Predicting seizures in untreated temporal lobe epilepsy using point-process nonlinear models of heartbeat dynamics

Area of Science:

• Neuroscience
• Cardiology
• Biomedical Engineering

Background:

• Temporal lobe epilepsy (TLE) is linked to autonomic dysregulation, a factor in sudden unexpected death in epilepsy (SUDEP).
• Cardiovascular changes during seizures are known, but pre-ictal autonomic cardiac effects remain unclear.

Purpose of the Study:

• To investigate heart rate variability (HRV) in TLE patients during inter-ictal and pre-ictal periods.
• To explore nonlinear, time-varying cardiovascular oscillation patterns preceding seizures.

Main Methods:

• ECG recordings from 12 TLE patients analyzed for HRV.
• Instantaneous, nonlinear HRV complexity (Lyapunov exponent, entropy) and higher-order statistics (bispectra) extracted.
• Inhomogeneous point-process nonlinear models with Volterra-Laguerre expansions used.

Main Results:

• Nonlinear, time-varying HRV features significantly improved inter-ictal vs. pre-ictal classification (73.91% balanced accuracy).
• Retaining the dynamic structure of heartbeat signals was crucial for accurate classification.

Conclusions:

• Nonlinear HRV analysis during pre-ictal periods shows promise for seizure prediction in TLE.
• Cardiovascular signals offer a potential non-invasive method for anticipating ictal events.