Effect of pressure and padding on motion artifact of textile electrodes

Abstract

Background: With the aging population and rising healthcare costs, wearable monitoring is gaining importance. The motion artifact affecting dry electrodes is one of the main challenges preventing the widespread use of wearable monitoring systems. In this paper we investigate the motion artifact and ways of making a textile electrode more resilient against motion artifact. Our aim is to study the effects of the pressure exerted onto the electrode, and the effects of inserting padding between the applied pressure and the electrode.

Method: We measure real time electrode-skin interface impedance, ECG from two channels, the motion artifact related surface potential, and exerted pressure during controlled motion by a measurement setup designed to estimate the relation of motion artifact to the signals. We use different foam padding materials with various mechanical properties and apply electrode pressures between 5 and 25 mmHg to understand their effect. A QRS and noise detection algorithm based on a modified Pan-Tompkins QRS detection algorithm estimates the electrode behaviour in respect to the motion artifact from two channels; one dominated by the motion artifact and one containing both the motion artifact and the ECG. This procedure enables us to quantify a given setup's susceptibility to the motion artifact.

Results: Pressure is found to strongly affect signal quality as is the use of padding. In general, the paddings reduce the motion artifact. However the shape and frequency components of the motion artifact vary for different paddings, and their material and physical properties. Electrode impedance at 100 kHz correlates in some cases with the motion artifact but it is not a good predictor of the motion artifact.

Conclusion: From the results of this study, guidelines for improving electrode design regarding padding and pressure can be formulated as paddings are a necessary part of the system for reducing the motion artifact, and further, their effect maximises between 15 mmHg and 20 mmHg of exerted pressure. In addition, we present new methods for evaluating electrode sensitivity to motion, utilizing the detection of noise peaks that fall into the same frequency band as R-peaks.

Figure 1

The electrode-skin interface model for gelled electrodes and dry electrodes. The diagram is modified from [29] and [30]. The electrical model for gelled electrodes is on the left, the electrical model for the dry electrodes is on the right.

Figure 2

Illustration of the motion range used in the experiments. The figure shows the end points of motion. The forearm was moved up and down between these end points so that motion was restricted to the elbow joint. The electrode locations are shown in Figure 3.

Figure 3

Depiction of electrode locations. Left hand side illustrates the locations from front; right hand side shows the detailed location for the arm electrode seen laterally. The chest electrodes are not affected by the motion described in Figure 2. The arm electrode is at a location that is minimally affected by the EMG of the arm and shoulder muscles that contract during movement.

Figure 4

The paddings used for the experiment. “d” is the diameter of the padding, “h” is the height. Padding 2 is an open celled viscoelastic high density foam with gel like properties; Paddings 3 and 4 are impact protection cushions that act as memory foams in the context of this study; Paddings 5 and 6 are open celled viscoelastic polyurethane memory foams. Padding 1 is not shown because this is the name used for the case where padding was not used between the ribbon and electrode.

Figure 5

The boxplots of various parameters obtained from the experiments utilizing the paddings. The parameters shown are the score obtained from R-peak detection, the scores obtained from the energies of the 1 Hz–7 Hz band components of the impedance signal, ECG signal and the motion artifact, the peak-to-peak values of the motion artifact and the baseline impedance.

Figure 6

Boxplots of the scores from R-peaks for the no padding case (1) and the paddings, in relation to pressure.

Figure 7

Flowchart summarizing the methods. The “end” decision box checks if measurement round 6 is finished. “P” is pressure and is incremented at each step by 5. After Padding 6 is measured, a new round starts with Padding 1. This loop is omitted for the sake of simplicity.

Figure 8

Measured impedance change, ECG and motion artifact, and the derived chest ECG at 20 mmHg applied pressure. Signals are presented after 0.3Hz–30 Hz band-pass filtering, and shown in a 10 second window.

Figure 9

Effect of increasing electrode pressure on the motion artifact related parameters. From top to bottom: the real time impedance signal, the ECG, the motion induced surface potential (motion artifact) and the reference ECG derived from the electrodes on the chest. The corresponding pressure for each graph box increases from left to right, from 5 mmHg to 25 mmHg in 5 mmHg increments.

Figure 10

Normalized variance graphs of various parameters, excluding the case of no padding. The data is normalized using the Matlab function “norm”, with the formula: normalized data = data/norm(data).