Estimating unmeasured invasive EEG signals using a reduced-order observer

Area of Science:

• Neuroscience
• Control Theory
• Computational Biology

Background:

• Epilepsy affects 50 million globally, with over 30% experiencing drug-resistant seizures.
• Surgical resection of the epileptogenic zone (EZ) is a key treatment, but success rates are limited (approx. 65%).
• Accurate EZ localization relies on dense electroencephalogram (EEG) electrode placement, leading to the "missing electrode problem".

Purpose of the Study:

• To develop a computational platform for estimating neural activity at "missing" electrode locations.
• To improve the precision of epileptogenic zone (EZ) identification for epilepsy surgery.
• To address the limitations of sparse EEG recordings in clinical practice.

Main Methods:

• Developed discrete-time Linear Time-Invariant (LTI) models from patient EEG data.
• Simulated EEG data and selectively removed signals representing "missing" states.
• Employed a reduced-order observer from control theory to estimate the removed neural signals.

Main Results:

• Successfully estimated neural activity at simulated "missing" electrode locations.
• Evaluated observer performance by comparing estimated signals to the original simulated EEG time series.
• Demonstrated the feasibility of using control theory for neural signal estimation.

Conclusions:

• The developed computational platform shows promise for estimating neural activity in unmonitored brain regions.
• This method could enhance the accuracy of epileptogenic zone (EZ) localization in epilepsy surgery.
• Further development is needed to translate this approach into clinical application for epilepsy management.