Detecting Epileptogenic Spikes Using Ultrafast 7T EEG-fMRI with 3D Paradigm Free Mapping

By Stephen E. Jones, MD, PhD

Patients suffering from intractable focal epilepsy potentially can be cured if the epileptogenic zone is identified and surgically resected. While current techniques such as scalp and intracranial EEG, structural MRI, MEG and PET are often helpful in identifying this zone, many patients remain untreated due to inadequacy of these techniques, a complicated disease presentation or both. For this reason, there is a continued need to develop new techniques to help identify the epileptogenic zone.

Promise and limits of EEG-fMRI

One promising technique is simultaneous EEG-fMRI, wherein sufficiently large epileptic spikes cause an associated change in local blood flow, which then can be visualized using a blood-oxygen-level-dependent (BOLD)-sensitive sequence during functional MRI (fMRI). This technique was first described nearly 20 years ago and classically requires EEG recordings to be obtained simultaneously during an MRI scan. These recordings then can be used to identify the time points of any spikes occurring during the MRI, which allows postprocessing methods to visualize the three-dimensional location of the epileptogenic spike. The advantage of adding MRI to identify the location is that its spatial accuracy is far superior to that of scalp EEG recordings alone.

Nevertheless, there are several drawbacks to this promising method:

• The BOLD effect elicited by a spike can be weak and may not be reliably detectable.
• The time resolution of MRI (typically 2 seconds) is much longer than that of EEG (less than a millisecond), so MRI cannot identify the temporal evolution of a spike.
• Spikes are not necessarily frequent and may occur only a few times an hour. This means that patients must be scanned for up to an hour, which also requires an epileptologist to carefully review scans of the entire session for just a few spikes.

Taking the technique to ultrahigh field strength

At Cleveland Clinic, we have recently explored ways to address these difficulties by extending the earlier technique to ultrahigh-field MRI (7 tesla [7T], versus standard 1.5T or 3T), which first required developing a new head coil for safe operation and extensive testing. Using 7T MRI provides several advantages:

• The BOLD effect is inherently stronger at 7T than at 3T, by a factor of two to three.
• New multiband techniques can reduce the time resolution from 2 seconds to 0.3 seconds.
• We incorporate a new detection algorithm called paradigm free mapping (PFM),1 which searches for the BOLD signature of any potential spikes within the data rather than requiring simultaneous EEG. Due to the long time required to scan these patients, special motion correction algorithms also have been adopted.2

We have tested the method using a simple model of a spike and its associated network: Volunteers have been scanned while in the resting state except for performing a single volitional tap of the index finger in response to an auditory stimulus. This method can easily detect the known BOLD activation maps for the associated auditory and motor networks. Thus we hypothesize that the method should be able to detect any spike whose BOLD response is at the level of a single finger tap or greater.

After confirming the model with tests, we have now applied this method to eight patients with epilepsy. Example findings from two patients are presented in Figures 1 and 2, both of which show close correspondence of EEG-fMRI localization to other localization modalities, as detailed in the captions. These early data provide evidence to support future studies to thoroughly investigate this noninvasive method.

Figure 1. Example of 7T EEG-fMRI applied to an epilepsy patient with a known occipital lobe focus, as displayed by the single-dipole models from a MEG study in the bottom row (performed before 7T EEG-fMRI). The top row shows three planes through the right occipital lobes, where the red overlaying the gray anatomy shows a strong focus of BOLD activation. Note the close correspondence of the BOLD location to that shown on MEG. PFM = paradigm free mapping.

Figure 2. Findings of 7T EEG-fMRI recently applied to a 24-year-old right-handed man with a three-year history of epilepsy now treated with multiple medications. While the structural MRI is negative, both PET and video-EEG indicated focal epilepsy likely arising from the right temporofrontal regions. The displayed EEG-fMRI maps were obtained without timing information from EEG, as computed from four BOLD spikes. Corresponding to the other findings, these maps indicate a small network of activation (yellow arrows) in the right frontotemporal region.

Conclusion

Any noninvasive MRI technique that can guide neurosurgeons to resect the region of brain tissue causing epilepsy holds enormous potential benefit. We are addressing the need for such a technique by extending traditional EEG-fMRI methods to 7T MRI, which markedly increases detection sensitivity and temporal resolution of epileptogenic spikes, and can be performed in a data-driven manner.

Acknowledgments. The author acknowledges contributions to this work from César Caballero-Gaudes (Basque Center of Cognition, Brain and Language, San Sabastian, Spain); from Anna Crawford; Mark Lowe, PhD; Sehong Oh, PhD; and Wanyong Shin, PhD (Imaging Institute, Cleveland Clinic); and from Balu Krishnan, PhD, and Imad Najm, MD (Neurological Institute, Cleveland Clinic).

References

1. Tan FM, Caballero-Gaudes C, Mullinger KJ, et al. Decoding fMRI events in sensorimotor network using sparse paradigm free mapping and activation likelihood estimates. Hum Brain Mapp. 2017;38:5778-5794.
2. Beall EB, Lowe MJ. SimPACE: generating simulated motion corrupted BOLD data with synthetic-navigated acquisition for the development and evaluation of SLOMOCO: a new, highly effective slicewise motion correction. Neuroimage. 2014;101:21-34.

Dr. Jones is a staff neuroradiologist and Vice Chair for Research and Academic Affairs in Cleveland Clinic’s Imaging Institute.