Electrocardiography (ECG) is widely used in clinical practice, for example to diagnose coronary artery disease or the cause of chest pain during a stress test, while the patient is running on a treadmill. Ambulatory ECG monitoring is used for long term recording of ECG signals, while the patient carries out his/her daily activities. Artefacts in ECG are caused by the patient’s movement, moving cables, interference from outside sources, electromyography (EMG) interference and electrical contact from elsewhere on the body. Most of these artefacts can be minimised by using proper electrode design and ECG circuitry. However, artefacts due to subject’s movement are hard to identify and eliminate and can be easily mistaken for symptoms of arrhythmia and the physiological effects of exercise, leading to misdiagnosis and false alarms.

Skin stretch has been identified as a major source of motion artefacts in ECG signals, which arise due to the flow of current, called the ‘injury current’ across the epidermis. Thus, the skin is generally abraded or punctured to minimize variations in injury current. This is unpleasant and not useful for long term monitoring, as the skin regrows after 24 hours. Present motion sensing approaches to artefact reduction in ECG do not measure motion in terms of skin stretch.

The main goal of this study is to quantify and eliminate motion artefacts from ECG pertaining to skin stretch. A polymer patch electrode with Young’s modulus lower than of skin has been developed to simultaneously measure ECG and skin stretch using an optical sensing technique. These signals were combined with infinitesimal strain theory to quantify skin stretch as two dimensional strains. Principal component analysis (PCA) and independent component analysis (ICA) were utilised for motion artefact removal from ECG signals.

A motion Artefact Rejection (AR) system has been developed to validate the approach implemented in this study. As this study mainly focuses on skin stretch induced artefacts, a plastic tube has been used to stretch the forearm skin of 7 subjects across the following age groups: 18–35 years (3 subjects), 36–55 years (2 subjects), and 56 years and above (2 subjects). ECG with motion artefacts were measured using CNT/PDMS electrodes and dry Ag electrodes. The reference ECG (ECG at rest) was measured from the chest using conventional Ag/AgCl electrodes. The average improvements in SNRs using PCA and ICA algorithms were found to be 4.249 dB and 9.586 dB respectively, while the average of maximum deviation from rest/reference ECG was 0.0843 for ECG with motion artefacts, 0.0702 for ECG after PCA and 0.0442 for ECG after ICA.

Both PCA and ICA algorithms also aided in removing baseline wander and high frequency noises in the cases of less or no motion artefact. The system performed well in removing artefacts generated due to EMG interference and stretching the skin perpendicular, diagonal and parallel to Langer’s lines. Higher SNRs were achieved when PCA and ICA were performed by using 2D strains as motion information than when no motion information was used. In conclusion, ICA used for motion artefact reduction in ECG signals shows better performance than other techniques employing adaptive filtering, PCA and ICA.

A novel, state-of-the-art technique to identify and eliminate motion artefacts from ECG signals has been developed through this study which is feasible for practical implementations.