Paper published on data assimilation for cortical activity during sleep

A new paper by ACDC member Daniele Avitabile, with Gabriel J. Lord and Khadija Meddouni (Radboud University, Nijmegen), has recently been published in the Journal of Mathematical Biology (DOI, arXiv).

The authors combine a Wilson–Cowan–Amari neural field model with a nested particle filter, a Bayesian data assimilation method that estimates the hidden state of the model and its parameters at the same time from noisy measurements. They first test the approach on synthetic data, then apply it to local field potential recordings from the cortex of a sleeping mouse, reduced from two to one spatial dimension with the Radon transform. The inferred parameters of the external forcing track the amplitude and the spatial and temporal frequencies of the recorded signal across different sleep stages, while the estimated state reproduces its dominant rhythms.

The work is a step towards numerical tools for studying how the cortex interacts with hidden, unobserved brain regions during sleep, when memories are thought to consolidate.