Graph energy based centrality measures to detect epileptogenic focal invasive EEG electrodes

Purpose: Medically intractable epilepsy can be treated with surgical interventions, which require localization of the cortical region where seizures start. This region is referred to as the epileptogenic zone (EZ). Good surgical outcomes depend on an exact localization of the EZ. Methods: We propose a graph theoretical approach providing a novel method to localize the epileptogenic zone using invasive electroencephalogram (EEG) data. The proposed methods employ centrality determination using three graph energies, namely simple graph energy, Laplacian energy, and distance energy. Centrality values of invasive EEG electrodes from 19 patients were analyzed at different frequency bands and at different time points. K-means clustering was used to distinguish focal (electrodes placed in the epileptogenic zone) from non-focal electrodes using the centrality values obtained. Results: Focal electrodes show higher centrality values when compared to non-focal electrodes. All three graph energy based centrality measures proposed show maximum f-score and accuracy during the early seizure phase in the gamma frequency band. Among the three proposed methods, simple graph energy based centrality outperforms Laplacian centrality and distance energy based centrality and also other related and competitive methods available in the literature in terms of accuracy and f-score. Conclusion: Graph energy based centrality measures are useful parameters for the delineation of the epileptogenic zone. Among the three centrality measures examined, simple graph energy based centrality proved best suited for this purpose.

Automated summary

A graph energy based centrality approach was proposed for localizing the epileptogenic zone, showing that simple graph energy based centrality outperforms other methods in accuracy and f-score.