Investigation into Electrocardiogram Motion Artifact Reduction, 10-9191

Printer Friendly Version

Principal Investigator
David A. Tong

Inclusive Dates: 04/01/2000 - 08/31/2001

Background - The electrocardiogram (ECG) is the body-surface manifestation of the electrical potentials produced by the beating heart. The magnitude, conduction, and duration of these potentials are detected by placing electrodes on the patient's skin. The ECG is the most prescribed diagnostic procedure in medicine and is routinely used to diagnose heart disease, identify irregular cardiac rhythms (arrhythmias), evaluate the effects of drugs, and monitor surgical procedures.

Motion artifact is the noise introduced to the ECG signal that results from motion of the electrode. More specifically, movement of the electrode or lead wire produces deformations of the skin around the electrode site. The deformations of the skin change the impedance and capacitance of the skin around the sensing electrode. The impedance and capacitance changes are sensed by the ECG electrode and result in artifacts that are manifest as large amplitude signals on the ECG. The presence of motion artifact may result in misdiagnosis, can prolong procedure duration, and may lead to delayed or inappropriate treatment decisions.

Approach - The hypothesis of this project was that because electrode motion is the catalyst for motion artifact, measuring electrode motion and using the measurements in an adaptive filtering system to remove the noise would lead to a reduction in the amount of motion artifact present in the ECG. The general approach was to use existing motion-sensing technologies to measure the amount of motion at the ECG electrode site and to use adaptive filtering techniques to remove the motion artifact.

Accomplishments - Two motion sensors, using different motion-sensing technologies, were developed. One sensor utilized an anisotrophic magnetoresistive (AMR) sensor, while the other sensor utilized accelerometers. The two sensors placed on ECG electrodes are shown in the first illustration below. It is envisioned that the motion sensor electronics can be miniaturized and incorporated into the black clip for manufacturing.

After successfully demonstrating that a correlation existed between the motion signal and the motion artifact induced in the ECG, data were collected from eight human subjects following an Institutional Review Board-approved protocol. Data sets were obtained that contained noise-free conditions and motion artifact conditions that were induced by manually manipulating the ECG/motion sensor combination in three ways: 1) pushing down on the combination, 2) pulling on the ECG lead wire, and 3) pushing on the skin around the ECG electrode.

Various adaptive filters were developed and evaluated for removing motion artifact from the ECG. These filters used the motion sensor signal as a reference of the motion artifact that was induced in the ECG. The data sets were processed using the adaptive filters. An example case using the accelerometer-based sensor is shown in the second illustration. Overall, the accelerometer-based sensor outperformed the AMR sensor, with a percent improvement before and after filtering of 83.6 and 67.6 percent, respectively. Several possible reasons for this improvement include 1) the accelerometer provided three-axis of motion measurement over the two provided by the AMR sensor, 2) the relationship between the accelerometer signal and the ECG noise is closer to a linear time-varying system, (3) the accelerometers have a better frequency response and matches better to that of the ECG noise. Furthermore, the accelerometer-based sensor is easier and less expensive to implement. A patent application describing this approach of reducing motion artifact has been filed with the U.S. Patent and Trademark Office.

 

A

B

The motion sensors developed for the project. A is based on an AMR sensing
 technology and can sense motion in two dimensions. B is based
on accelerometers and can sense motion in three dimensions.

Processing results on the accelerometer mixed-noise data. Upper plot is unprocessed, noisy ECG waveform. The second plot is the magnitude of the accelerometer motion sensor waveform. The 
third plot is the adaptive filter's estimate of the motion noise.
 The lower plot is the output of the filter.

2001 IR&D Home SwRI Home