ECG signal quality in intermittent long-term dry electrode recordings with controlled motion artifacts

Abstract

Wearable long-term monitoring applications are becoming more and more popular in both the consumer and the medical market. In wearable ECG monitoring, the data quality depends on the properties of the electrodes and on how they interface with the skin. Dry electrodes do not require any action from the user. They usually do not irritate the skin, and they provide sufficiently high-quality data for ECG monitoring purposes during low-intensity user activity. We investigated prospective motion artifact-resistant dry electrode materials for wearable ECG monitoring. The tested materials were (1) porous: conductive polymer, conductive silver fabric; and (2) solid: stainless steel, silver, and platinum. ECG was acquired from test subjects in a 10-min continuous settling test and in a 48-h intermittent long-term test. In the settling test, the electrodes were stationary, whereas both stationary and controlled motion artifact tests were included in the long-term test. The signal-to-noise ratio (SNR) was used as the figure of merit to quantify the results. Skin-electrode interface impedance was measured to quantify its effect on the ECG, as well as to leverage the dry electrode ECG amplifier design. The SNR of all electrode types increased during the settling test. In the long-term test, the SNR was generally elevated further. The introduction of electrode movement reduced the SNR markedly. Solid electrodes had a higher SNR and lower skin-electrode impedance than porous electrodes. In the stationary testing, stainless steel showed the highest SNR, followed by platinum, silver, conductive polymer, and conductive fabric. In the movement testing, the order was platinum, stainless steel, silver, conductive polymer, and conductive fabric.

Keywords: Biomedical electrodes; Electrocardiography; Impedance; Materials testing; Signal-to-noise ratio; Skin–electrode interface.

Figure 1

Six dry electrode types were manufactured for evaluation. The electrodes had identical contact surface areas and lead lengths, ending in a snap fastener enabling an easy connection to ECG recorder lead wires.

Figure 2

The instrumentation for producing the controlled mounting force and motion artifacts as well as for acquiring ECG recordings. (A) Schematic presentation and a lateral view of the instrumentation. Black arrows indicate commands and force sensor data, and blue arrows indicate ECG data. (B) Frontal view of the instrumentation. (C) The instrumentation’s servo and force sensor in place during electrode testing on a test subject’s medial forearm. AL and AH stand for the distal and proximal forearm electrode locations, respectively.

Figure 3

The electrode locations and recording channels. (A) Chest, Arm Low, and Arm High electrode positions, and the voltage channels Ch1 and Ch2 recorded from them. (B) and (C) electrode labels and inter-electrode distances. The electrodes in a set of three electrodes are of the same type. The electrode labels were C1, AH1, and AL1 in set 1; C2, AH2, and AL2 in set 2; and C3, AH3, and AL3 in set 3.

Figure 4

Example data from stainless steel test electrodes during the settling and long-term test. The figures illustrate the Ch1 (derivation C-AH, blue) and Ch2 (derivation AL-AH, red) after filtering. In the initial minute of the settling test, both channels exhibit noisy data, with only the R-peaks being discernible in Ch1. During the 10th and final minute of the settling test, the ECG waveform is distinctly visible in Ch1, whereas Ch2 appears as nearly isoelectric. Thereafter, one minute of movement provocation and stationary data were acquired intermittently at 0 h, 24 h, and 48 h. Here, only the movement test data are shown. The 2 Hz movement of the AH electrode and the resulting motion artifact are evident as periodic fluctuations in both channels. The motion artifact varies based on the properties of the skin and the skin–electrode interface at the time of the recording.

Figure 5

The single time constant model of the skin–electrode interface.

Figure 6

The ECG SNR in dB for all electrodes during the settling test (top row) and intermittent long-term tests (bottom row). The top and bottom borders of the box represent the first and third quartiles, respectively, and the line in the middle is the median. The whiskers indicate the highest and the lowest values. In the long-term test, the blue boxes show the result for the stationary and the red boxes for the movement provocation test.

Figure 7

Impedance data over the recorded 1 Hz–1 MHz band from all of the test electrodes during the settling and long-term test. The far left panel depicts 1st- and 10th-minute results of the settling test. The following three panels show the intermittent long-term movement provocation and stationary test results. Overall, the solid materials consistently exhibit impedances that are lower than those observed in porous materials.

Figure 8

Test electrode’s skin–electrode interface impedance in the 1–100 Hz biosignal band and SNR from the movement and stationary tests. The stationary data include average impedances from each minute of the settling test, as well as the one-minute tests at 0 h, 24 h, and 48 h. The movement data contain average impedances in the minute-long movement provocation testes at 0 h, 24 h, and 48 h.

Figure 9

ECG over one cardiac cycle from a test subject wearing five test electrodes, one of each surface material. The data have been recorded during the 10th minute of the settling test. Due to limitations of the recording equipment, the data were recorded sequentially and processed and aligned in post-processing. Some residual, 10 Hz common-mode interference is seen in the porous electrode data. The skin–electrode interface model component values for the presented case are in the figure legend.