Neuronal synchrony and the relation between the blood-oxygen-level dependent response and the local field potential

The most widespread measures of human brain activity are the blood-oxygen-level dependent (BOLD) signal and surface field potential. Prior studies report a variety of relationships between these signals. To develop an understanding of how to interpret these signals and the relationship between them, we developed a model of (a) neuronal population responses and (b) transformations from neuronal responses into the functional magnetic resonance imaging (fMRI) BOLD signal and electrocorticographic (ECoG) field potential. Rather than seeking a transformation between the two measures directly, this approach interprets each measure with respect to the underlying neuronal population responses. This model accounts for the relationship between BOLD and ECoG data from human visual cortex in V1, V2, and V3, with the model predictions and data matching in three ways: across stimuli, the BOLD amplitude and ECoG broadband power were positively correlated, the BOLD amplitude and alpha power (8–13 Hz) were negatively correlated, and the BOLD amplitude and narrowband gamma power (30–80 Hz) were uncorrelated. The two measures provide complementary information about human brain activity, and we infer that features of the field potential that are uncorrelated with BOLD arise largely from changes in synchrony, rather than level, of neuronal activity.Author summary: There are several methods for measuring activity in the living human brain. Here, we studied functional magnetic resonance imaging (fMRI), which depends on the vascular response to neuronal activity, and surface field potentials, which measure electrical activity from many neurons. These two widely used measurements of human brain activity often provide different and potentially conflicting results. We propose a quantitative model for how these two measurements integrate activity from neuronal populations. The fMRI signal is highly sensitive to the average level of local neuronal activity but not the degree of synchrony between neurons. In contrast, the field potential is most sensitive to synchronous neuronal signals. Our model accounts for several observations seen in fMRI and field potential data: some very large features of field potential recordings, such as gamma oscillations, can occur with little to no associated fMRI signal. The model predicts this because the gamma oscillations result more from increased neuronal synchrony than increased neuronal activity. Other field potential signals, such as broadband changes, which are likely driven by the level of neuronal activity rather than a change in synchrony, are highly correlated with fMRI. The two measures thus provide complementary information about human brain activity.