Deconvolution of Sustained Neural Activity From Large-Scale Calcium Imaging Data

Area of Science:

• Neuroscience
• Biophysics
• Computational Biology

Background:

• Light-sheet microscopy enables whole-brain functional imaging at cellular resolution using calcium indicators.
• Analyzing large-scale imaging data is difficult due to noisy fluorescence signals and limited temporal resolution, hindering conventional spike inference.

Purpose of the Study:

• To develop a robust method for extracting meaningful neuronal activity information from large-scale, noisy functional imaging data.
• To model sustained neuronal activity moments rather than individual spikes for improved analysis.

Main Methods:

• Deconvolution of calcium response using a linear system model.
• Incorporation of a differential operator inverse within a regularization term for sparsity.
• Application of generalized total variation for promoting activity transient sparsity.

Main Results:

• Demonstrated numerical performance on simulated signals, showing model superiority over spike inference at specific firing rate transitions.
• Successfully applied the algorithm to experimental zebrafish larval data.
• Validated retrieved neural activation by correlating it with unknown locomotor behavior.

Conclusions:

• The proposed deconvolution and activity moment modeling approach effectively extracts neuronal information from challenging large-scale functional imaging data.
• This method offers a significant advancement over conventional spike inference for analyzing complex brain activity patterns.
• The algorithm's ability to link neural activity to behavior highlights its potential for understanding neural circuit function.