Mapping effective connectivity in the human brain with concurrent intracranial electrical stimulation and BOLD-fMRI
Introduction
Large-scale brain connectivity is a key aspect of brain function, providing important insights into cognition and behavior, in both health and disease. Large databases such as those from the Human Connectome Project are providing detailed maps of structural and functional connectivity in the human brain (Van Essen and Barch, 2015). Structural connectivity is estimated from diffusion-weighted MRI, but is limited to relatively macroscopic white matter connections, has fundamental limitations in resolution (Thomas et al., 2014), and of course ignores function. The distributed functional networks of the human brain, typically obtained from resting-state-fMRI, show reliable large-scale organization that is thought to subserve aspects of emotion, attention, and cognition (Andrews-Hanna et al., 2010, Buckner et al., 2008, Simmons and Martin, 2012, Yeo et al., 2011), and that are altered in many psychiatric and neurodegenerative diseases (Fox et al., 2014, Greicius, 2008, Greicius and Kimmel, 2012, Greicius et al., 2004, Zhang and Raichle, 2010). Such functional networks reveal a complex structure, often featuring anatomical hubs that can serve to organize communication within the network (van den Heuvel and Sporns, 2013), showing dynamic shifts over time (Allen et al., 2014), and opening the possibility of finding functional variations that correlate with individual differences, including psychopathology (Smith et al., 2013).
Functional brain connectivity is typically estimated from fMRI data in two ways (Friston, 2011). By far the most common depends on covariance in resting-state fluctuations of BOLD time series that are not induced by an experimental stimulus (Biswal et al., 2010). This type of functional connectivity analysis (resting-state fMRI connectivity) has revealed reliable large-scale brain networks, but cannot distinguish true causal coupling from common drivers. For instance, disparate brain regions may be correlated due to interaction with the external environment, as well as somatic signals such as respiration or heart rate. Another source of network-level correlation may arise from complex large-scale dynamics that are not well reflected in direct anatomical connections (Honey et al., 2009). Effective connectivity, by contrast, does aim to estimate actual causal relationships. One way of approaching such an estimation is by temporal precedence. Granger causality capitalizes on this feature, but unfortunately it is insufficient to establish true causality, since temporal precedence is still possible with more complex common drivers. Dynamic causal modeling (Daunizeau et al., 2011, Penny, 2012) does yield evidence for causal models but depends on specifying models in the first place, requires task-driven fMRI to manipulate activation at nodes in those models, does not scale well to large-scale connectivity, and rests on a number of assumptions.
Unambiguous isolation of the causal influence that one region has on another can be achieved by applying focal external perturbation (such as direct electrical stimulation) at a specific node (or multiple nodes) while concurrently recording responses from rest of the brain. Clinically, electrical stimulation has been successfully used to treat movement disorders, and has more recently been applied also to treat psychiatric disorders, such as depression. However, the mediating effects of such stimulation on the rest of the brain are at present poorly understood, making the rational development of such interventions challenging. Combining these clinical applications of deep-brain stimulation with the wide field-of-view of fMRI would yield important new information on the acute effects that local brain stimulation has on the rest of the brain.
Application of concurrent electrical stimulation and BOLD fMRI (es-fMRI) in humans requires a unique setting and careful safety testing. In this paper, we report on the safety, feasibility, experimental setup and scientific promise of es-fMRI as applied in a series of epilepsy patients with depth electrodes.
Section snippets
Safety testing with a gel phantom
Our approach poses poorly understood risks to research participants. We therefore first characterized in detail the safety of concurrent electrical stimulation and fMRI, which allowed us to converge on a range of physiologically feasible parameters to use in our research protocol. Safety concerns derive from three effects: (1) induction of currents in the electrode from time-varying magnetic field gradients produced by the scanner; (2) heating of the electrodes due to RF energy deposition from
Heating of the electrodes
Results of temperature measurement of the electrodes during MRI scanning at 3 T are presented in Fig. 2A for three different scanning sequences (GE-EPI, MPRAGE and TSE). For GE-EPI with head transmit coil, fifty-six temperature measurements were obtained. The maximum temperature rise was 0.78 °C. The median value of maximum temperature rise was 0.17 °C with inter-quartile range of 0.13 °C. For MPRAGE with head transmit coil, nine temperature measurements were obtained. The maximum temperature rise
Discussion
In this report, we document the results of safety testing and initial observation of BOLD activity induced by direct electrical stimulation of the brain through depth electrodes in humans.
We established the safety and feasibility of es-fMRI through extensive studies in a gel phantom followed by in vivo studies in sixteen neurosurgical patients. No complications were encountered in any case. We converged on a protocol using a bi-phasic charge-balanced pulse of peak amplitude of 12 mA at 100 Hz, 9
Acknowledgements
Support for this work was provided by the National Institute on Deafness and Other Communication Disorders (R01-DC04290), National Center for Research Resources (UL1RR024979), and an NIMH Conte Center (P50MH094258). We thank Haiming Chen, Phillip Gander, Rick Reale for assistance with conducting the experiments, and Jeremy Greenlee, Goldie Boone, Julien Dubois, Swaloop Guntapalli and Tim Tierney for helpful discussion. We thank our patients for their contribution.
References (39)
- et al.
Functional-anatomic fractionation of the brain's default network
Neuron
(2010) - et al.
Functional MRI with active, fully implanted, deep brain stimulation systems: safety and experimental confounds
Neuroimage
(2007) - et al.
Dynamic causal modelling: a critical review of the biophysical and statistical foundations
Neuroimage
(2011) - et al.
Sequence of information processing for emotions based on the anatomic dialogue between prefrontal cortex and amygdala
Neuroimage
(2007) - et al.
Contributions of the amygdala to reward expectancy and choice signals in human prefrontal cortex
Neuron
(2007) - et al.
Ventromedial prefrontal cortex is critical for the regulation of amygdala activity in humans
Biol. Psychiatry
(2015) Comparing dynamic causal models using AIC, BIC and free energy
Neuroimage
(2012)- et al.
Functional connectomics from resting-state fMRI
Trends Cogn. Sci.
(2013) - et al.
FIACH: a biophysical model for automatic retrospective noise control in fMRI
Neuroimage
(2016) - et al.
Mapping cortical activity elicited with electrical microstimulation using FMRI in the macaque
Neuron
(2005)