researchers at King's College London report TomoFLIM, a compressed multiphoton fluorescence lifetime imaging method. In a preprint benchmark, it reproduced a 60-second raster-scanned dataset in 3.75 seconds.

Researchers at King's College London report a fluorescence lifetime imaging method that uses compressed multiphoton measurements and computational reconstruction to accelerate image acquisition. In a bioRxiv preprint posted on September 16, 2026, the team says its TomoFLIM system reproduced a raster-scanned fluorescence lifetime imaging dataset acquired in 60 seconds in 3.75 seconds—a 16-fold increase in frame rate.

The reported method also achieved compression ratios above 90%, with Pearson correlation coefficients above 80% relative to reference images. The experiments used calibrated fluorescence lifetime beads and biological specimens. The authors present the technique as a potential way to capture faster biological processes, including events deep within turbid samples.

Contents

Why fluorescence lifetime matters

Fluorescence lifetime imaging microscopy, or FLIM, measures how long fluorescent molecules remain in an excited state before emitting photons. That timing carries information about molecular interactions, biochemical surroundings and cellular physiology. Unlike an image based only on brightness, fluorescence lifetime can provide quantitative contrast that is less dependent on the concentration of the fluorescent label.

The method requires enough time-resolved photon statistics to estimate those lifetimes accurately. This makes FLIM relatively slow, particularly when each location in an image must be measured separately. In dynamic applications such as calcium signalling or vesicular trafficking, the acquisition can take longer than the biological event being observed. Changes may then be averaged together, while movement can introduce artefacts and transient information can be lost.

How TomoFLIM compresses the measurement

TomoFLIM excites two-photon fluorescence with a line focus projected tomographically across the sample, rather than relying only on a conventional raster-scanned acquisition. The emitted fluorescence is time-tagged using time-correlated single-photon counting, which records the timing of individual detected photons relative to excitation.

The time-resolved tomographic measurements are reconstructed computationally. The reported pipeline applies Lucy–Richardson deconvolution and then uses a centre-of-mass method to estimate fluorescence lifetimes. The researchers also developed TomoFLIM Net, a physics-informed neural network that directly reconstructs fluorescence intensity and lifetime from compressed time-resolved tomographic data.

This design shifts part of the measurement task from collecting a complete raster scan to recovering quantitative information from fewer, structured observations. The intended benefit is a shorter acquisition while retaining the lifetime information needed to distinguish different molecular or biochemical environments.

What the preprint demonstrated

The team benchmarked TomoFLIM against raster-scanned FLIM using calibrated fluorescence lifetime beads and biological specimens. The system operated at compression ratios exceeding 90%, while the reconstructed images had Pearson correlation coefficients above 80% relative to reference images.

In one comparison, a raster-scanned FLIM dataset collected over 60 seconds was reproduced using TomoFLIM in 3.75 seconds. The authors report this as a 16-fold increase in frame rate and an equivalent reduction in accumulated detector dark counts. TomoFLIM Net also recovered distinct experimental bead lifetime populations, indicating that the compressed measurements retained separable lifetime information in that test.

The evidence is from a bioRxiv preprint and its reported benchmarks, not from a clinical or commercial imaging deployment. The abstract describes live-cell imaging and imaging deep within turbid biological specimens as potential applications; the reported demonstrations are framed around calibrated beads and biological specimens. If the approach performs similarly in dynamic samples, faster acquisition could reduce temporal averaging and motion artefacts while preserving quantitative lifetime maps.

Sources