Robust Muscle Activity Onset Detection Using an Unsupervised Electromyogram Learning Framework

Accurate muscle activity onset detection is an essential prerequisite for many applications of surface electromyogram (EMG). This study presents an unsupervised EMG learning framework based on a sequential Gaussian mixture model (GMM) to detect muscle activity onsets. The distribution of the logarithmic power of EMG signal was characterized by a two-component GMM in each frequency band, in which the two components respectively correspond to the posterior distribution of EMG burst and non-burst logarithmic powers. The parameter set of the GMM was sequentially estimated based on maximum likelihood, subject to constraints derived from the relationship between EMG burst and non-burst distributions. An optimal threshold for EMG burst/non-burst classification was determined using the GMM at each frequency band, and the final decision was obtained by a voting procedure. The proposed novel framework was applied to simulated and experimental surface EMG signals for muscle activity onset detection. Compared with conventional approaches, it demonstrated robust performance for low and changing signal to noise ratios in a dynamic environment. The framework is applicable for real-time implementation, and does not require the assumption of non EMG burst in the initial stage. Such features facilitate its practical application.     

A hierarchical clustering method for estimating copy number variation

Microarray technologies allow for simultaneous measurement of DNA copy number at thousands of positions in a genome. Gains and losses of DNA sequences reveal themselves through characteristic patterns of hybridization intensity. To identify change points along the chromosomes, we develop a marker clustering method which consists of 2 parts. First, a "circular clustering tree test statistic" attaches a statistic to each marker that measures the likelihood that it is a change point. Then construction of the marker statistics is followed by outlier detection approaches. The method provides a new way to build up a binary tree that can accurately capture change-point signals and is easy to perform. A simulation study shows good performance in change-point detection, and cancer cell line data are used to illustrate performance when regions of true copy number changes are known.     

Sample entropy analysis of surface EMG for improved muscle activity onset detection against spurious background spikes

Voluntary surface electromyogram (EMG) signal is sometimes contaminated by spurious background spikes of both physiological and extrinsic or accidental origins. A novel method of muscle activity onset detection against such spurious spikes was proposed in this study based primarily on the sample entropy (SampEn) analysis of the surface EMG. The method takes advantage of the nonlinear properties of the SampEn analysis to distinguish voluntary surface EMG signals from spurious background spikes in the complexity domain. To facilitate muscle activity onset detection, the SampEn analysis of surface EMG was first performed to highlight voluntary EMG activity while suppressing spurious background spikes. Then, a SampEn threshold was optimized for muscle activity onset detection. The performance of the proposed method was examined using both semi-synthetic and experimental surface EMG signals. The SampEn based methods effectively reduced the detection error induced by spurious background spikes and achieved improved performance over the methods relying on conventional amplitude thresholding or its extended version in the Teager Kaiser Energy domain.     

Muscle activation characteristics associated with differences in physiological tremor amplitude between the dominant and non-dominant hand

The aim of this study was to assess differences in physiological tremor amplitude of the hand between the dominant and non-dominant side of right-handed individuals. Mechanical loading of the hand and frequency analysis were used in an attempt to identify the physiological mechanisms involved in observed differences. Seventeen healthy right-handed adults participated in a single session where physiological tremor of the outstretched left and right hands was recorded under different loading conditions (0 g up to 5614 g). Physiological tremor amplitude was quantified through accelerometry and electromyographic (EMG) signals of wrist extensor and flexor muscles were also recorded. The main findings were: ～30% greater amplitude of fluctuations in acceleration for the non-dominant compared with the dominant hand, no difference in the frequency content of acceleration or demodulated EMG signals between dominant and non-dominant sides across all loads, and condition-dependent associations between the amplitude of fluctuations in acceleration and EMG amplitude and frequency content. These associations suggest a potential role of central modulation of neural activity to explain dominance-related differences in physiological tremor amplitude of the hand.     

Discrimination of Parkinsonian tremor from essential tremor by implementation of a wavelet-based soft-decision technique on EMG and accelerometer signals

A wavelet-decomposition with soft-decision algorithm is used to estimate an approximate power-spectral density (PSD) of both accelerometer and surface EMG signals for the purpose of discrimination of Parkinson tremor from essential tremor. A soft-decision wavelet-based PSD estimation is used with 256 bands for a signal sampled at 800 Hz. The sum of the entropy of the PSD in band 6 (7.8125–9.375 Hz) and band 11 (15.625–17.1875 Hz) is used as a classification factor. The data has been recorded for diagnostic purposes in the Department of Neurology of the University of Kiel, Germany. Two sets of data are used. The training set, which consists of 21 essential-tremor (ET) subjects and 19 Parkinson-disease (PD) subjects, is used to obtain the threshold value of the classification factor differentiating between the two subjects. The test data set, which consists of 20 ET and 20 PD subjects, is used to test the technique and evaluate its performance. A “voting” between three results obtained from accelerometer signal and two EMG signals is applied to obtain the final discrimination. A total accuracy of discrimination of 85% is obtained.     

Multi-channel surface EMG classification using support vector machines and signal-based wavelet optimization

The study proposes a method for supervised classification of multi-channel surface electromyographic signals with the aim of controlling myoelectric prostheses. The representation space is based on the discrete wavelet transform (DWT) of each recorded EMG signal using unconstrained parameterization of the mother wavelet. The classification is performed with a support vector machine (SVM) approach in a multi-channel representation space. The mother wavelet is optimized with the criterion of minimum classification error, as estimated from the learning signal set. The method was applied to the classification of six hand movements with recording of the surface EMG from eight locations over the forearm. Misclassification rate in six subjects using the eight channels was (mean ± S.D.) 4.7 ± 3.7% with the proposed approach while it was 11.1 ± 10.0% without wavelet optimization (Daubechies wavelet). The DWT and SVM can be implemented with fast algorithms, thus, the method is suitable for real-time implementation.     

Removing ECG contamination from EMG recordings: a comparison of ICA-based and other filtering procedures

Trunk muscle electromyography (EMG) is often contaminated by the electrocardiogram (ECG), which hampers data analysis and potentially yields misinterpretations. We propose the use of independent component analysis (ICA) for removing ECG contamination and compared it with other procedures previously developed to decontaminate EMG. To mimic realistic contamination while having uncontaminated reference signals, we employed EMG recordings from peripheral muscles with different activation patterns and superimposed distinct ECG signals that were recorded during rest at conventional locations for trunk muscle EMG. ICA decomposition was performed with and without a separately collected ECG signal as part of the data set and contaminated ICA modes representing ECG were identified automatically. Root mean squared relative errors and correlations between the linear envelopes of uncontaminated and contaminated EMG were calculated to assess filtering effects on EMG amplitude. Changes in spectral content were quantified via mean power frequencies. ICA-based filtering largely preserved the EMG's spectral content. Performance on amplitude measures was especially successful when a separate ECG recording was included. That is, the ICA-based filtering can produce excellent results when EMG and ECG are indeed statistically independent and when mode selection is flexibly adjusted to the data set under study.     

Detection of feedback in the central nervous system using system identification techniques

An analysis method to detect the presence of feedback between biological signals, particularly those associated with the central nervous system, is presented. The technique is based on recent results in the system identification literature involving the concept of a feedback free process. It may be applied to volume conducted signals such as EEG and EMG, as well as to neuronal spike trains through the use of a data transformation procedure. The utility of the technique is then demonstrated in a study of the relationship between Parkinsonian tremor and certain tremor cells found in the thalamus of Parkinsonian patients, using data collected during thalamotomies. The results obtained suggest that feedback mechanisms may be an important factor contributing to Parkinsonian tremor.This research was supported by the Natural Sciences and Engineering Research Council of Canada, The Medical Research Council of Canada, and Bell-Northern Research     

Corticomuscular Transmission of Tremor Signals by Propriospinal Neurons in Parkinson's Disease

Cortical oscillatory signals of single and double tremor frequencies act together to cause tremor in the peripheral limbs of patients with Parkinson''s disease (PD). But the corticospinal pathway that transmits the tremor signals has not been clarified, and how alternating bursts of antagonistic muscle activations are generated from the cortical oscillatory signals is not well understood. This paper investigates the plausible role of propriospinal neurons (PN) in C3–C4 in transmitting the cortical oscillatory signals to peripheral muscles. Kinematics data and surface electromyogram (EMG) of tremor in forearm were collected from PD patients. A PN network model was constructed based on known neurophysiological connections of PN. The cortical efferent signal of double tremor frequencies were integrated at the PN network, whose outputs drove the muscles of a virtual arm (VA) model to simulate tremor behaviors. The cortical efferent signal of single tremor frequency actuated muscle spindles. By comparing tremor data of PD patients and the results of model simulation, we examined two hypotheses regarding the corticospinal transmission of oscillatory signals in Parkinsonian tremor. Hypothesis I stated that the oscillatory cortical signals were transmitted via the mono-synaptic corticospinal pathways bypassing the PN network. The alternative hypothesis II stated that they were transmitted by way of PN multi-synaptic corticospinal pathway. Simulations indicated that without the PN network, the alternating burst patterns of antagonistic muscle EMGs could not be reliably generated, rejecting the first hypothesis. However, with the PN network, the alternating burst patterns of antagonist EMGs were naturally reproduced under all conditions of cortical oscillations. The results suggest that cortical commands of single and double tremor frequencies are further processed at PN to compute the alternating burst patterns in flexor and extensor muscles, and the neuromuscular dynamics demonstrated a frequency dependent damping on tremor, which may prevent tremor above 8 Hz to occur.     

Surface EMG: The issue of electrode location

This paper contributes to clarifying the conditions under which electrode position for surface EMG detection is critical and leads to estimates of EMG variables that are different from those obtained in other nearby locations. Whereas a number of previous works outline the need to avoid the innervation zone (or the muscle belly), many authors place electrodes in the central part or bulge of the muscle of interest where the innervation zone is likely to be. Computer simulations are presented to explain the effect of the innervation zone on amplitude, frequency and conduction velocity estimates from the signal and the need to avoid placing electrodes near it. Experimental signals recorded from some superficial muscles of the limbs and trunk (abductor pollicis brevis, flexor pollicis brevis, biceps, upper trapezius, vastus medialis, vastus lateralis) were processed providing support for the findings obtained from simulations. The use of multichannel techniques is recommended to estimate the location of the innervation zone and to properly choose the optimal position of the detection point(s) allowing meaningful estimates of EMG variables during movement analysis.     

Dynamical correlation of non-stationary signals in time domain—A comparative study

In a thorough study, the multitaper (MTM) and the extended continuous wavelet-transform (CWT) coherence-analysis methods were compared in terms of there application in determining the dynamics from the electroencephalogram (EEG) and electromyogram (EMG) signals of patients with Parkinsonian tremor. The main aim of the study in a biological point of view is to analyze whether the basic tremor frequency and its “first harmonic” frequency of Parkinsonian tremor are really harmonically related or are in fact distinct processes.The extension of the CWT is achieved by using a Morlet wavelet as the analysis window with an adjustable relative bandwidth which gives the flexibility in setting a desired frequency resolution. In order to obtain a perspective view of the two methods, they were applied to two different model signals to determine their actual threshold in detecting short-lived changes in the analysis of non-stationary signals and to determine their noise thresholds by adding external noise to the signals to test the reduction in coherence to be not merely due to the random fluctuations in stochastic signals. Beyond applying an autoregressive 2nd-order and a coupled van der Pol model system, however, also true EEG and EMG data from five Parkinson patients were used. The results were compared in terms of the time and frequency resolutions of these two methods, and it was determined that the multitaper method was able to detect reduction in power and coherence as short as 1 s. The extended CWT analysis only revealed gaps that were longer than 3 s.The time gaps in the coherence indicate the loss of connection between the cortex and muscle during the respective time intervals. This more accurate analysis of the MTM was also seen in the dynamical EEG–EMG coherence at the tremor frequency and its “first harmonic” of Parkinsonian patients.In terms of our “biological” aim, this shows distinct prevalence of the corticomuscular coupling at those frequencies over time. Applying this method to biological data reveals important aspects about their dynamics, e.g., the distinct dynamics between basic frequency and “first harmonic” frequency over time in Parkinsonian tremor.     

EMG and kinematics of normal subjects performing trunk flexion/extensions freely in space.

Ten normal subjects performed continuous trunk flexion/extensions (F/E) without any restraining apparatus at free, 3, 2.25 and 1.5 s periods and a fatiguing task of F/E at 1.5 s period during 45 s. Kinematics of the trunk was obtained with bilateral electromyographic (EMG) activity of the erector spinae (three levels), the abdominal oblique muscles and the rectus abdominis muscles. The free period chosen by the subjects was found to vary between 3.05 and 1.47 s. Lateral flexion of the spine was similar in each task but rotation about its longitudinal axis increased as the F/E period shortened. When left and right side EMG signals were grouped by level of recording, a significant difference in activity was found. Subjects who produced the slowest free F/E displayed larger fatigue indexes derived from the EMG signals for some of their back muscles than for other subjects. The flexion/relaxation phenomenon was considered present in a muscle if a level <10% of the maximum signal recorded during extension was detected. The phenomenon was found in >50% of the observations and occurred at a similar angle in each task. Kinematics and several characteristics of the EMG signals of the trunk were statistically independent of the speed of motion.     

EMG signal morphology and kinematic parameters in essential tremor and Parkinson’s disease patients

The aim of this work was to differentiate patients with essential tremor from patients with Parkinson’s disease. Electromyographic data from biceps brachii muscles and kinematic data from arms during isometric tension of the arms were measured from 17 patients with essential tremor, 35 patients with Parkinson’s disease and 40 healthy controls.The EMG signals were divided to smaller segments from which histograms were calculated. The histogram shape was analysed with a feature dimension reduction method, the principal component analysis, and the shape parameters were used to differentiate between different subject groups. Three parameters, RMS-amplitude, sample entropy and peak frequency were determined from the kinematic measurements of the arms.The height and the side differences of the histogram were the most effective for differentiating between essential tremor and Parkinson’s disease groups. The histogram parameters of patients with essential tremor were more similar to patients with Parkinson’s disease than healthy controls. With this method it was possible to discriminate 13/17 patients with essential tremor from 26/35 patients with Parkinson’s disease and 14/17 patients with essential tremor from 29/40 healthy controls. The kinematic parameters of patients with essential tremor were closer to parameters of patients with Parkinson’s disease compared to healthy controls. Combining EMG and kinematic analysis did not increase discrimination efficiency but provided more reliability to the discrimination of subject groups.     

Estimation of monopolar signals from sphincter muscles and removal of common mode interference

The detection of surface electromyogram (EMG) by multi-electrode systems is applied in many research studies. The signal is usually recorded by means of spatial filters (linear combination of the potential under at least two electrodes) with vanishing sum of weights. Nevertheless, more information could be extracted from monopolar signals measured with respect to a reference electrode away from the muscle. Under certain conditions, surface EMG signal along a curve parallel to the fibre path has zero mean (property approximately satisfied when EMG is sampled by an array of electrodes that covers the entire support of the signal in space). This property allows estimating monopolar from single differential (SD) signals by pseudoinversion of the matrix relating monopolar to SD signals. The method applies to EMG signals from the external anal sphincter muscle, recorded using a specific cylindrical probe with an array of electrodes located along the circular path of the fibres. The performance of the algorithm for the estimation of monopolar from SD signals is tested on simulated signals. The estimation error of monopolar signals decreases by increasing the number of channels. Using at least 12 electrodes, the estimation error is negligible. The method applies to single fibre action potentials, single motor unit action potentials, and interference signals.The same method can also be applied to reduce common mode interference from SD signals from muscles with rectilinear fibres. In this case, the last SD channel defined as the difference between the potentials of the last and the first electrodes must be recorded, so that the sum of all the SD signals vanishes. The SD signals estimated from the double differential signals by pseudoinvertion are free of common mode.     

The linear electrode array: a useful tool with many applications.

In this review we describe the basic principles of operation of linear electrode arrays for the detection of surface EMG signals, together with their most relevant current applications. A linear array of electrodes is a system which detects surface EMG signals in a number of points located along a line. A spatial filter is usually placed in each point for signal detection, so that the recording of EMG signals with linear arrays corresponds to the sampling in one spatial direction of a spatially filtered version of the potential distribution over the skin. Linear arrays provide indications on motor unit (MU) anatomical properties, such as the locations of the innervation zones and tendons, and the fiber length. Such systems allow the investigation of the properties of the volume conductor and its effect on surface detected signals. Moreover, linear arrays allow to estimate muscle fiber conduction velocity with a very low standard deviation of estimation (of the order of 0.1-0.2 m/s), thus providing reliable indications on muscle fiber membrane properties and their changes in time (for example with fatigue or during treatment). Conduction velocity can be estimated from a signal epoch (global estimate) or at the single MU level. In the latter case, MU action potentials are identified from the interference EMG signals and conduction velocity is estimated for each detected potential. In this way it is possible, in certain conditions, to investigate single MU control and conduction properties with a completely non-invasive approach. Linear arrays provide valuable information on the neuromuscular system properties and appear to be promising tools for applied studies and clinical research.     

Neurons as ideal change-point detectors

Every computational unit in the brain monitors incoming signals, instant by instant, for meaningful changes in the face of stochastic fluctuation. Recent studies have suggested that even a single neuron can detect changes in noisy signals. In this paper, we demonstrate that a single leaky integrate-and-fire neuron can achieve change-point detection close to that of theoretical optimal, for uniform-rate process, functions even better than a Bayes-optimal algorithm when the underlying rate deviates from a presumed uniform rate process. Given a reasonable number of synaptic connections (order 10⁴) and the rate of the input spike train, the values of the membrane time constant and the threshold found for optimizing change-point detection are close to those seen in biological neurons. These findings imply that biological neurons could act as sophisticated change-point detectors.     

Fetal ECG extraction during labor using an adaptive maternal beat subtraction technique.

Fetal ECG (FECG) monitoring using abdominal maternal signals is a non-invasive technique that allows early detection of changes in fetal wellbeing. Several other signal components have stronger energy than the FECG, the most important being maternal ECG (MECG) and, especially during labor, uterine EMG. This study proposes a new method to subtract MECG after detecting and removing abdominal signal segments with high-amplitude variations due to uterine contractions. The method removes MECG from abdominal signals using an approximation of the current MECG segment based on a linear combination of previous MECG segments aligned on the R-peak. The coefficients of the linear model are computed so that the squared error of the approximation over the whole current segment is minimized. Abdominal signal segments strongly affected by uterine contractions are detected by applying median filtering. The methods proposed are tested on real abdominal data recorded during labor, with FECG recorded using scalp electrodes synchronously recorded for comparison.     

Persistence of improvements in postural strategies following motor control training in people with recurrent low back pain.

This study investigated long-term effects of training on postural control using the model of deficits in activation of transversus abdominis (TrA) in people with recurrent low back pain (LBP). Nine volunteers with LBP attended four sessions for assessment and/or training (initial, two weeks, four weeks and six months). Training of repeated isolated voluntary TrA contractions were performed at the initial and two-week session with feedback from real-time ultrasound imaging. Home program involved training twice daily for four weeks. Electromyographic activity (EMG) of trunk and deltoid muscles was recorded with surface and fine-wire electrodes. Rapid arm movement and walking were performed at each session, and immediately after training on the first two sessions. Onset of trunk muscle activation relative to prime mover deltoid during arm movements, and the coefficient of variation (CV) of EMG during averaged gait cycle were calculated. Over four weeks of training, onset of TrA EMG was earlier during arm movements and CV of TrA EMG was reduced (consistent with more sustained EMG activity). Changes were retained at six months follow-up (p<0.05). These results show persistence of motor control changes following training and demonstrate that this training approach leads to motor learning of automatic postural control strategies.     

Paraspinal muscle reflex dynamics

Neuromuscular control of spinal stability may be represented as a control system wherein the paraspinal muscle reflex acts as feedback response to kinetic and kinematic disturbances of the trunk. The influence of preparatory muscle recruitment for the control of spinal stability has been previously examined, but there are few reported studies that characterize paraspinal reflex gain as feedback response. In the current study, the input-output dynamics of paraspinal reflexes were quantified by means of the impulse response function (IRF), with trunk perturbation force representing the input signal and EMG the output signal. Surface EMGs were collected from the trunk muscles in response to a brief anteriorly directed impact force applied to the trunk of healthy participants. Reflex behavior was measured in response to three levels of force impulse, 6.1, 9.2 and 12.0 Ns, and two different levels of external trunk flexion preload, 0 and 110 N anterior force. Reflex EMG was quantifiable in response to 91% of the perturbations. Mean reflex onset latency was 30.7+/-21.3 ms and reflex amplitude increased with perturbation amplitude. Impulse response function gain, G(IRF), was defined as the peak amplitude of the measured IRF and provided a consistent measure of response behavior. EMG reflex amplitude and G(IRF) increased with force impulse. Mean G(IRF) was 2.27+/-1.31% MVC/Ns and demonstrated declining trend with flexion preload. Results agree with a simple systems model of the neuromechanical feedback behavior. The relative contribution of the reflex dynamics to spinal stability must be investigated in future research.