7T MRI maps submillimeter hubs in risky decisions

A risky choice can turn on a brain structure only 2 mm across. In a collaboration between National Taiwan University and the University of Illinois Urbana-Champaign, researchers used 7-Tesla MRI to watch living participants process stakes, make decisions, and assess outcomes in a specialized task.

The higher-field scanner opened a window that conventional imaging leaves blurred. The researchers describe 3-Tesla MRI as resolving neural activity at roughly 3 mm, enough to distinguish broad brain areas but not the smaller functional hubs involved in a decision. At 7T, the team examined structures ranging from hundreds of microns to 1 millimeter.

The map separates jobs that had often appeared blended together. When participants encountered salient, high stakes, the locus coeruleus signaled the rest of the brain—especially the frontal cortex—to allocate more neural resources. In the 3 mm-thick frontal cortex, activity shifted between deeper layers while a choice was formed and superficial layers while its result was evaluated. The VTA/SN and nucleus accumbens then differentiated gains from losses, helping update future behavior.

One of the clearest findings concerned gray matter bridges, or GMBs, structures about 1 mm wide that span the striatum between the caudate and putamen. Their activity was higher during faster decisions. The pattern suggests that these bridges may act as information gates, signaling when accumulated evidence is sufficient to pass downstream toward action in the putamen, or when further processing is required upstream in the caudate or cortex.

So what changes, concretely? This is not yet a treatment or a clinical test, but it gives researchers a finer map for looking at decision-making breakdown. Because deterioration may begin in discrete neural microcircuits rather than whole brain regions, the team says the method could help detect individual vulnerabilities earlier and target future interventions more precisely. The researchers also acquired structural and functional images simultaneously, limiting the misalignment that can undermine submillimeter localization.