Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Implications of temporal sampling in voltage imaging microscopy

Jakub Czuchnowski, Jerome Mertz

TL;DR

This paper analyzes how temporal sampling strategies in voltage imaging microscopy shape spike-detection fidelity. Using a mathematical framework that combines analytical models with Monte-Carlo simulations, it characterizes the impact of duty cycle and frame rate on the measured spike waveform and detection performance under two detection schemes: peak detection and template matching. The results show that scanning approaches are advantageous at low sampling rates or for detecting a small subset of spikes, while wide-field imaging is more robust under heavy undersampling; as sampling improves, both modalities converge, with template matching outperforming peak detection. A key practical takeaway is a recommended sampling threshold around $f^* \approx \lambda$ (approximately 1 kHz for common voltage indicators), and a strong admonition against using scanning at frame rates below 500 Hz due to dramatic degradation in detection fidelity.

Abstract

Significance: Voltage imaging microscopy has emerged as a powerful tool to investigate neural activity both in vivo and in vitro. Various imaging approaches have been developed, including point-scanning, line-scanning and wide-field microscopes, however the effects of their different temporal sampling methods on signal fidelity have not yet been fully investigated. Aim: To provide an analysis of the inherent advantages and disadvantages of temporal sampling in scanning and wide-field microscopes and their effect on the fidelity of voltage spike detection. Approach: We develop a mathematical framework based on a mixture of analytical modeling and computer simulations with Monte-Carlo approaches. Results: Scanning microscopes outperform wide-field microscopes in low signal-to-noise conditions and when only a small subset of spikes needs to be detected. Wide-field microscopes outperform scanning microscopes when the measurement is temporally undersampled and a large fraction of the spikes needs to be detected. Both modalities converge in performance as sampling increases and the frame rate reaches the decay rate of the voltage indicator. Conclusions: Our work provides guidance for the selection of optimal temporal sampling parameters for voltage imaging. Most importantly it advises against using scanning voltage imaging microscopes at frame rates below 500 Hz.

Implications of temporal sampling in voltage imaging microscopy

TL;DR

This paper analyzes how temporal sampling strategies in voltage imaging microscopy shape spike-detection fidelity. Using a mathematical framework that combines analytical models with Monte-Carlo simulations, it characterizes the impact of duty cycle and frame rate on the measured spike waveform and detection performance under two detection schemes: peak detection and template matching. The results show that scanning approaches are advantageous at low sampling rates or for detecting a small subset of spikes, while wide-field imaging is more robust under heavy undersampling; as sampling improves, both modalities converge, with template matching outperforming peak detection. A key practical takeaway is a recommended sampling threshold around (approximately 1 kHz for common voltage indicators), and a strong admonition against using scanning at frame rates below 500 Hz due to dramatic degradation in detection fidelity.

Abstract

Significance: Voltage imaging microscopy has emerged as a powerful tool to investigate neural activity both in vivo and in vitro. Various imaging approaches have been developed, including point-scanning, line-scanning and wide-field microscopes, however the effects of their different temporal sampling methods on signal fidelity have not yet been fully investigated. Aim: To provide an analysis of the inherent advantages and disadvantages of temporal sampling in scanning and wide-field microscopes and their effect on the fidelity of voltage spike detection. Approach: We develop a mathematical framework based on a mixture of analytical modeling and computer simulations with Monte-Carlo approaches. Results: Scanning microscopes outperform wide-field microscopes in low signal-to-noise conditions and when only a small subset of spikes needs to be detected. Wide-field microscopes outperform scanning microscopes when the measurement is temporally undersampled and a large fraction of the spikes needs to be detected. Both modalities converge in performance as sampling increases and the frame rate reaches the decay rate of the voltage indicator. Conclusions: Our work provides guidance for the selection of optimal temporal sampling parameters for voltage imaging. Most importantly it advises against using scanning voltage imaging microscopes at frame rates below 500 Hz.
Paper Structure (4 sections, 22 equations, 3 figures)

This paper contains 4 sections, 22 equations, 3 figures.

Table of Contents

  1. Introduction
  2. Theory
  3. Results
  4. Conclusions

Figures (3)

  • Figure 1: (a) Different imaging configurations result in different temporal sampling of cells within a camera frame. (b) Shape of the measured voltage transient. (c) Sampling functions for wide-field ($\tau=1$) and scanning ($\tau=0.1$) configurations. (d-e) Averaged spike shapes for wide-field (d) and scanning (e) configurations ($f=0.4$) where $\sigma[X] = \sqrt{V[X]}$ is the standard deviation. (e-f) Dependence of the expected value of the voltage signal peak on sampling frequency (e - wide-field, f - line-scan). (g) Ratio of the expected peak values between scanning and wide-field configurations. (h-k) Z-score comparison of peak detection (PD) fidelity between wide-field and scanning configurations. (l-o) Z-score comparison of template matching (TM) fidelity between wide-field and scanning configurations for different numbers of detected photons.
  • Figure 2: (a) Probability distributions of spike heights calculated both analytically ($f=1$) and by Monte Carlo (MC). (b-i) Comparison of spike detection efficiency for both scanning and wide-field systems using PD (b-e) and TM (f-i). (j-m) Comparison of the relative sensitivity (wide-field/scanning) for both PD and TM for different sampling frequencies.
  • Figure 3: (a-d) Estimated lower bound of the SNR for peak detection (PM) and template matching (TM) using scanning and wide-field systems for different number of detected photons, assuming a target $E=95\%$. (e-h) Estimated lower bound of the SNR for PD and TM using scanning and wide-field systems for different number of detected photons, assuming a target $E=99\%$.