Functional magnetic resonance imaging measures brain activity indirectly by monitoring changes in blood oxygenation levels, known as the blood-oxygenation-level-dependent (BOLD) signal, rather than directly measuring neuronal activity. This approach crucially relies on neurovascular coupling, the mechanism that links neuronal activity to changes in cerebral blood flow.

Here we found that about 40% of voxels with significant BOLD signal changes during various tasks showed reversed oxygen metabolism, particularly in the default mode network.

A voxel is a cube of a 3D brain scan and are used to determine increases and decreases in blood oxygenation and therefore oxygen metabolism, suggesting brain activity or inactivity. (I think I got that right.)

What they seem to have found is that change in blood oxygenation as measured by fMRI does not always track changes in oxygen metabolism, because the way oxygen is used in the brain varies in more ways than the BOLD signal assumes.

Neuronal activity is the primary energy consumer in the brain, driven by oxygen metabolism and quantified as the cerebral metabolic rate of oxygen (CMRO2). Functional magnetic resonance imaging (fMRI) maps this activity indirectly by detecting regional changes in blood oxygenation. The resulting blood-oxygenation-level-dependent (BOLD) signal originates from fluctuations in deoxygenated hemoglobin, rather than from neuronal activity itself.

The BOLD signal itself reflects a complex interplay among changes in CBF, cerebral blood volume (CBV) and the oxygen extraction fraction (OEF) during capillary passage, making its interpretation region-dependent. Consequently, various studies have reported inconsistencies between BOLD signal responses and cognitive or neuronal activity in humans.

They find BOLD signals are especially misleading when a region of the brain extracts more oxygen from blood without the large blood flow that we'd expect. They conclude that it's a unreliable means of measuring actual brain activity.

In this study, we used both quantitative and conventional BOLD imaging to test the hypothesis that ∆BOLD would not reliably reflect changes in oxygen metabolism throughout the entire cortex. We found that in a substantial fraction of voxels with significant BOLD responses, oxygen metabolism changes in the opposite direction to both positive and negative BOLD signals. Notably, these discordant voxels regulated oxygen demand primarily via changes in OEF, rather than CBF. These findings challenge the canonical hemodynamic response model, demonstrating that ∆BOLD alone can lead to misleading interpretations of underlying neuronal activity.