UCLA researchers have developed an automated and unsupervised digital signal processing method to quickly and efficiently minimize and reject artifacts from scalp Electroencephalogram (EEG) and intracranial EEG recordings.
Electroencephalogram (EEG) is designed to record cerebral activity. However, in addition to cerebral activity, other electrical activity from regions other than the brain are also recorded. This added activity is characterized by abrupt vertical transient changes in signal creating a significant amount of noise that masks the original signal. Although techniques and existing technology work to retrieve this original signal, they typically are mediocre at best and require supervision. There is a need for an automated, unsupervised, reliable method to filter EEG signal for interpretation and clinical diagnosis.
Automated and unsupervised method for electroencephalogram artifact minimization and rejection
Filter the EEG digitally above 16 Hz, then determine an epoch of artifact in the EEG. Calculate association between electrode and epoch. Identify faulty electrodes and remove data from faulty electrodes.
Process high pass data (>16Hz) in consecutive bins, then process using independent component analysis (ICA). Identify independent components of data (EEG) in each bin and reassemble the signal from each bin. Low pass filter the reconstructed signal.
DSP method, software, algorithm, artifact removal, noise removal, contamination removal, electroencephalogram, EEG, HFO detection, artifact removal intracranial, High Frequency Oscillation, binning