A probabilistic method for determining cortical dynamics during seizures.

Area of Science:

• Computational Neuroscience
• Biophysics
• Medical Signal Processing

Background:

• Epilepsy is characterized by recurrent seizures, often studied using electrocorticogram (ECoG) signals.
• Mathematical models of the cortex are used to understand seizure generation, but linking model parameters to real-world ECoG data is challenging.
• The stochastic and noisy nature of ECoG signals and computational models requires robust analytical frameworks.

Purpose of the Study:

• To develop a probabilistic method for inferring physiologically relevant parameter ranges in a cortical model that produce seizures.
• To generate probabilistic temporal pathways of cortical physiological state evolution during seizures.
• To provide insights into seizure etiology and potential new therapeutic targets.

Main Methods:

• Utilized a probabilistic Bayesian framework to map ECoG signal features to likelihood distributions over model parameter states.
• Employed a Hidden Markov Model (HMM) to capture temporal continuity and reproducibility of cortical physiology between seizures.
• Compared likelihood ratios from HMMs run under different hypothesized parameter regions to identify those most likely causing observed seizures.

Main Results:

• Demonstrated consistency in likelihood ratios between hypothesized parameter regions across individual seizures.
• Showcased reproducible temporal pathways of cortical state evolution during seizures.
• Successfully separated seizure from non-seizure time segments using likelihood maps derived from the model.

Conclusions:

• The developed probabilistic method effectively infers seizure-related cortical parameters from ECoG data.
• The findings suggest reproducible physiological dynamics underlying seizure generation and progression.
• This approach holds potential for advancing our understanding of epilepsy and guiding treatment strategies.