High frequency oscillation network dynamics predict outcome in non-palliative epilepsy surgery

Area of Science:

• Neuroscience
• Epilepsy Research
• Computational Biology

Background:

• High frequency oscillations (HFOs) are recognized biomarkers for epilepsy outcomes.
• Previous research focused on individual channel HFOs, limiting clinical applicability.
• Understanding HFOs as network phenomena is crucial for improving predictive models.

Purpose of the Study:

• To develop a predictive tool for epilepsy surgery outcomes using HFO network properties.
• To assess the value of HFO network connectivity in predicting patient outcomes before surgical resection.
• To compare the predictive accuracy of HFO network analysis versus seizure onset zone resection alone.

Main Methods:

• Correlational analysis of HFO functional connectivity in 28 epilepsy patients with intracranial electrodes.
• Utilized eigenvector and outcloseness centrality to identify important channels within HFO networks.
• Developed a Naïve Bayes model integrating HFO rate, network centralities, and seizure onset zone resection data.

Main Results:

• HFOs frequently manifest as local network discharges, not isolated events.
• The developed Naïve Bayes model achieved 100% positive predictive value for predicting surgical outcomes.
• Predictive accuracy was significantly higher when analyzing HFO network properties compared to seizure onset zone resection alone (71% PPV).
• Outcomes were predictable in definitive surgeries (≥80% seizure onset zone resection) but not in palliative surgeries (<80% resection).

Conclusions:

• Network properties of HFOs offer superior accuracy in predicting epilepsy surgery outcomes compared to seizure onset zone resection alone, particularly in cases with substantial resection.
• This HFO network analysis tool holds significant promise for guiding clinical decisions in refractory epilepsy surgery.
• Incorporating HFO network characteristics can refine pre-surgical planning and improve patient outcomes.