Trunk muscle onset detection technique for EMG signals with ECG artefact

Abstract

The timing of trunk muscle activation has become an important element in the understanding of human movement in normal and chronic low back pain populations. The detection of anticipatory postural adjustment via trunk muscle onsets from electromyographic (EMG) signals can be problematic due to baseline noise or electro-cardiac (ECG) artefact. Shewhart protocols or whole signal analyses may show different degrees of sensitivity under different conditions.

Muscle activity onsets were determined from surface EMG of seven muscles for five trials before and after fatigue were examined in four subjects (n=280). The objective of this study was to examine two detection methods (Shewhart and integrated protocol (IP)) in determining the onsets of trunk muscles. The variability of the baseline amplitude and the impact of added Gaussian noise on the detected onsets were used to test for robustness.

The results of this study demonstrate that before and after fatigue there is a large degree of baseline variance in the trunk muscles (coefficients of variation between 40–65%) between trials. This could be normal response to body sway. The IP method was less susceptible to false onsets (detecting onsets in the baseline window) 3 vs. 51%. The findings suggest the IP method is robust with large variance in the baseline if the signal to noise ratio is greater than six.

In spite of the robustness of the algorithm, the findings would suggest that statistical assessments should be used to target trials for selective visual inspection for subtle trunk muscle onsets.

Introduction

The relative timing of trunk muscle activity during sudden movements is important in understanding the motor control strategies of individuals with and without chronic low back pain (LBP) [1], [2]. Both surface and fine wire electrodes are used to record the muscle activity profile from which a muscle activity ‘event’, over and above the underlying background signal, is detected [3].

During the assessment of feed forward anticipatory postural adjustments of trunk muscles the detection algorithms need to be sensitive to the very small increments over and above the baseline signal. The reason for this is that initial feed forward onsets are followed directly by significantly larger amplitude feedback responses and may partly explain why the visual detection methods are preferred [1], [2], [4], [5].

The easiest and most common reported algorithm for determining the onset of muscle activity is where the threshold is derived from the baseline signal amplitude characteristics (mean and standard deviation SD) (see Fig. 1). [3], [6]. The literature widely reports the use of different threshold as multiples of the baseline SD, for example, 1.4 SD [7], 2 SD [1] and three or more SD [6], [8], [9], [10], [11]. In all cases, the use of the baseline amplitude characteristics (mean, SD) to define the threshold follow the generic principles described widely in the quality literature and known as the Shewhart protocol [9], [12].

Following initial filtering (via the amplifiers) some researchers use visual detection methods on the raw/rectified data [7], [13], [14], [15]. Some algorithms however, use the duration for number of consecutive samples for which the signal must be maintained above the threshold. Such algorithms can only work on linear envelops created by secondary EMG signals processing. Since the methods used to create linear envelops also impact on the frequency content of the signal, the optimal combination of duration and amplitude thresholds can not be judged without reference to the EMG processing methods.

There are various definitions of baseline window and target windows [3] and there are physiological and statistical reasons why the researcher would select specific window cut-offs.

Initiating the detection algorithm at the target window or the physiological window limits the potential of detecting false onsets (see Fig. 1). The use of the physiological window is based on physiological limits and expectations of latency of response. For example, in assessing peroneal latency, Fernades et al. [16] used a conservative target window of 50 and 200 ms after the sudden inversion perturbation. Onsets that were detected outside this window reflected a response that is not consistent with a spinal reflex response to the perturbation stimulus. Similar specific physiological window techniques have been utilised for trunk muscle onsets [9], [17], [18] and the selection of up to ≈50 ms after the onset of the prime mover deltoid has been used for anticipatory postural adjustment physiological window [4], [5], [14], [15].

An alternative approach to selecting a baseline and target window combination is to utilise the complete signal. Abbink et al. [19] for example, suggested that using the whole signal profile assists in determining when the muscles bursts occur during rhythmical muscle activities (chewing). Santello and McDonagh [20] used the whole signal to normalise the amplitude. Strubb et al. [3] suggested that such single observations are simple but are less reliable suggesting a sliding window be used.

Most onset detection algorithms are robust when signals have a high signal to noise ratio (SNR) and fast rates of amplitude increase (see Fig. 1A). Since amplitude threshold detection methods utilise the baseline amplitude characteristics the subsequent threshold value is very sensitive to changes in the baseline. Low signal baseline and subsequent reduced threshold may detect earlier onsets and high variability may increase the threshold and therefore delay the onset detection. Clearly electro-cardiac (ECG) and movement artefacts and muscle activity bursts (pre-emption) during the baseline will increase the threshold for which an onset will be detected by an algorithm (see Fig. 1B) [6], [8]. The impact of the baseline variability is accentuated by variability in the rate of amplitude increase. These issues are problems in detecting feed forward responses in trunk muscles since the muscles are modulated with body sway, ECG artefact is present in some muscles and the feed forward responses are low amplitude graduated onsets. In such circumstances the error in the baseline may significantly impact on the detection of onsets during subtle trunk activation patterns.

The purpose of this study, therefore, was to examine two onset detection algorithms of the trunk muscle onsets before and after a trunk muscle fatiguing task. This was done by determining firstly, the algorithm’s capacity to detect onsets in muscles with ECG artefact and secondly the algorithms robustness as incremental noise was superimposed on the surface EMG signals.