Maximum a posteriori estimation of change points in the EEG

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Electroencephalography (EEG) is crucial for studying brain activity.
• Accurate segmentation of EEG signals is challenging due to non-stationarity.
• Existing methods may lack flexibility in resolution or model-based rigor.

Purpose of the Study:

• Introduce a new model-based methodology for optimal EEG signal segmentation.
• Develop a flexible approach for analyzing non-stationary brain rhythms and artifacts.
• Demonstrate the method's efficacy across various spectral changes in EEG.

Main Methods:

• Utilized maximum a posteriori (MAP) estimation for signal segmentation.
• Developed a model-based, non-sequential segmentation approach.
• Applied the methodology to EEG recordings with diverse spectral characteristics.

Main Results:

• Successfully segmented EEG signals exhibiting normal and pathological spectral variations.
• Demonstrated effective segmentation of spontaneous brain rhythmic activities.
• Showcased robust performance in segmenting physiological artifacts within EEG data.

Conclusions:

• The proposed MAP-based methodology offers an optimal approach for EEG segmentation.
• The method provides flexibility in analyzing non-stationary EEG signals at multiple resolutions.
• This technique is valuable for characterizing brain activity and artifacts in neurological studies.