Synopsis

Keywords: Contrast mechanisms: fMRI, Neuro: Brain function

Because BOLD-fMRI probes neuronal activity indirectly and with a lag of a few seconds, based on neurovascular coupling mechanisms, several studies have attempted to clarify its neuronal correlates in humans by combining it with the simultaneous recording of the electroencephalogram (EEG). Like other electrophysiology techniques, EEG provides direct measures of neuronal activity with sub-millisecond temporal resolution, albeit poorer spatial resolution and coverage than BOLD-fMRI. In this talk, I will overview the main characteristics of electrophysiology relative to BOLD-fMRI as well as the evidence contributed by EEG-fMRI studies towards our understanding of the neuronal correlates of different types of BOLD-fMRI measurements.

In contrast to the indirect measures of neuronal activity obtained with BOLD fMRI, noninvasive electrophysiology techniques commonly used in humans such as electroencephalography (EEG) and magnetoencephalography (MEG) directly measure the electrical activity of neuronal populations (1). Moreover, in contrast to the slow haemodynamic response captured by BOLD signals, lagging neuronal activity by a few seconds, electrophysiology techniques can assess neuronal dynamics with sub-millisecond temporal resolution. On the other hand, the homogeneous full brain coverage allowed by BOLD fMRI and its excellent spatial localisation power, down to sub-millimetre resolution, are unbeatable when compared with noninvasive electrophysiology techniques.

All combined, the high complementarity between the EEG and fMRI has motivated their multimodal integration. In particular, their simultaneous recording is required to study spontaneous neuronal fluctuations during rest or trial-by-trial variations in evoked activity due to ongoing neuronal fluctuations. Since the first study by Bonmassar in 1999 (2), various methodological challenges including the severe artefacts induced on the EEG in the MRI environment have been addressed (3), and a growing body of literature has provided evidence for the electrophysiology correlates of different aspects of BOLD fMRI measurements (4).

Biophysical models have been proposed for the integration of the two types of signal, based on the relationship between the BOLD signal and local field potentials measured using intracortical recordings in monkeys in the seminal experiment by Logothetis in 2001 (5). Simpler heuristics have been derived to postulate relationships between the BOLD signal and noninvasive EEG features such as the total power, the power in specific frequency bands, or different measures of the power-weighted mean frequency (4). Empirically, several human studies have reported relationships between the BOLD signal and EEG band power, partly supporting the idea that BOLD correlates positively with higher EEG frequencies (e.g., gamma band) and negatively with lower EEG frequencies (e.g., alpha band). This idea is elegantly summarised in the heuristic proposed by Kilner in 2005 (6), whereby the BOLD signal elicited by neuronal metabolic activity would vary with the EEG spectral density.

Despite some consistent correlations with EEG power metrics, the multitude of BOLD measurements ranging from evoked changes to resting-state fluctuations and from regional activation to connectivity across whole-brain networks has revealed a more complex landscape for the relationship between BOLD signals and the underlying electrophysiology. This goes well beyond EEG spectral features to exploit also phase synchronization functional connectivity measures and microstates, among others (7–9).

In this talk, I will overview the main characteristics of electrophysiology measurements relative to BOLD-fMRI as well as the evidence contributed by EEG-fMRI studies towards our understanding of the neuronal correlates of BOLD-fMRI measurements.

Acknowledgements

Portuguese Science Foundation (FCT) through Grant LARSyS UID/EEA/50009/2019.