High frequency spike inference with particle Gibbs sampling.

Area of Science:

• Neuroscience
• Computational Biology
• Biophysics

Background:

• Calcium-sensitive fluorescent indicators are crucial for monitoring neuronal activity in vivo.
• Existing spike-time inference methods struggle with high firing rates (>20 Hz) and lack uncertainty quantification.
• Accurate estimation of spike timing and its statistical uncertainty is vital for interpreting neuronal population dynamics.

Purpose of the Study:

• To develop a robust Bayesian inference method for accurate neuronal spike-time estimation from calcium imaging data.
• To address limitations in current methods regarding high firing rates and uncertainty quantification.
• To provide a flexible framework for analyzing neuronal activity with genetically encoded calcium indicators.

Main Methods:

• Developed a statistical model incorporating bursting activity and baseline fluorescence modulation.
• Applied a particle Gibbs with ancestor sampling Monte Carlo strategy for joint posterior distribution estimation.
• Validated the method against state-of-the-art algorithms on benchmark datasets (CASCADE).

Main Results:

• The proposed Bayesian method demonstrates competitive performance with existing algorithms.
• Successfully resolves interspike intervals as short as 5 ms using the GCaMP8f indicator.
• Provides unbiased estimates of spike times and model parameters, quantifying statistical uncertainties.

Conclusions:

• The developed Bayesian inference method enhances the accuracy of neuronal spike detection from calcium imaging.
• Quantification of uncertainty provides crucial confidence levels for inferred spike patterns.
• The particle Gibbs framework offers a flexible approach for advancing calcium indicator data analysis.