Calcium imaging permits optical measurement of neural activity. Since intracellular calcium concentration is an indirect measurement of neural activity, computational tools are necessary to infer the true underlying spiking activity from fluorescence measurements. Bayesian model inversion can be used to solve this problem, but typically requires either computationally expensive MCMC sampling, or faster but approximate maximum-a-posteriori optimization. Here, we introduce a flexible algorithmic framework for fast, efficient and accurate extraction of neural spikes from imaging data. Using the framework of variational autoencoders, we propose to amortize inference by training a deep neural network to perform model inversion efficiently.

This paper presents a variational auto-encoder-style method for extracting spike-trains from two-photon fluorescence microscopy time traces. This is a task that scientists are interested in for the purposes of interpreting one- two- or three- photon microscopy data, which are becoming a staple method in neuroscience. The authors introduce 3 different spike-to-fluorescence models and use one of two neural networks to create an decoder to recover the (discritized) spike trains from the fluorescence. Overall, the method does show some promise. That said, my biggest concern with this work lies with the characterization of the method.

Calcium imaging permits optical measurement of neural activity. Since intracellular calcium concentration is an indirect measurement of neural activity, computational tools are necessary to infer the true underlying spiking activity from fluorescence measurements. Bayesian model inversion can be used to solve this problem, but typically requires either computationally expensive MCMC sampling, or faster but approximate maximum-a-posteriori optimization. Here, we introduce a flexible algorithmic framework for fast, efficient and accurate extraction of neural spikes from imaging data. Using the framework of variational autoencoders, we propose to amortize inference by training a deep neural network to perform model inversion efficiently.