Dimensionality reduction of calcium-imaged neuronal population activity

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Calcium imaging is a key technique for recording large neuronal populations.
• Dimensionality reduction is useful for summarizing neural activity, but its application to calcium imaging is not well-established.
• Existing methods for spiking activity may not be optimal for calcium imaging data.

Purpose of the Study:

• To systematically evaluate dimensionality reduction methods for calcium imaging data.
• To develop a novel method for simultaneous deconvolution and dimensionality reduction.
• To assess the performance of new and existing methods on simulated and experimental calcium imaging data.

Main Methods:

• Systematic study of standard dimensionality reduction techniques applied to calcium imaging.
• Development of the Calcium Imaging Linear Dynamical System (CILDS) method for joint deconvolution and dimensionality reduction.
• Validation using simulated calcium imaging data and recordings from larval zebrafish and mice.

Main Results:

• CILDS demonstrated superior accuracy in recovering single-trial, low-dimensional neural time courses from simulated data.
• CILDS outperformed other tested methods on real calcium imaging recordings from zebrafish and mice.
• The study provides a robust framework for analyzing large-scale neuronal population activity.

Conclusions:

• Dimensionality reduction is effective for summarizing calcium imaging data.
• CILDS offers a powerful and accurate approach for analyzing neural population dynamics.
• This work establishes a foundation for applying dimensionality reduction to calcium imaging across various experimental contexts.