Using Intracranial EEG-fMRI Mapping of Intermittent Discharges to Predict Epilepsy Surgery Outcomes

Background and Objective

Epilepsy is a common neurological disorder, and many patients are unresponsive to pharmacological treatments, making surgery one of the primary therapeutic options. However, accurately localizing the seizure onset zone (SOZ) to maximize surgical efficacy remains a significant clinical challenge. The blood oxygen level-dependent (BOLD) response associated with interictal epileptiform discharges (IEDs) can be crucial for delineating the epileptic focus. Although such techniques have been applied in research since the 1990s, they remain underestimated in clinical practice. To address this, William Wilson et al. conducted a study to assess the efficacy of various subjective and objective methods in localizing clinically relevant areas and comparing these regions with postoperative outcomes.

Source of the Study

This study was conducted by the Hotchkiss Brain Institute, Cumming School of Medicine, Seaman Family MR Research Centre, among others. The research was published in 2024 by Oxford University Press on behalf of the Guarantors of Brain. It is an open-access article, with unrestricted distribution and reuse.

Research Methods

Participants

A total of 70 patients who underwent intracranial video-EEG monitoring were recruited for this study. All subjects met the following criteria: aged over 18, capable of providing informed consent, no contraindications to MRI, and diagnosed with focal epilepsy. The study was approved by the Research Ethics Board of the University of Calgary.

Data Acquisition and Preprocessing

Intracranial EEG

The intracranial video-EEG monitoring data of each participant were reviewed by experienced epileptologists, and up to eight electrodes were selected for EEG-fMRI acquisition. The EEG data acquisition process employed real-time artifact rejection and filtering.

MRI

Participants underwent scanning on different types of 3T GE Signa LX and GE Discovery MR750 whole-body scanners. MRI scans included multilayer anatomical images and functional MRI images. All subjects were advised to sleep during the scan to increase the probability of IED occurrences. A total of 60 minutes of functional data was collected.

Data Analysis

Intracranial EEG

Post-processing of the data used customized Matlab algorithms for re-referencing, bandpass filtering, average artifact subtraction, and principal component analysis. Two experienced doctors then separately reviewed and annotated the datasets.

fMRI

fMRI analysis was conducted using the FMRIB Software Library (FSL), including time correction, spatial smoothing, high-pass filtering, and motion correction. Event-related analysis of IED-related BOLD responses was performed by convolving the timing of IED events with four different HRF models.

Cluster Identification and Refinement

In each IED analysis, only positive BOLD responses were considered. The study identified and compared the most significant cluster, the second most significant cluster, the cluster closest to the IED electrode, and the clinically relevant cluster. Distances of each cluster to the electrode contacts were measured and compared, and patients underwent postoperative imaging to identify resected areas.

Research Results

Participant Characteristics

Out of all 70 participants, 60 completed at least 10 minutes of functional imaging acquisition. The average age of the participants was 35.8 years, with an average epilepsy duration of 18.2 years. Ultimately, 38 patients underwent surgery, with good prognoses (Engel Class I and II) achieved in 23 patients (61%).

Clinically Relevant Clusters

A total of 117 types of IEDs were analyzed, generating 106 significant BOLD response maps. Clinically relevant clusters were identified in ⁸⁵⁄₁₀₆ (80%) IED analyses, featuring significant z-scores and larger cluster volumes.

Surgical Outcomes and Consistency

In 68 significant cluster maps, a significant correlation was found between the resection of the peak clusters and good postoperative outcomes. High benchmark maps showed high negative predictive value and sensitivity in predicting surgical outcomes.

Discussion

The study found that both subjective methods (clinically relevant clusters) and objective methods (confidence level of the largest clusters) improved the accuracy of SOZ localization, but only the objective method showed a significant correlation with good surgical outcomes. Furthermore, although resection of the peak voxel is crucial for surgical success, it alone is insufficient to guarantee complete surgical success.

Conclusion

The study demonstrates that further refining IED-related BOLD clusters can improve the prognosis of epilepsy surgery. Such advanced screening methods should be applied in future clinical practice and research to improve surgical outcomes for epilepsy patients. Additionally, intracranial EEG-fMRI has a higher detection rate of IED and higher BOLD activation compared to scalp EEG-fMRI, showing its extensive clinical application prospects.

Research Highlights

1. This study proposes a method of further refining IED-related BOLD clusters, improving the accuracy of epileptic focus localization.
2. The research shows that resecting the peak clusters of the IED-related BOLD response is crucial for surgical success, being a necessary but not sufficient condition.
3. Intracranial EEG-fMRI outperforms scalp EEG-fMRI in analyzing IED and BOLD activation, indicating its higher clinical value.

Limitations and Future Research Directions

1. Signal loss caused by intracranial electrodes limits the ability to detect BOLD signals.
2. Future research needs to further explore the clinical value of low-confidence level maps and determine the specific role of EEG-fMRI in preoperative work.
3. The study’s methods may have limitations in promotion and application across different centers, requiring cross-center validation in the future.

This study presents a novel method for precisely locating the epileptic focus using EEG-fMRI, emphasizing the importance of further research and rational application of these high-confidence clusters, providing strong support for future epilepsy surgery planning.