Two sources can generate motion artifact: the electrode metal-to-solution interface and fluctuations in skin potential due to skin stretch. The first source of artifact has been significantly reduced with modern Ag-AgCl recessed electrodes, but skin stretch remains a serious problem. Standard methods for reducing motion artifact due to skin stretch require skin abrasion, but this method has significant drawbacks: it causes patient discomfort, it results in more work for the ECG technician, and it can be ineffective for long term recordings.
We have been investigating methods for using adaptive filters to remove motion artifact from ECG recordings. An adaptive filter can remove noise from an ECG recording if a signal is available that is linearly related to the motion artifact, but contains no ECG signal. The adaptive filter uses noise corrupted ECG and the signal that is related to the motion artifact to estimate the motion artifact. This motion artifact estimate is then subtracted from the noisy ECG, reducing the motion artifact in the resulting signal.
By placing stretch sensors directly onto ECG electrodes, we have been able to generate signals that are related to the skin stretch near the electrode and the resulting motion artifact. We have been able to produce significant motion artifact reduction using an adaptive filter and these signals. Figure 1 shows a recording of an ECG, the electrode mounted stretch sensor signal, and the signal that results after the motion artifact has been removed. We are presently investigating the use of a skin/electrode impedance signal for adaptive motion artifact removal. An impedance based artifact removal system has the potential to be less expensive than one based on electrode mounted sensors because standard ECG electrodes could be used.