Power-to-power cross-frequency coupling as a novel approach for temporal lobe seizure detection and analysis

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biomedical Engineering

Background:

• Seizures are complex neurological events involving dynamic spatio-temporal interactions between brain rhythms.
• High-frequency oscillations (HFOs) are implicated in epileptogenic processes.
• Intracranial EEG (iEEG) provides high-resolution data for studying brain activity during seizures.

Purpose of the Study:

• To evaluate power-to-power cross-frequency coupling (PPC) for seizure detection in iEEG.
• To characterize unique PPC signatures of common seizure components: spike, ripple-on-spike, and ripple-on-oscillation.
• To advance the understanding of epileptogenic processes through CFC analysis.

Main Methods:

• Deep learning networks, including Stacked Sparse Autoencoder (SSAE) and Long Short Term Memory (LSTM), were trained for seizure detection.
• iEEG data from 120 Temporal Lobe Epilepsy (TLE) seizures across 26 patients were analyzed.
• Pairwise power-to-power coupling was calculated for frequencies 1-250 Hz, with CFC matrices used as input for deep learning models.

Main Results:

• The trained deep learning network achieved 90.2% sensitivity, 96.8% specificity, and 93.4% overall accuracy in distinguishing seizure from background activity.
• Distinct PPC signatures were identified for the three analyzed seizure components.
• The study demonstrated the feasibility of using deep learning with CFC for seizure detection.

Conclusions:

• Power-to-power coupling (PPC) is a relevant method for seizure detection in iEEG.
• Unique PPC signatures exist for common seizure components, offering insights into epileptogenesis.
• This approach holds promise for improving seizure detection and understanding neurological disorders.