Recording the electrical chatter of hundreds of neurons at once has long been the holy grail of neuroscience. Traditional patch-clamp techniques measure one cell at a time, and even the most advanced cameras force you to trade spatial detail for speed. A new study by Kim, Yoon, Ko, Kang, Tian, Fan, Li, Xiao, Zhang, Cohen, Wu, Dai & Choi in Nature Communications introduces DeMOSAIC, an ingenious optical segmentation strategy that squeezes maximal temporal information out of each pixel—making it possible to capture circuit-scale voltage dynamics at over 5 kHz without sacrificing subcellular resolution.
The problem:
When you use a high-speed camera to watch neurons fire, you end up with massive image files—only to throw away most of that data later when you average each region of interest (ROI) into a single intensity trace. Not only is that wasteful, it also forces a choice between recording fast enough to see millisecond electrical events and maintaining enough pixels to resolve tiny neuronal processes.
The DeMOSAIC solution:
DeMOSAIC (Diffractive Multisite Optical Segmentation Assisted Image Compression) flips the script: before you start recording, you take a snapshot, draw your ROIs, and then display a custom “grating” pattern on spatial light modulators. Each ROI is diffracted to its own angle, focused through a microlens array, and imaged onto just one detector pixel. The result? Your raw data are already in the exact form you need—one pixel per ROI—so you can sample at lightning speeds without colossal files.
Proof of principle:
-
100 kHz “graffiti” test: By optically segmenting a dynamic light pattern spelling “i AM CODES,” they showed DeMOSAIC can faithfully record 9 intertwined subimages at 125 kHz.
-
Calcium imaging: An EMCCD camera plus pixel binning recorded 21 neurons’ calcium transients with compressed 120×3-pixel frames at several hundred hertz.
-
Voltage imaging in vitro: Using the dye BeRST1 and field stimulation, they tracked 23 subcellular ROIs at 5.5 kHz—then applied deep-learning denoising (DeepCAD-RT) to boost single-trial spike detection.
-
In vivo demos: In anesthetized mice, DeMOSAIC measured blood-cell flow in pial vessels at 4 kHz and, with targeted illumination, captured spontaneous action potentials of inhibitory interneurons through a cranial window at 4 kHz.
Why it matters:
-
Unprecedented combination of subcellular spatial precision and kilohertz sampling over tens of ROIs.
-
Modular add-on: DeMOSAIC can retrofit most widefield microscopes without rewriting your analysis pipeline.
-
New neuroscience insights: Track rapid voltage waveforms and submillisecond delays across entire microcircuits.
As neuroscience pushes toward truly “all-optical” brain mapping, DeMOSAIC offers a powerful, data-efficient way to capture the fleeting electrical conversations that underlie perception, movement, and cognition.
Cosmael ThinkLab commentary:
By rethinking how we acquire and compress optical data, DeMOSAIC sidesteps a fundamental bottleneck in high-speed imaging. This approach could transform not only basic neuroscience but also any field where fast, multiplexed signals must be recorded, from cardiac tissue to microfluidics.
Source:
Kim, S., Yoon, J., Ko, G., Kang, I., Tian, H., Fan, L. Z., Li, Y., Xiao, G., Zhang, Q., Cohen, A. E., Wu, J., Dai, Q. & Choi, M. “Optical segmentation-based compressed readout of neuronal voltage dynamics.” Nature Communications (2025). optical