Since the first demonstration of BOLD fMRI, systems neuroscience has been revolutionised by our ability to non-invasively map functionality in the human brain. BOLD has elucidated the structure and function of distributed networks and allowed assessment of functional connectivity between spatially separate regions. However, despite its wide utility, the relationship between BOLD and the underlying electrical activity is still under debate. A number of studies address neurovascular coupling, and advances in multi-modal imaging are allowing increasingly complex questions to be posed, resulting in a wealth of new data. In light of such progress, here we revisit the question; What drives the BOLD response?
First, we question the role of neural oscillations and their relationship with haemodynamics. Investigations of scalp based (E/MEG) and invasive (ECoG) measures of oscillations describe the co-expression of finite, discrete frequency bands spanning 0.05 to 500 Hz. Parametric modulation of frequency specific oscillatory power has been related to cortical processing and, for low level sensory stimuli, has shown a good spatial correspondence with BOLD fMRI. On this basis, the powers in several frequency bands have been proposed as strong candidates for the signal driving the BOLD response. However, results for task-related amplitude correlation between neural oscillations and haemodynamics are somewhat inconsistent, and more variable with the move to more cognitively demanding tasks studying multiple brain regions.
Second, we ask whether phase-locked evoked components of neural activity from E/MEG recordings show significant correlation with BOLD. The results to date are again variable, with amplitude and latency of evoked components exhibiting both positive and negative correlation with BOLD depending on the task. It is noteworthy that many analysis techniques in E/MEG are tuned to either evoked (e.g. minimum norm) or induced (e.g. beamforming) responses, meaning that many studies only examine aspects of either evoked or oscillatory responses to compare to BOLD, leaving the question open as to which of these more closely correlates.
Finally, we ask what information can be gained from multi-modal connectivity measurements? Synchronisation of oscillatory activity between regions has long been posed as a mechanism for long range interaction. The recent shift in fMRI towards a brain-wide view of inter-regional connectivity shows that networks drive correlated BOLD fluctuations. New approaches to E/MEG data analysis have enabled the identification of corresponding electrodynamic networks involving oscillatory rhythms. However, the mechanisms underlying these disparate phenomena remain unclear.
We challenge the multi-modal neuroimaging community to put forward their latest data addressing the question of how electrophysiological metrics relate to BOLD fMRI, and how this advances our understanding of brain function. Specifically we ask the questions: After comparing multiple metrics of neural activity, which best explains the most variance of the BOLD signal (for example, across trials, conditions, or subjects)? To what extent is BOLD confounded by vascular non-linearity? To what extent are electrophysiological metrics confounded by analysis techniques? And finally, which aspect of E/MEG or BOLD activity best explains the true cognitive processing (i.e. perception or behaviour)?
Since the first demonstration of BOLD fMRI, systems neuroscience has been revolutionised by our ability to non-invasively map functionality in the human brain. BOLD has elucidated the structure and function of distributed networks and allowed assessment of functional connectivity between spatially separate regions. However, despite its wide utility, the relationship between BOLD and the underlying electrical activity is still under debate. A number of studies address neurovascular coupling, and advances in multi-modal imaging are allowing increasingly complex questions to be posed, resulting in a wealth of new data. In light of such progress, here we revisit the question; What drives the BOLD response?
First, we question the role of neural oscillations and their relationship with haemodynamics. Investigations of scalp based (E/MEG) and invasive (ECoG) measures of oscillations describe the co-expression of finite, discrete frequency bands spanning 0.05 to 500 Hz. Parametric modulation of frequency specific oscillatory power has been related to cortical processing and, for low level sensory stimuli, has shown a good spatial correspondence with BOLD fMRI. On this basis, the powers in several frequency bands have been proposed as strong candidates for the signal driving the BOLD response. However, results for task-related amplitude correlation between neural oscillations and haemodynamics are somewhat inconsistent, and more variable with the move to more cognitively demanding tasks studying multiple brain regions.
Second, we ask whether phase-locked evoked components of neural activity from E/MEG recordings show significant correlation with BOLD. The results to date are again variable, with amplitude and latency of evoked components exhibiting both positive and negative correlation with BOLD depending on the task. It is noteworthy that many analysis techniques in E/MEG are tuned to either evoked (e.g. minimum norm) or induced (e.g. beamforming) responses, meaning that many studies only examine aspects of either evoked or oscillatory responses to compare to BOLD, leaving the question open as to which of these more closely correlates.
Finally, we ask what information can be gained from multi-modal connectivity measurements? Synchronisation of oscillatory activity between regions has long been posed as a mechanism for long range interaction. The recent shift in fMRI towards a brain-wide view of inter-regional connectivity shows that networks drive correlated BOLD fluctuations. New approaches to E/MEG data analysis have enabled the identification of corresponding electrodynamic networks involving oscillatory rhythms. However, the mechanisms underlying these disparate phenomena remain unclear.
We challenge the multi-modal neuroimaging community to put forward their latest data addressing the question of how electrophysiological metrics relate to BOLD fMRI, and how this advances our understanding of brain function. Specifically we ask the questions: After comparing multiple metrics of neural activity, which best explains the most variance of the BOLD signal (for example, across trials, conditions, or subjects)? To what extent is BOLD confounded by vascular non-linearity? To what extent are electrophysiological metrics confounded by analysis techniques? And finally, which aspect of E/MEG or BOLD activity best explains the true cognitive processing (i.e. perception or behaviour)?