Voxel-based state space modeling recovers task-related cognitive states in naturalistic fMRI experiments

A voxel-based state space modeling method that recovers low-dimensional, task-related cognitive state spaces from human fMRI data recorded during a visual attention task and a video game task.

Human neuroimaging has largely focused on static representations, yet real-world behavior is continuous and time-varying, and electrophysiology suggests that dynamic task variables occupy a low-dimensional subspace within the activity of neural populations. We developed a voxel-based state space modeling method that recovers task-related state spaces from human fMRI data and applied it to a visual attention task and a video game task. Each task induces distinct brain states that embed in a low-dimensional state space capturing task parameters, and attention increases the separation between states within the task-related subspace. These results demonstrate that the state space framework offers a powerful approach for modeling brain activity elicited by complex natural tasks.