Estimating the Memory Order of Electrocorticography Recordings.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electrocorticography (ECoG) recordings are crucial for understanding brain activity.
• Estimating the memory order of ECoG signals is vital for analyzing causal associations.
• Current methods may lack precision in characterizing ECoG signal dependence.

Purpose of the Study:

• To develop and present a novel data-driven method for estimating the memory order of ECoG sequences.
• To quantify the statistical dependence of ECoG samples on preceding ones.
• To improve the localization of the epileptogenic zone for epilepsy surgery.

Main Methods:

• The method estimates memory order by finding the number of past samples that minimize least squares error (LSE) in predicting the next sample.
• A collection of diverse predictors is combined to overcome the lack of an analytical ECoG model.
• The approach was applied to ECoG recordings from six epilepsy patients during rest and pre-ictal states.

Main Results:

• Memory order was estimated for ECoG recordings, with values typically in the tens of milliseconds (up to 100 ms).
• Empirical cumulative density functions (ECDFs) of memory orders varied between patients and recording intervals.
• Distinct ECoFs were observed between rest and pre-ictal states, and for seizure-onset-zone electrodes.

Conclusions:

• The developed method provides a robust way to estimate ECoG memory order, crucial for causal association analysis.
• Improved estimation of causal associations can enhance algorithms for epileptogenic zone localization.
• This work has significant implications for pre-surgical planning in epilepsy patients.