D2 Functional magnetic resonance imaging (fMRI, BOLD signal)

Topic

Functional magnetic resonance imaging (fMRI) indirectly measures neuronal activity via a signal that depends on blood oxygen levels: when a brain region becomes active, the proportion of oxygen in the blood supplying it changes, and this change is what fMRI detects as the blood-oxygen-level-dependent (BOLD) signal. The complete formulation of how neuronal activity translates into this BOLD signal was introduced in the neurovascular coupling model, of which this signal is a direct extension.

Standard statistical analysis of the BOLD signal employs a general linear model: the signal's time series, observed in each voxel, is explained as a linear combination of regressors derived from the experimental design, convolved with the canonical hemodynamic response function. This general linear model allows for the generation of statistical activation maps based on that combination, showing the voxels where the BOLD signal is best explained by the experimental design.

The quality of the BOLD signal acquisition is quantified by the temporal signal-to-noise ratio: the higher this ratio, the more reliable the BOLD signal recorded in each voxel.

In resting-state studies, where there is no task-based experimental design to generate regressors, functional connectivity between different brain regions is estimated differently: by correlating the BOLD signal time series recorded in each of those regions. The more highly correlated the BOLD signal time series of two regions are, the greater the estimated functional connectivity between them. This functional connectivity, estimated from the blood-oxygen-level-dependent signal during resting-state studies, is a widely used measure for characterizing network alterations in various pathologies described in other models within this same catalog—such as schizophrenia and autism spectrum disorder. Ultimately, it is this blood-oxygen-level-dependent signal—measured indirectly via functional magnetic resonance imaging—that enables both the generation of statistical activation maps and the estimation of the functional connectivity characterizing these pathologies.

ID:4298