Removing cardiac interference from the electroencephalogram using a modified Pan-Tompkins algorithm and linear regression

Cardiac interference can alter the results of quantitative electroencephalograms (qEEG) used for medical diagnoses. The methods currently employed for the automated removal of cardiac interference, which rely solely on the electroencephalogram (EEG), are susceptible to non-cardiac interference commonly encountered in EEGs. Methods that rely on the electrocardiogram (ECG) - besides being unreliable when non-cardiac artifacts corrupt the ECG - either assume periodicity of the cardiac (QRS) peaks or alter uncorrupted EEG segments. This paper proposes a robust method for the automated removal of cardiac interference from EEGs by identifying QRS peaks in the ECG without assuming periodicity. Artificial signals consisting only of QRS peaks and the zero-lines in between are computed. Linear regression of the EEG channels on the “QRS signals” removes cardiac interference without altering uncorrupted EEG segments. The QRS-based regression method was tested on 30 multi-channel EEGs exhibiting cardiac interference of elderly subjects (15 male, 15 female). Achieving a correction rate of 80%, the QRS-based regression method has proved effective in removing cardiac interference from the EEG even in presence of additional non-cardiac interference in the EEG.