Effective and Efficient Neural Networks for Spike Inference fromIn VivoCalcium Imaging

Abstract

Calcium imaging technique provides irreplaceable advantages in monitoring large population of neuronal activities simultaneously. However, due to the generally low signal to noise ratio (SNR) of the calcium signal and variability in dye properties, it is still challenging to faithfully infer neuronal spikes from these calcium signals, especially from in vivo experiments. In this study, we tackled the problem of both spike-rate and spike-event predictions using a data-driven approach, based on a public pool of dataset with simultaneously recorded calcium and electrophysiological signals using different dyes and recorded from different brain regions. We proposed the ENS2 (effective and efficient neural networks for spike inference from calcium signals) system using raw calcium inputs and it consistently outperforms state-of-the-arts algorithms in both spike-rate and spike-event predictions with reduced computational load. We have also demonstrated that factors such as sampling rates, smoothing window sizes and parametric evaluation metrics could readily bias the interpretation of inference performance. We concluded that optimizing our system for spike-event prediction could produce a more versatile inference system for real neuroscience studies.

Movement-specific signaling is differentially distributed across motor cortex layer 5 projection neuron classes

“…Although we found that the majority of layer 5B neuron signaling was movement invariant, a relatively small proportion of neurons displayed response bias toward push or pull movements. The relatively low level of movement-specific signaling is unlikely to be due to masking of subtle changes in spike rate when using calcium reporters ( Wei et al., 2020 ; Zhou and Tin, 2021 ) because we observed similar proportions of movement-specific signaling when performing high-density extracellular recordings of putative layer 5B projection neurons. The firing rates of individual neurons in motor cortex reflect a complex combination of signals that correlate with joint angle, direction, and speed ( Georgopoulos et al., 1982 ; Moran and Schwartz, 1999 ; Paninski et al., 2004 ; Thach, 1978 ), whereas population dynamics reflect time-varying changes in neural state during the transition from rest to movement execution ( Churchland et al., 2010 , 2012 ; Kaufman et al., 2014 , 2016 ; Kurtzer et al., 2005 ; Sauerbrei et al., 2019 ).…”

“…In contrast to primate motor cortex, we found widespread movement-invariant responses at the single-neuron level ( Kaufman et al., 2016 ). This is unlikely to reflect differences in recording sensitivity, given that our imaging and electrophysiology approaches identified similar proportions of movement-invariant neurons across layer 5B ( Wei et al., 2020 ; Zhou and Tin, 2021 ), or the limited number of movements in our task because movement-invariant responses have been shown in relatively simple tasks requiring few actions ( Evarts, 1968 ; Hocherman and Wise, 1991 ; Messier and Kalaska, 2000 ; Riehle et al., 1994 ; Weinrich et al., 1984 ) and in complex tasks involving more than 20 separate actions ( Kaufman et al., 2016 ). Instead, this might reflect evolutionary differences in how motor cortex recruits and controls muscle activation during the transition from rest to movement execution.…”