REDUCTION OF SKIN STRETCH INDUCED MOTION ARTIFACTS IN ELECTROCARDIOGRAM MONITORING USING ADAPTIVE FILTERING

Abstract

Cardiovascular disease (CVD) is the leading cause of death in many regions worldwide, accounting for nearly one third of global deaths in 2001. Wearable electrocardiographic cardiovascular monitoring devices have contributed to reduce CVD mortality and cost by enabling the diagnosis of conditions with infrequent symptoms, the timely detection of critical signs that can be precursor to sudden cardiac death, and the long-term assessment/monitoring of symptoms, risk factors, and the effects of therapy. However, the effectiveness of ambulatory electrocardiography to improve the treatment of CVD can be significantly impaired by motion artifacts which can cause misdiagnoses, inappropriate treatment decisions, and trigger false alarms. Skin stretch associated with patient motion is a main source of motion artifact in current ECG monitors. A promising approach to reduce motion artifact is the use of adaptive filtering that utilizes a measured reference input correlated with the motion artifact to extract noise from the ECG signal. Previous attempts to apply adaptive filtering to electrocardiography have employed either electrode deformation or acceleration, body acceleration, or skin/electrode impedance as a reference input, and were not successful at reducing motion artifacts in a consistent and reproducible manner. This has been essentially attributed to the lack of correlation between the reference input selected and the induced noise. In this study, motion artifacts are adaptively filtered by using skin strain as the reference signal. Skin strain is measured non-invasively using a light emitting diode (LED) and an optical sensor incorporated in an ECG electrode. The optical strain sensor is calibrated on animal skin samples and finally in-vivo, in terms of sensitivity and measurement range. Skin stretch induced artifacts are extracted in-vivo using adaptive filters. The system and method are tested for different individuals and under various types of ambulatory conditions with the noise reduction performance quantified.