Forecasting seizures in dogs with naturally occurring epilepsy

Area of Science:

• Neuroscience
• Epilepsy Research
• Biomedical Engineering

Background:

• Epilepsy treatment can be advanced by seizure forecasting for patient warnings and preemptive therapy.
• Progress in seizure forecasting is limited by insufficient data to confirm if seizures are preceded by physiological changes or are random.
• Canine models offer a valuable platform for studying epilepsy and developing forecasting methods.

Purpose of the Study:

• To investigate the feasibility of long-term seizure forecasting in dogs with naturally occurring focal epilepsy.
• To determine if physiological changes precede seizures, supporting the hypothesis that seizures are not random events.
• To evaluate the performance of a seizure forecasting algorithm using intracranial EEG data.

Main Methods:

• Continuous intracranial EEG (iEEG) was recorded from three dogs with focal epilepsy.
• Spectral power in six frequency bands (delta, theta, alpha, beta, low-gamma, high-gamma) was extracted from iEEG data.
• Logistic regression classifiers were trained to differentiate pre-ictal and inter-ictal states, with performance assessed using 10-fold cross-validation.

Main Results:

• A total of 125 spontaneous seizures were detected over 6.5 to 15 months of recording in the three dogs.
• The seizure forecasting algorithm significantly outperformed a chance predictor, even when accounting for seizure clustering.
• Forecasting performance remained above chance for specific parameters when analyzing seizures separated by at least 4 hours.

Conclusions:

• Seizures in canine epilepsy are not random events and are preceded by detectable physiological changes.
• Long-term seizure forecasting using iEEG monitoring is feasible in a naturalistic epilepsy model.
• These findings support the development of advanced therapeutic strategies for epilepsy.