Deconvolving Passive Diffusion on the Structural Network from Functional Brain Signals

Motivation: Neuroscience seeks to investigate active signaling in the brain. However, since macroscopic brain signals spread through the white matter structural network, signal measurements will mix passive network diffusion with active signaling. Goal(s): Here, we use a Network Diffusion Model to estimate how much signal is related to passive diffusion through the network, then we deconvolve that signal to recover a regional "innovation" signal. Approach: We apply Network Diffusion Deconvolution (NDD) to task functional MRI data to assess whether brain regional power becomes more localized to task-related regions. Results: We found that NDD significantly amplified expected task-related fMRI signals and suppressed diffuse signals. Impact: Signal passively spreading through the brain's white matter network mixes with active "innovation" signals in fMRI time series. Deconvolving this passive signal reveals a task-relevant innovation signal, which may lead to novel interpretations of task-related fMRI activity not possible previously.