Multiphoton minimal inertia scanning for fast acquisition of neural activity signals

Renaud Schuck, Mary Ann Go, Stefania Garasto, Stephanie Reynolds, Pier Luigi Dragotti, Simon R. SchultzbioRxiv 2017 · DOI

TL;DR

This paper introduces the Adaptive Spiral Scanning (SSA) algorithm, a novel method for multiphoton microscopy that significantly increases the speed of neural activity imaging. SSA achieves this by optimizing scanning trajectories for the mechanical limitations of galvanometric scanners, enabling the simultaneous observation of hundreds of neurons with high temporal precision, outperforming traditional raster scanning and existing optimized methods like Traveling Salesman Scanning (TSS) in many scenarios.

What Problem Does This Paper Address?

Monitoring the activity of neural circuits in real-time is crucial for understanding brain function. Multiphoton laser scanning microscopy (MPLSM) is a powerful tool for this, but its inherent point-scanning nature limits how quickly it can acquire data. Standard "raster scanning" (moving the laser back and forth like an old TV) is slow, especially when trying to image many neurons across a large field of view. While methods like Traveling Salesman Scanning (TSS) try to speed things up by only visiting regions of interest (like neuron cell bodies), they don't fully account for the mechanical inertia of the scanning mirrors, which can cause them to overshoot or undershoot their targets at high speeds, leading to missed data or inaccurate positioning. This paper tackles the challenge of achieving high cellular sampling rates and temporal precision in neural activity imaging without requiring specialized, expensive hardware.

Methods

The core contribution is the Adaptive Spiral Scanning (SSA) algorithm. Unlike TSS, which aims for the shortest path between all points (like a real Traveling Salesman), SSA generates a series of near-circular trajectories. It starts by sorting neuron locations by their distance from the center of the field of view and then uses a model of galvanometric scanner mechanics to create spiral paths. This approach is designed to minimize the inertial effects of the mirrors, allowing for faster and more accurate movement. The researchers also implemented and tested the Traveling Salesman Scanning (TSS) algorithm, which uses a genetic algorithm to find an optimized path between target neurons. They compared both SSA and TSS against standard raster scanning. To evaluate signal quality, they used the Cramér-Rao Bound (CRB), a theoretical measure of the lowest possible uncertainty in estimating spike timing from calcium imaging data, alongside simultaneous electrophysiology recordings for ground truth.

Datasets

The paper utilizes two main types of data:

  • Simulated neuron spatial distributions: The authors generated synthetic datasets of neuron locations within a defined field of view. These simulations allowed for controlled testing of scanning strategies across various neural densities.
  • In vitro hippocampal mouse brain slices: They used actual brain slices from juvenile mice, loaded with the calcium indicator dye Cal-520 AM. This allowed for real-world validation of the scanning strategies and comparison with simultaneous electrophysiology recordings.

Key Insights and Ideas

  • Inertia is the enemy of speed: The paper highlights that the mechanical inertia of galvanometric scanners is a major bottleneck for fast scanning. Simply finding the shortest path (TSS) isn't enough if the mirrors can't physically make those sharp turns quickly.
  • Spiral paths are smoother: SSA's approach of using near-circular trajectories is inherently better suited to the physics of mirror movement, allowing for higher speeds without missing targets.
  • Speed vs. Signal Quality Trade-off: While faster scanning is desirable, it can sometimes come at the cost of signal-to-noise ratio (SNR) because the laser spends less time on each neuron. SSA offers a good balance, and TSS can achieve higher SNR by "parking" on neurons, but at a lower maximum speed.
  • Temporal Precision Matters: For accurately reconstructing neural activity, especially spike timing, both sampling frequency (how often you check a neuron) and SNR are important. The CRB analysis shows that SSA and TSS can achieve comparable temporal precision to raster scanning, even with lower SNR, due to their significantly higher sampling rates.
  • Adaptability is Key: The paper emphasizes that the "best" scanning strategy depends on the specific experimental setup, neural density, and research question. SSA is presented as a flexible and easily implementable solution for standard MPLSM systems.

Key Results

  • SSA significantly outperforms raster scanning: SSA achieved sampling rates an order of magnitude higher than raster scanning for the same field of view and number of cells.
  • SSA generally outperforms TSS: For neural densities typical of the cortex, SSA achieved higher sampling frequencies than TSS. For example, in a simulated population of 200 neurons, SSA reached 103 Hz while TSS maxed out at 51 Hz before missing cells.
  • TSS can offer higher SNR: When precise targeting is possible, TSS can achieve higher SNR than SSA because it allows for longer dwell times on each neuron. However, this comes at the cost of maximum sampling speed.
  • Comparable temporal precision: Despite differences in SNR and sampling speed, both SSA and TSS achieved comparable accuracy in spike timing estimates (as measured by CRB) to raster scanning, demonstrating their effectiveness for capturing neural dynamics.
  • High neural densities are feasible: The study showed that SSA is particularly advantageous at higher neural densities (e.g., >1400 neurons/mm²), which are common in many brain regions.

Limitations and Future Work

  • Dye loading dependency: The effectiveness of these scanning strategies, particularly in in vitro experiments, is dependent on how well calcium indicator dyes can be loaded into the neurons. This can be challenging in older brain slices or deeper tissues.
  • Voltage indicators: While the paper focuses on calcium imaging, the authors note that genetically encoded voltage indicators, which have faster kinetics, might benefit even more from the high sampling rates offered by SSA and TSS. However, achieving sufficient SNR with these indicators remains an open question.
  • Future directions: The authors suggest that further research into adaptable scanning strategies is needed to fine-tune parameters for optimal signal quality across different experimental conditions. They also point to the potential of voltage indicators and the need to assess their SNR capabilities with these advanced scanning methods. The paper also mentions that the SSA algorithm is easily implementable on standard MPLSMs, implying its broad applicability for future research.

Summary updated

Abstract

Abstract Multi-photon laser scanning microscopy provides a powerful tool for monitoring the spatiotemporal dynamics of neural circuit activity. It is, however, intrinsically a point scanning technique. Standard raster scanning enables imaging at subcellular resolution; however, acquisition rates are limited by the size of the field of view to be scanned. Recently developed scanning strategies such as Travelling Salesman Scanning (TSS) have been developed to maximize cellular sampling rate by scanning only regions of interest in the field of view corresponding to locations of interest such as somata. However, such strategies are not optimized for the mechanical properties of galvanometric scanners. We describe here the Adaptive Spiral Scanning (SSA) algorithm, which fits a set of near-circular trajectories to the cellular distribution to avoid inertial drifts of galvanometer position. We compare its performance to raster scanning and TSS in terms of cellular sampling frequency and signal-to-noise ratio (SNR). Using surrogate neuron spatial position data, we show that SSA acquisition rates are an order of magnitude higher than those for raster scanning and generally exceed those achieved by TSS for neural densities comparable with those found in the cortex. We show that this result also holds true for in vitro hippocampal mouse brain slices bath loaded with the synthetic calcium dye Cal-520 AM. The ability of TSS to ”park” the laser on each neuron along the scanning trajectory, however, enables higher SNR than SSA when all targets are precisely scanned. Raster scanning has the highest SNR but at a substantial cost in number of cells scanned. To understand the impact of sampling rate and SNR on functional calcium imaging, we used the Cramér-Rao Bound on evoked calcium traces recorded simultaneously with electrophysiology traces to calculate the lower bound estimate of the spike timing occurrence. The results show that TSS and SSA achieve comparable accuracy in spike time estimates compared to raster scanning despite their lower SNR. SSA is an easily implementable way for standard multi-photon laser scanning systems to gain temporal precision in the detection of action potentials while scanning hundreds of active cells.
BibTeX
@misc{schuck2017multiphoton,
  author = {Renaud Schuck and Mary Ann Go and Stefania Garasto and Stephanie Reynolds and Pier Luigi Dragotti and Simon R. Schultz},
  title = {{Multiphoton minimal inertia scanning for fast acquisition of neural activity signals}},
  year = {2017},
  doi = {10.1101/196428},
  howpublished = {bioRxiv},
}
Figure