Interictal intracranial EEG asymmetry lateralizes temporal lobe epilepsy

Erin C. Conrad, Alfredo Lucas, William K.s. Ojemann, Carlos A. Aguila, Marissa Mojena, Joshua J. LaRocque, Akash R. Pattnaik, Ryan S. Gallagher, Adam Greenblatt, Ashley Tranquille, Alexandra Parashos, Ezequiel Gleichgerrcht, Leonardo Bonilha, Brian Litt, Saurabh R. Sinha, Lyle Ungar, Kathryn A. Davis

Abstract

Patients with drug-resistant temporal lobe epilepsy (TLE) often undergo invasive intracranial EEG (iEEG) monitoring to capture spontaneous seizures and identify the seizure onset zone (SOZ). This process is associated with significant morbidity, high costs, and the risk of postoperative seizure recurrence. While abundant interictal (between-seizure) data is recorded during this period, it currently plays a minimal role in surgical planning. This study aimed to predict SOZ laterality using interictal iEEG data. We developed a logistic regression model using data from a single-center internal cohort (University of Pennsylvania) and validated it using a two-center external cohort (Medical University of South Carolina). Our findings demonstrate that interictal spike rate asymmetries accurately identify left-sided seizure onset zones and correlate with successful surgical outcomes, suggesting that long-term seizure monitoring may not always be necessary for lateralization.

Introduction

For patients with drug-resistant temporal lobe epilepsy, surgical intervention offers the best chance for seizure freedom. However, the success of surgery depends on the accurate identification of the Seizure Onset Zone (SOZ). The current clinical gold standard involves implanting intracranial electrodes and monitoring the patient for one to two weeks to record multiple spontaneous seizures. This "waiting game" is not only expensive and taxing for the patient but also carries inherent risks of infection and hemorrhage.

Furthermore, identifying patients with bilateral epilepsy—where seizures originate from both hemispheres—can take weeks of monitoring, delaying treatment decisions. This research addresses the need for faster, safer biomarkers by analyzing the vast amounts of "interictal" or background brain activity that occur between seizures, which are typically overlooked in favor of the seizures themselves.

Methods

We employed a data-driven approach to determine if the brain's quiet periods contain signatures indicative of the SOZ laterality.

Feature Extraction and Asymmetry Index

We analyzed 12-hour blocks of iEEG data from 47 patients in an internal cohort. From these recordings, we extracted 180 features, including univariate measures (such as spike rates and signal power) and bivariate measures (functional connectivity). To simplify these data for clinical use, we calculated an Asymmetry Index (AI) for each feature using the formula: $(Left - Right) / (Left + Right)$. A positive AI indicates a feature is more prominent in the left hemisphere, while a negative AI indicates right-side prominence.

Model Development

We developed logistic regression classifiers using LASSO (Least Absolute Shrinkage and Selection Operator) to prevent overfitting and Principal Component Analysis (PCA) for dimensionality reduction. We compared complex models utilizing all 180 features against simpler models focused exclusively on interictal spike rates.

Validation

The study utilized three distinct datasets:

  1. Internal Cohort (HUP): 47 patients used for model development and cross-validation.
  2. External Cohort (MUSC): 19 patients used for independent validation to ensure the model's generalizability across different clinical environments.
  3. fMRI Cohort: 62 patients used to validate whether the observed electrical asymmetries were mirrored in functional MRI connectivity patterns.

Key Insights

The Power of Spike Rates

Contrary to expectations that complex connectivity measures would provide the most insight, a simple model using only interictal spike rates outperformed more complex versions. In machine learning, simpler models are often more robust; here, additional features appeared to introduce more noise than predictive signal.

The Left-Side Advantage

Across both iEEG and fMRI data, left-sided seizure zones were significantly easier to identify than right-sided ones. This asymmetry may stem from fundamental differences in brain network organization—such as the localization of language centers—or potential selection biases in which right-sided patients are referred for invasive monitoring.

Temporal Efficiency

One of the most striking findings was that the model's accuracy stabilized with as little as five minutes of interictal data. This suggests that the current practice of multi-day recordings might be significantly shortened if interictal biomarkers are prioritized for lateralization.

Results

In the internal cohort, the spike-only model achieved an Area Under the Curve (AUC) of 0.83 for predicting left-sided onset. The model proved robust during external validation, achieving balanced accuracies of 79.3% for left-sided and 78.9% for right-sided predictions in the independent MUSC dataset.

Crucially, the model's predictions correlated with clinical success. Patients whose model-predicted laterality matched the side of their actual surgery had significantly better seizure-freedom outcomes at one year (p = 0.037 for Engel outcomes; p = 0.014 for ILAE outcomes). This suggests the model captures the underlying pathology relevant to surgical success.

Discussion and Limitations

While these results represent a significant step toward quantitative, seizure-independent lateralization, several limitations remain. The sample sizes for the cohorts are relatively small, and while the model excels at lateralization (identifying the side), it does not yet provide localization (pinpointing the specific area within the lobe). Precise localization remains critical for procedures like laser interstitial thermal therapy (LITT).

Future research will focus on whether these quantitative asymmetries can be detected using non-invasive scalp EEG. If successful, this could revolutionize the pre-surgical workup for epilepsy, reducing the need for invasive electrode implantation and making surgical evaluation safer and more accessible. To encourage further validation, we have provided a free online calculator for clinicians to test the model in their own practices.

Summary updated

Abstract

Patients with drug-resistant temporal lobe epilepsy often undergo intracranial EEG recording to capture multiple seizures in order to lateralize the seizure onset zone. This process is associated with morbidity and often ends in postoperative seizure recurrence. Abundant interictal (between-seizure) data is captured during this process, but these data currently play a small role in surgical planning. Our objective was to predict the laterality of the seizure onset zone using interictal (between-seizure) intracranial EEG data in patients with temporal lobe epilepsy. We performed a retrospective cohort study (single-center study for model development; two-center study for model validation). We studied patients with temporal lobe epilepsy undergoing intracranial EEG at the University of Pennsylvania (internal cohort) and the Medical University of South Carolina (external cohort) between 2015 and 2022. We developed a logistic regression model to predict seizure onset zone laterality using interictal EEG. We compared the concordance between the model-predicted seizure onset zone laterality and the side of surgery between patients with good and poor surgical outcomes. 47 patients (30 women; ages 20-69; 20 left-sided, 10 right-sided, and 17 bilateral seizure onsets) were analyzed for model development and internal validation. 19 patients (10 women; ages 23-73; 5 left-sided, 10 right-sided, 4 bilateral) were analyzed for external validation. The internal cohort cross-validated area under the curve for a model trained using spike rates was 0.83 for a model predicting left-sided seizure onset and 0.68 for a model predicting right-sided seizure onset. Balanced accuracies in the external cohort were 79.3% and 78.9% for the left- and right-sided predictions, respectively. The predicted concordance between the laterality of the seizure onset zone and the side of surgery was higher in patients with good surgical outcome. In conclusion, interictal EEG signatures are distinct across seizure onset zone lateralities. Left-sided seizure onsets are easier to distinguish than right-sided onsets. A model trained on spike rates accurately identifies patients with left-sided seizure onset zones and predicts surgical outcome.
BibTeX
@inproceedings{conrad2024interictal,
  author = {Erin C. Conrad and Alfredo Lucas and William K.S. Ojemann and Carlos A. Aguila and Marissa Mojena and Joshua J. LaRocque and Akash R. Pattnaik and Ryan S. Gallagher and Adam Greenblatt and Ashley Tranquille and Alexandra Parashos and Ezequiel Gleichgerrcht and Leonardo Bonilha and Brian Litt and Saurabh R. Sinha and Lyle Ungar and Kathryn A. Davis},
  title = {{Interictal intracranial EEG asymmetry lateralizes temporal lobe epilepsy}},
  booktitle = {Brain},
  year = {2024},
  doi = {10.1093/braincomms/fcae284},
}
Figure