Voltage imaging, a powerful tool for recording neural activity, has exploded in popularity in recent years. It directly translates neuronal membrane potential into fluorescence signals via fluorescent probes, offering a direct and powerful way to capture neural activity. This technique combines the strengths of optical imaging and electrophysiology, overcoming the limitations of both. Surprisingly, this seemingly perfect technique isn't a recent invention; it predates even calcium imaging¹, the current gold standard in optical neural recording. So, what hindered its development?

The Challenges of Voltage Imaging
One of major obstacles has been the inability of optical imaging technologies to meet the demanding requirements of voltage imaging. To effectively record electrical signals from neurons, the tool must offer millisecond-scale sampling rates. For observing neural activity in behaving mice, high throughput and extended imaging durations become crucial. Moreover, capturing subtle subthreshold activity necessitates a high signal-to-noise ratio (SNR)².
Currently available voltage imaging methods present a difficult trade-off: wide-field microscopy offers high-speed imaging but suffered from background interference, while two-photon microscopy is free of background interference but has low throughput. While recent optimizations, such as incorporating targeted illumination^3,4 into wide-field microscopy and introducing spatial and temporal multiplexing⁵ in two-photon microscopy, have been attempted, but the demonstrated imaging throughputs still lagged significantly behind calcium imaging. 

An opportunity from Light Field Microscopy
To overcome these limitations, we turned our attention to the light-field microscopy that was under development in our lab⁶. It can capture an entire 3D field of view in a single exposure, offering inherent advantages in imaging speed and throughput. More importantly, all information can be captured with high efficiency without wasting fluorescence signal that is especially precious in voltage imaging. These characteristics align remarkably well with voltage imaging requirements.

The Challenging Development Journey
The journey to develop a confocal light field voltage imaging system, however, was not as straightforward as we had anticipated. Initial concerns centered around finding a suitable camera. Light-field microscopy requires larger sensors, but a high-speed camera with a large chip is extremely rare. Fortunately, a newly released camera met our needs—but it wasn't perfect. To achieve the necessary speed, it sacrificed dynamic range, potentially compromising the SNR—a critical factor in voltage imaging. However, we employed the principle of generalized confocal detection to selectively filter out background noise, thereby reducing the signal baseline. By effectively integrating information from multiple views, we succeeded in efficiently capturing weak voltage signals using cameras with low dynamic range.

Subsequent in-vivo experiments were unsuccessful at the beginning due to severe noise contamination. To address this, we systematically investigated noise sources in light-field imaging. We identified several significant factors degrading the SNR in voltage imaging, including intensity fluctuations from laser sources, synchronization noise from scanning galvanometers, and laser speckle noise caused by blood flow in animals. To overcome these challenges, we innovatively proposed a combination of approaches, including a single-mirror double-sided scanning mechanism, a high numerical aperture (NA) illumination strategy, and advanced data processing methods. This combination of innovative techniques effectively reduced system noise to the theoretical Poisson noise limit. 

Finally, the volume of data generated by the high-speed voltage imaging system presented a significant computational challenge. To handle the camera's data transmission rate, we upgraded our computer's RAM to 96 GB and equipped it with a high-speed SSD for faster data. The immense data volume also dramatically increased the computational cost of 3D reconstruction. To meet the challenge of processing large amount of data, we opted for a simple and fast algorithm--synthetic refocusing, which has a minor reduction in spatial resolution to significantly reduce the computational burden.

The Rewards of Perseverance
Through persistent effort and continuous refinement, we successfully integrated several key innovations into a new confocal light-field microscope, achieving simultaneous recording of voltage signals from hundreds of neurons in a three-dimensional field of view (800 μm in diameter, 180 μm in depth) in the brain of awake mice. Our system can continuously image at 400 volumes per second for over 20 minutes. This development significantly enhances the practicality of voltage imaging and represents a valuable advancement in meeting the growing demand for in vivo voltage recording across larger neuronal populations and within 3D volumes.

Reference

1.    Knöpfel, T. & Song, C. Optical voltage imaging in neurons: moving from technology development to practical tool. Nat Rev Neurosci 20, 719–727 (2019).
2.    Wu, Z., Lin, D. & Li, Y. Pushing the frontiers: tools for monitoring neurotransmitters and neuromodulators. Nat Rev Neurosci 23, 257–274 (2022).
3.    Xiao, S. et al. Large-scale voltage imaging in behaving mice using targeted illumination. iScience 24, 103263 (2021).
4.    Fan, L. Z. et al. All-Optical Electrophysiology Reveals the Role of Lateral Inhibition in Sensory Processing in Cortical Layer 1. Cell 180, 521-535.e18 (2020).
5.    Platisa, J. et al. High-speed low-light in vivo two-photon voltage imaging of large neuronal populations. Nat Methods 20, 1095–1103 (2023).
6.    Zhang, Z. et al. Imaging volumetric dynamics at high speed in mouse and zebrafish brain with confocal light field microscopy. Nat Biotechnol 39, 74–83 (2021).