RADICaL 1p: Deep inference of neural population dynamics from one-photon calcium imaging
Author(s)
Ke, Jingyang
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
One-photon (1p) calcium imaging is a valuable tool for studying large-scale neural activity in freely moving animals, but accurately inferring the underlying neural activity is challenging due to the low-pass filtering nonlinear transformation with strong noise in recording. Recently, a deep learning framework Recurrent Autoencoder for Discovering Imaged Calcium Latents (RADICaL) has shown promising performance in inferring calcium neural event rates in two-photon (2p) calcium imaging data using a zero-inflated gamma (ZIG) observation model that captures the dynamics of deconvolved calcium signals. This project RADICaL 1p extends the RADICaL framework application to 1p calcium imaging and demonstrates its superior performance compared to the Gaussian smoothing method in neural decoding experiments on 1p calcium imaging neural recording in a mouse’s dorsolateral striatum (DLS) and synchronized behavior data. RADICaL 1p has the potential to enhance a variety of neuroscience research topics involving 1p calcium imaging, particularly those concerning the relationship between large-scale neural activity and the behavior of freely moving animals.
Sponsor
Date
Extent
Resource Type
Text
Resource Subtype
Undergraduate Research Option Thesis