Resting-state functional-connectivity MRI has identified abnormalities in patients with a wide range of neuropsychiatric disorders, including epilepsy due to malformations of cortical development. Transcranial Magnetic Stimulation in combination with EEG can demonstrate that patients with epilepsy have cortical hyperexcitability in regions with abnormal connectivity.
Resting-state functional connectivity MRI (rs-fcMRI) is a technique that identifies connectivity between different brain regions based on correlations over time in the blood-oxygenation level dependent signal. rs-fcMRI has been applied extensively to identify abnormalities in brain connectivity in different neurologic and psychiatric diseases. However, the relationship among rs-fcMRI connectivity abnormalities, brain electrophysiology and disease state is unknown, in part because the causal significance of alterations in functional connectivity in disease pathophysiology has not been established. Transcranial Magnetic Stimulation (TMS) is a technique that uses electromagnetic induction to noninvasively produce focal changes in cortical activity. When combined with electroencephalography (EEG), TMS can be used to assess the brain’s response to external perturbations. Here we provide a protocol for combining rs-fcMRI, TMS and EEG to assess the physiologic significance of alterations in functional connectivity in patients with neuropsychiatric disease. We provide representative results from a previously published study in which rs-fcMRI was used to identify regions with abnormal connectivity in patients with epilepsy due to a malformation of cortical development, periventricular nodular heterotopia (PNH). Stimulation in patients with epilepsy resulted in abnormal TMS-evoked EEG activity relative to stimulation of the same sites in matched healthy control patients, with an abnormal increase in the late component of the TMS-evoked potential, consistent with cortical hyperexcitability. This abnormality was specific to regions with abnormal resting-state functional connectivity. Electrical source analysis in a subject with previously recorded seizures demonstrated that the origin of the abnormal TMS-evoked activity co-localized with the seizure-onset zone, suggesting the presence of an epileptogenic circuit. These results demonstrate how rs-fcMRI, TMS and EEG can be utilized together to identify and understand the physiological significance of abnormal brain connectivity in human diseases.
Transcranial magnetic stimulation (TMS) is a means of noninvasively stimulating regions of cortex via electromagnetic induction. In TMS, a large but spatially restricted magnetic flux is used to induce an electrical field in a target cortical area, and thereby modulate the activity of the underlying neural tissue. TMS to motor cortex results in motor evoked potentials that can be measured peripherally via electromyography (EMG). When applied in pairs or triplets of pulses, TMS can be used to assess the activity of specific intracortical GABAergic and glutaminergic circuits1-3, and thus assess the balance of excitation and inhibition in vivo in human patients. In epilepsy specifically, TMS studies have shown that cortical hyperexcitability is present in patients with epilepsy4,5, and may normalize with successful anti-epileptic drug therapy and thus predict response to medication6. Furthermore, TMS measures of cortical excitability show intermediate values in patients with a single seizure7 and in siblings of patients with both idiopathic generalized and acquired focal epilepsies8. These findings suggest that TMS measures of cortical excitability may allow us to identify endophenotypes for epilepsy. However, the sensitivity and specificity of these measures are limited, likely because TMS-EMG can only be assessed with stimulation of motor cortical circuits, and many patients with epilepsy have seizure foci outside the motor cortex.
Electroencephalography (EEG) provides an opportunity to directly measure the cerebral response to TMS, and can be used to assess cerebral reactivity across wide areas of neocortex. Studies integrating TMS with EEG (TMS-EEG) have shown that TMS produces waves of activity that reverberate throughout the cortex9,10 and that are reproducible and reliable11-13. By evaluating the propagation of evoked activity in different behavioral states and in different tasks, TMS-EEG has been used to causally probe the dynamic effective connectivity of human brain networks10,14-16. TMS-EEG measures have shown significant abnormalities in diseases ranging from schizophrenia17 to ADHD18, and in disorders of consciousness such as persistent vegetative state19. Furthermore, several groups have identified EEG correlates of the paired-pulse TMS-EMG metrics that are abnormal in patients with epilepsy20,21. Of particular relevance, previous studies have also suggested that abnormal stimulation-evoked EEG activity is seen in patients with epilepsy22-25.
Another means of evaluating brain circuits is via resting-state functional connectivity MRI (rs-fcMRI), a technique that evaluates the correlations over time in the blood oxygenation level dependent (BOLD) signal from different brain regions26. Studies using rs-fcMRI have demonstrated that the human brain is organized into distinct networks of interacting regions26-29, that neuropsychiatric diseases may occur within specific large-scale distributed neural networks identified by rs-fcMRI30, and that the brain networks identified via rs-fcMRI are often abnormal in neuropsychiatric disease states31,32. In terms of potential clinical applications, rs-fcMRI has several advantages over conventional task-based fMRI application33, including less reliance on subject cooperation and concern over variable performance. Consequently, there has recently been an explosion of studies exploring rs-fcMRI changes in different disease states. However, one of the limitations of rs-fcMRI is the difficulty in determining whether and how correlations (or anticorrelations) in the BOLD signal relate to the electrophysiological interactions that form the basis of neuronal communication. A related problem is that it is often unclear whether the rs-fcMRI changes seen in various disease states have physiologic significance. In particular with regards to epilepsy, it is unclear whether abnormalities in rs-fcMRI are due solely to interictal epileptiform transients, or exist independently of such electrophysiological abnormalities; simultaneous EEG-fMRI is needed to help evaluate between these possibilities34.
As TMS can be used to produce transient or sustained changes in the activations of different cortical regions, TMS studies provide a means of causally assessing the significance of different resting-state fMRI connectivity patterns. One approach is to use rs-fcMRI to guide therapeutic stimulation efforts in different disease states; it could be expected that TMS targeted to regions that are functionally connected to areas known to be involved in different disease states is more likely to be therapeutically effective than TMS targeted to regions without such functional connectivity, and indeed several studies have found preliminary evidence for this35,36. Another approach would involve using TMS-EEG to causally assess the physiologic significance of different resting-state fcMRI patterns. Specifically, one can test the hypothesis that regions that show abnormal functional connectivity in a specific disease state should show a different response to stimulation in patients than in healthy subjects, and that these physiologic abnormalities are present specifically (or primarily) with stimulation of the abnormally connected region.
To illustrate the above, we provide an example of a recent study in which rs-fcMRI, TMS and EEG were combined to explore cortical hyperexcitability in patients with epilepsy due to the developmental brain abnormality periventricular nodular heterotopia (PNH)37. Patients with PNH present clinically with adolescent- or adult-onset epilepsy, reading disability, and normal intelligence, and have abnormal nodules of gray matter adjacent to the lateral ventricles on neuroimaging38,39. Previous studies have shown that these periventricular nodules of heterotopic gray matter are structurally and functionally connected to discrete foci in the neocortex40,41, and that epileptic seizures may originate from neocortical regions, heterotopic gray matter, or both simultaneously42, suggesting that epileptogenesis in these patients is a circuit phenomenon. By using resting-state fc-MRI to guide TMS-EEG, we demonstrated that patients with active epilepsy due to PNH have evidence of cortical hyperexcitability, and that this hyperexcitability appears to be limited to regions with abnormal functional connectivity to the deep nodules.
The protocol is conducted in two separate sessions. During the first session, structural and resting-state blood-oxygenation level-dependent (BOLD) contrast MRI sequences are acquired (for patients), or just structural MRI sequences (for the healthy controls). Between the first and second sessions, resting-state functional connectivity analysis is used to define the cortical targets for the patients, and the MNI coordinates for these targets are obtained. The equivalent cortical targets (based on MNI coordinates) are then identified for each healthy control subject. In the second session, the TMS-EEG data is obtained.
In the example given in this paper, functional-connectivity MRI analyses were performed using an in-house software toolbox and the MRI software43,44. Neuro-navigated TMS was performed with a transcranial magnetic stimulator with real-time MRI neuronavigation. EEG was recorded with a 60-channel TMS-compatible system, which utilizes a sample-and-hold circuit to avoid amplifier saturation by TMS. EEG data were analyzed using custom scripts and the EEGLAB toolbox45 (version 12.0.2.4b) running in MATLAB R2012b.
The protocol described here was approved by the institutional review boards of the Beth Israel Deaconess Medical Center and the Massachusetts Institute of Technology.
1. Subject Selection
2. Generating the Stimulation Targets
3. TMS-EEG Experimental Setup
4. Experimental Session
5. EEG Data Pre-processing and Analysis
NOTE: TMS-EEG data usually contains large stimulation-related artifacts, particularly when stimulating away from the midline/vertex or with high stimulation intensities, and significant preprocessing may be necessary to obtain clean analyzable data. Independent Component Analysis (ICA) is one method that has been utilized for removal of TMS artifacts, and can be applied using publicly available toolboxes (e.g., EEGLAB45) on the MATLAB platform. One validated approach60 is as follows, describing the analysis of data collected using the Eximia EEG system:
6. Assess for Evidence of Cortical Hyperexcitability
7. Source Estimation of Evoked Electrical Activity
Resting-state functional connectivity fMRI can be used to identify regions of cortex that demonstrate high functional connectivity with the heterotopic periventricular gray matter nodules (Figure 1), and control regions without such connectivity. To determine whether such abnormal functional connectivity has physiologic significance, the cortical region with correlated resting-state activity can be chosen as the "connected" target sites for neuronavigated TMS, and the evoked EEG results compared to the EEG potentials produced by stimulation of a control non-connected target in the same patients. Furthermore, the same regions can be targeted in healthy control subjects (Figure 2) to determine whether the abnormal functional connectivity seen in the PNH patients has pathophysiologic significance for the patients' clinical epilepsy syndrome. Specifically, the presence of cortical hyperexcitability can be assessed on the individual patient level by determining the normalized area-under-the-curve of the Global Mean Field Potential, and then evaluating whether this value is larger for the epilepsy patient than his or her matched control (Figure 2). Source localization of the abnormal late peaks in the TMS-evoked potentials in patients with epilepsy can identify the brain regions from which the abnormal activity arises, and may spatially co-localize with the patient's seizure focus (Figure 3).
Figure 1. Resting-state Functional Connectivity and TMS Targets. (A, B) Regions with significant correlations in functional activation (blue/green) to the resting-state BOLD signal in the heterotopic nodules in two patients with periventricular nodular heterotopia and epilepsy. (C, D) The connected target site (red) and the non-connected target site (blue) in these two patients. (Modified with permission from Shafi et al., 201537). Please click here to view a larger version of this figure.
Figure 2. TMS-evoked Potentials and Global Mean Field Potentials. (A) The TMS-evoked potential produced by stimulation of the connected target in a patient with PNH and epilepsy. (B) The TMS-evoked potential produced by stimulation of the same region in the above patient's matched healthy control subject. (C) The global mean field potential (GMFP) produced by stimulation of the connected and non-connected targets for this patient and his matched control. (D) The normalized area-under-the-curve of the GMFP produced by stimulation of the connected and non-connected targets for this subject pair. Please click here to view a larger version of this figure.
Figure 3. Source Localization of TMS-evoked Activity and Seizure Onsets. (A) Electrical source imaging results of a late TMS-evoked peak in a patient with epilepsy; scale is the estimated currents multiplied by 10-11. (B) Electrical source imaging results of a previously captured seizure onset in that same patient. (Modified with permission from Shafi et al, 201537) Please click here to view a larger version of this figure.
Resting-state functional connectivity MRI has been used to identify network connectivity in the human brain, and to identify alterations of connectivity that occur in different disease states26,31,32. However, as fMRI functional connectivity is based on identifying correlations in the BOLD signal, and as blood oxygenation changes have a non-trivial relationship with underlying neural activity, the causal significance and physiological relevance of these fMRI connectivity findings is unclear. TMS enables spatially and temporally targeted manipulations of brain activity in specific cortical regions; when combined with EEG, TMS can be used to assess the brain's response to stimulation across different brain regions. Consequently, TMS-EEG can be applied to regions with altered fMRI functional connectivity to assess whether the observed alterations in connectivity have a physiologic correlate that might relate to the underlying disease pathophysiology.
This article presents a protocol using connectivity-guided TMS-EEG to assess cortical excitability in patients with epilepsy due to a malformation of brain development, periventricular nodular heterotopia, which is associated with the development of abnormal functional connectivity networks37. This protocol is used to demonstrate that patients with active epilepsy have cortical hyperexcitability that is specific to the regions with altered resting-state fMRI functional connectivity, and that hyperexcitability can be assessed on an individual subject level. In a patient with seizures previously captured on EEG, the abnormal late TMS-evoked activity is seen in the same region (distant from the stimulation site) from which the patient's seizures originate, suggesting that the region of abnormal functional connectivity is indeed part of the seizure-generating network.
There are a number of critical steps to successful completion of this protocol. Technical expertise with resting-state fMRI data collection, high-quality resting-state data, and experience with rs-fcMRI data processing and analysis techniques are essential for accurate determination of connectivity-based targets. Another important constraint in designing and executing TMS-EEG studies is the need for TMS-compatible EEG equipment; furthermore, for studies where precise targeting is critical, neuronavigation equipment is also necessary. Another limitation is that TMS often generates substantial EEG artifacts, particularly when stimulating over the frontopolar and lateral temporal regions, and therefore it may be difficult to obtain high-quality data when the stimulation target is located in these regions. The data collection and EEG recording process also needs to be optimized to minimize artifacts in the EEG signal, and experiments should ideally be run by persons familiar with EEG data so that artifacts that do arise (e.g., poor impedance as conducting paste dries) can be rapidly identified and minimized. One important step involves demonstrating the effects of eye blinks, muscle contraction and movement on the EEG to the research subject, as this can be critical in helping the subject to understand and minimize these types of artifacts.
Another important consideration is the minimization of biological artifacts that may limit interpretation of the results. One particularly important such biological artifact is the auditory evoked-potential produced by the TMS coil "click", which is known to contribute to the magnitude of the TMS-evoked potential, particularly at 100 and 180 milliseconds55,67,68 when the TMS-evoked potential is also typically maximal. One method that has been shown to minimize the TMS auditory evoked potential is noise masking via the use of white or colored noise, with the addition of a thin layer of foam between the coil and scalp10,55. In the absence of such noise masking, differences in the TMS auditory-evoked potential could potentially contribute to differences in the evoked activity between subject groups (although not the observed differences in the evoked potentials between different sites within subjects.)
Finally, even if care is taken to optimize the recording, significant preprocessing is often necessary to recover clean data for analysis. Fortunately, validated methods for removing artifacts from TMS-EEG recordings have been published60; however, even with these techniques, recovery of very early signals (< 15 msec) can be very difficult or unreliable. An additional challenge is that EEG data is high-dimensional and complex, and therefore a clear prior hypothesis is often necessary to extract meaningful information. Furthermore, because TMS effects and EEG signals can vary significantly between subjects because of a wide range of non-cerebral factors that are difficult or impossible to control (e.g., skull thickness, skull-cortex distance, concomitant medications, quality of sleep the night prior), outcome measures that are less reliant on the raw magnitude of evoked responses are likely to be more informative or meaningful.
Although technically challenging, the integration of rs-fcMRI, TMS and EEG together in one experiment enables tests of a wide array of hypotheses regarding the significance of specific connectivity findings on cortical physiology. In disease states, these technologies can be integrated together to assess the relationship between fMRI network connectivity changes, pathophysiologic alterations in cortical excitability and evoked brain activity, and disease expression. Notably, this protocol can be used to investigate cortical physiology via common outcome measures even when the focus of abnormal connectivity (and thus the stimulated region) differs from one subject to another, providing an output that may be meaningful at the individual subject level, and opening up the possibility of a personalized approach to investigation and ultimately treatment.
The protocol described in this study could also be expanded to assess specific features of cortical physiology in different subject groups. For example, a number of recent studies have suggested that the N45 component of the TMS-evoked EEG response represents activity of GABA-A receptors69, whereas the N100 component of the TMS-evoked EEG response is a measure of GABA-B mediated inhibition21,69. Paired-pulse TMS-EEG with a long-interval intracortical inhibition protocol provides another measure of GABAergic activity, and has been shown to be altered in frontal regions in patients with schizophrenia relative to controls70. Thus, the above protocol could be modified to specifically address questions regarding GABAergic activity in regions with altered functional connectivity. Additionally, source localization of the peaks in the TMS-evoked potential may identify distant regions that are engaged by stimulation, and thus help inform which of the functional connections identified by conventional resting-state fMRI are capable of causally transmitting evoked activity. For situations in which the key network hubs are deep, rs-fcMRI can also be used to identify cortical targets that are accessible to stimulation, and thereby enable modulation of specific brain networks involved in normal behavior and in disease states35,36,71. In such cases, the techniques described in this study can be used to assess local and distributed single-pulse TMS-evoked activity before and after a repetitive plasticity protocol, to determine whether the plasticity protocol has indeed changed cortical excitability locally, and/or network excitability distally.
In summary, the integration of rs-fcMRI, TMS and EEG enables the exploration of how brain connectivity influences cortical physiology and behavior in human subjects. Moreover, these techniques can also be combined to assess how alterations in connectivity are related to pathophysiology in disease states, as illustrated in the protocol described above.
The authors have nothing to disclose.
The authors would like to thank Emily L. Thorn, B.A., for her assistance with the Source estimation of evoked electrical activity Section. MMS was supported by a KL2/Catalyst Medical Research Investigator Training award from Harvard Catalyst/The Harvard Clinical and Translational Science Center (National Center for Research Resources and the National Center for Advancing Translational Sciences, National Institutes of Health Award KL2 TR001100). CJC was supported by a grant from the National Institutes of Health (5K12NS066225). APL was supported in part by grants from the Sidney R. Baer Jr. Foundation, the National Institutes of Health (R01 HD069776, R01 NS073601, R21 MH099196, R21 NS082870, R21 NS085491, R21 HD07616), and Harvard Catalyst/The Harvard Clinical and Translational Science Center (NCRR and the NCATS, NIH UL1 RR025758). BSC was supported by the National Institute of Neurological Disorders and Stroke (R01 NS073601).
3T MRI scanner | |||
MRI functional connectivity software | |||
MRI image viewing software | MRICron | ||
Transcranial Magnetic Stimulator | Nexstim | eXimia Stimulator | Can use stimulators from other suppliers e.g. Magventure, Magstim |
MRI neuronavigation system | Nexstim | NBS v3.2.1 | Alternative MRI neuronavigation system e.g. Brainsight, Localite |
TMS-compatible EEG system | Nexstim | Eximia EEG | Alternatives: Brain Products, Synamps, ANT |
Matlab | Mathworks | R2012b | Alternatives: Octave |
EEGLab | |||
Minimum Norm Estimate (MNE) software | |||
FreeSurfer |