Ictal time-irreversible intracranial EEG signals as markers of the epileptogenic zone

Kaspar Schindler, Christian Rummel, Ralph G. Andrzejak, Marc Goodfellow, Frédéric Zubler, Eugenio Abela, Roland Wiest, Claudio Pollo, Andreas Steimer, Heidemarie GastClinical 2016 · DOI

TL;DR

This paper introduces a novel method using "time-irreversibility" analysis of intracranial EEG (iEEG) signals to pinpoint the epileptogenic zone in epilepsy patients. They found that brain regions generating highly time-irreversible iEEG signals during seizures are more likely to be the source of the epilepsy, and their removal leads to better seizure control.

What Problem Does This Paper Address?

Epilepsy surgery aims to remove the "epileptogenic zone" – the specific brain area responsible for generating seizures – to achieve seizure freedom. However, precisely identifying this zone can be challenging, as no single diagnostic method is perfect. Intracranial EEG (iEEG) offers high-resolution brain activity data, but extracting definitive markers of the epileptogenic zone from these complex signals remains an active area of research. This paper tackles the problem of finding more robust and objective markers within iEEG signals that can reliably indicate the epileptogenic zone, thereby improving pre-surgical evaluation.

Methods

The core of this research lies in the application of a recently developed method called Directed Horizontal Visibility Graphs (dHVG). This technique transforms a time series (like an iEEG signal) into a network. Each point in time becomes a "node," and connections (edges) are drawn between nodes based on a specific geometric condition related to the signal's amplitude.

The key innovation here is the analysis of time-irreversibility. In simple terms, a time-irreversible signal behaves differently when played forwards versus backwards. The dHVG method allows for the quantification of this asymmetry by comparing the "input" and "output" degrees of the nodes in the network. A statistically significant difference between these degrees indicates time-irreversibility. The authors use a Kolmogorov-Smirnov test to assess this difference and quantify the degree of irreversibility using a metric derived from the p-value. This analysis is performed in a time-resolved manner using a moving window to capture how irreversibility changes over time, particularly around seizure events.

Datasets

The study utilized iEEG recordings from 16 patients undergoing epilepsy surgery at the Inselspital Bern. These patients had a history of pharmacoresistant epilepsy and underwent pre- and post-operative MRI. The dataset included recordings of the first two seizures for each patient, totaling 32 seizure recordings.

Key Insights and Ideas

  • Time-irreversibility as a Seizure Marker: The study strongly suggests that time-irreversible iEEG signals are not just random fluctuations but are specifically associated with seizure activity and the epileptogenic zone.
  • Saw-tooth Patterns: The most time-irreversible signals were often characterized by a "saw-tooth" shape, with slow increases and rapid decreases, which is a visually recognizable but previously hard-to-quantify feature.
  • Predictive Power for Surgery: Critically, brain regions removed from patients who became seizure-free after surgery exhibited significantly higher time-irreversibility in their iEEG signals compared to regions removed from patients who did not achieve seizure freedom. This points to time-irreversibility as a potential predictor of surgical success.
  • Efficiency of dHVG: The dHVG method is highlighted as being computationally efficient and robust, requiring no arbitrary parameter settings, making it suitable for analyzing large iEEG datasets.

Key Results

  • Statistically significant time-irreversible iEEG signals were found in 31 out of 32 seizure recordings (97%).
  • The maximally time-irreversible signals occurred during seizures, most frequently in the middle of the first seizure half (relative seizure time period 1/6 < t < 2/6).
  • Brain regions removed from post-surgically seizure-free patients generated iEEG signals with significantly stronger time-irreversibility (p = 0.002) compared to those from non-seizure-free patients.

Limitations and Future Work

  • Sample Size: The authors acknowledge that the study's limitation is the relatively small number of patients (16). They suggest replicating the findings in a larger cohort with a greater diversity of epilepsy types.
  • Speculative Mechanisms: While the study demonstrates the utility of time-irreversibility, the precise neurophysiological mechanisms underlying it remain speculative. The authors suggest it might reflect dysbalances in synchronization/desynchronization or excitation/inhibition in neural networks.
  • Interictal Signals: Future work could explore whether time-irreversible signals can also be detected during the interictal (between seizures) period, particularly during different sleep stages.
  • Network Analysis: Extending the HVG approach to analyze dynamic interdependencies between multiple iEEG signals (network measures) is proposed as a promising avenue.
  • Model Building: Time-irreversible signals could aid in developing Bayesian statistical generative models for multi-channel iEEG, potentially reducing computational load in complex simulations.

Summary updated

BibTeX
@inproceedings{schindler2016ictal,
  author = {Kaspar Schindler and Christian Rummel and Ralph G. Andrzejak and Marc Goodfellow and Frédéric Zubler and Eugenio Abela and Roland Wiest and Claudio Pollo and Andreas Steimer and Heidemarie Gast},
  title = {{Ictal time-irreversible intracranial EEG signals as markers of the epileptogenic zone}},
  booktitle = {Clinical},
  year = {2016},
  doi = {10.1016/j.clinph.2016.07.001},
}
Figure