Event-based cameras fire when individual pixels detect brightness changes, achieving microsecond temporal resolution — but they can't distinguish colors. Diffractive spectral imagers encode wavelength information through spatial dispersion — but they operate at conventional frame rates. Each sensor has what the other lacks.
This paper fuses both into a single fluorescence microscope. Asynchronous event streams provide the temporal backbone while diffraction-encoded CMOS measurements provide spectral information. The computational merger synchronizes the two data streams and achieves spectrally resolved tracking at 100,000 frames per second — without scanning, without filter switching, with 23 nm spectral discrimination.
Note (2026-11-13; read against the paper's full text): the paper does not say the fluorophores are distinguished in real time; I found no 'real-time' or processing-speed statement in it. The spectral identity comes from the diffraction-encoded CMOS sensor at its own frame rate, with a 100 ms exposure (used because mKate2 and mCherry are dim), and a neural network classifies it; those spectral labels are then assigned to the event-camera trajectories, so the spectral information is not itself measured at 100 microseconds. The 23 nm figure is one test pair, mKate2 and mCherry (emission peaks 23 nm apart, with more than 62 percent spectral overlap), not a general discrimination limit, and the system's spatial resolution is about 3.9 micrometres over a 0.5 mm field of view. The paper's 100 microseconds is 'effective temporal resolution down to' that value, with 100,000 frames per second described as an effective rate for the event camera. The behavioural measurement it reports is a response-latency difference in C. elegans omega turns (0.18 s in wild type, 0.88 s in unc-101), with sub-millisecond granularity in how the latency is quantified; the essay's 'microsecond regime of behavioral dynamics' echoes the paper's title, but the behaviour measured is at the sub-second scale.
The 100 μs effective temporal resolution reaches into the microsecond regime of behavioral dynamics in microscopy. The 23 nm spectral separation means fluorophores with nearly overlapping emission spectra can be distinguished in real time. Neither sensor alone could do this: the event camera doesn't see color, the spectral imager doesn't see fast enough. The fusion creates a capability that neither component contains. Decoupling temporal and spectral sampling — measuring fast in one modality and accurate in another — produces better combined performance than trying to do both in one sensor.