Temporal dynamics and neural assemblies underlying fMRI BOLD signals

Determine the precise temporal dynamics of neuronal activity and identify the exact neural assemblies underlying the macroscopic fMRI BOLD signals recorded during naturalistic video watching, which remain unresolved when using only fMRI data in the TRIBE framework.

Background

TRIBE is an end-to-end, multimodal (text, audio, video), multisubject, nonlinear encoder trained to predict whole-brain fMRI responses during naturalistic video viewing. The approach operates on fMRI, whose BOLD signal is an indirect, temporally sluggish measure of neuronal activity aggregated over macroscopic spatial scales.

Because of these intrinsic properties of fMRI, the work acknowledges that it does not resolve which precise neural assemblies give rise to the observed macroscopic BOLD fluctuations, nor their exact temporal dynamics. Clarifying these aspects would require information beyond what can be inferred from fMRI alone.