Accurate spike estimation from noisy calcium signals for ultrafast three-dimensional imaging of large neuronal populations in vivo

Abstract

Extracting neuronal spiking activity from large-scale two-photon recordings remains challenging, especially in mammals in vivo, where large noises often contaminate the signals. We propose a method, MLspike, which returns the most likely spike train underlying the measured calcium fluorescence. It relies on a physiological model including baseline fluctuations and distinct nonlinearities for synthetic and genetically encoded indicators. Model parameters can be either provided by the user or estimated from the data themselves. MLspike is computationally efficient thanks to its original discretization of probability representations; moreover, it can also return spike probabilities or samples. Benchmarked on extensive simulations and real data from seven different preparations, it outperformed state-of-the-art algorithms. Combined with the finding obtained from systematic data investigation (noise level, spiking rate and so on) that photonic noise is not necessarily the main limiting factor, our method allows spike extraction from large-scale recordings, as demonstrated on acousto-optical three-dimensional recordings of over 1,000 neurons in vivo.

In Vivo Imaging of Dentate Gyrus Mossy Cells in Behaving Mice

“…2B-C). To estimate activity in both populations in a manner independent of our transient detection procedure, we also estimated the spike trains underlying the fluorescence signals (Deneux et al, 2016). This analysis further supported a highly significant difference in spiking activity between iMCs and GCs (Fig S1).…”

Long-Term Optical Access to an Estimated One Million Neurons in the Live Mouse Cortex

“…To sort cells from one-photon Ca 2+ videos, we used a method based on maximum likelihood maximization (Deneux et al, 2016; Harris et al, 2016). To sort cells from two-photon Ca 2+ movies, we used the constrained non-negative matrix factorization (CNMF) cell-sorting algorithm (Pnevmatikakis et al, 2016).…”

CosMIC: A Consistent Metric for Spike Inference from Calcium Imaging

“…To calculate the bound, knowledge of the calcium transient pulse parameters and the standard deviation of the noise are required. These parameters are typically used by algorithms in the spike detection process (Vogelstein et al, 2010;Deneux et al, 2016).…”

“…Spike inference performance has been assessed using true and false positive rates (Rahmati et al, 2016), precision and recall analysis (Reynolds et al, 2017) and using the complement of the success rate, the error rate (Deneux et al, 2016). We study this class of metrics under the umbrella of the success rate, which we define here.…”

“…Rather, it reads out a relatively reliable indicator of spiking activity -a cell's intracellular calcium concentration -from which spike times must be inferred. A diverse array of techniques have been proposed for this task, including deconvolution approaches (Vogelstein et al, 2010;Friedrich et al, 2017;Pachitariu et al, 2017), methods that identify the most likely spike train given a signal model (Vogelstein et al, 2009;Deneux et al, 2016) and approaches that exploit the sparsity of the underlying spike train (Oñativia et al, 2013). To enable the investigation of neural coding hypotheses, reconstructed spike trains must have sufficient temporal precision for analysis of synchrony between neurons and behavioural variables (Huber et al, 2012), whilst accurately inferring the rate of spiking activity.…”