Dynamic muscle force predictions from EMG: an artificial neural network approach

EMG signals of dynamically contracting muscle have never been used to predict experimentally known muscle forces across subjects. Here, we use an artificial neural network (ANN) approach to first derive an EMG–force relationship from a subset of experimentally determined EMGs and muscle forces; second, we use this relationship to predict individual muscle forces for different contractile conditions and in subjects whose EMG and force data were not used in the derivation of the EMG–force relationship; and third, we validate the predicted muscle forces against the known forces recorded in vivo. EMG and muscle forces were recorded from the cat soleus for a variety of locomotor conditions giving a data base from three subjects, four locomotor conditions, and 8–16 steps per subject and condition. Considering the conceptual differences in the tasks investigated (e.g. slow walking vs. trotting), the intra-subject results obtained here are superior to those published previously, even though the approach did not require a muscle model or the instantaneous contractile conditions as input for the force predictions. The inter-subject results are the first of this kind to be presented in the literature and they typically gave cross-correlation coefficients between actual and predicted forces of >0.90 and root mean square errors of <15%, thus they were considered excellent.

From the results of this study, it was concluded that ANNs represent a powerful tool to capture the essential features of EMG–force relationships of dynamically contracting muscle, and that ANNs might be used widely to predict muscle forces based on EMG signals.     

Fatigue and muscle activation during submaximal elbow flexion in children with cerebral palsy

The purpose of this study was to investigate whether children with cerebral palsy (CP), like typically developing peers, would compensate for muscle fatigue by recruiting additional motor units during a sustained low force contraction until task failure.Twelve children with CP and 17 typically developing peers performed one submaximal isometric elbow flexion contraction until the task could no longer be sustained at on average 25% (range 10–35%) of their maximal voluntary torque. Meanwhile surface electromyography (EMG) was measured from the biceps brachii and triceps brachii, and acceleration variations of the forearm were detected by an accelerometer. Slopes of the change in EMG amplitude and median frequency and accelerometer variation during time normalised to their initial values were calculated.Strength and time to task failure were similar in both groups. Children with CP exhibited a lower increase in EMG amplitude of the biceps brachii and triceps brachii during the course of the sustained elbow flexion task, while there were no significant group differences in median frequency decrease or acceleration variation increase. This indicates that children with CP do not compensate muscle fatigue with recruitment of additional motor units during sustained low force contractions.     

Eccentric exercise increases EMG amplitude and force fluctuations during submaximal contractions of elbow flexor muscles.

The purpose of this study was to determine the effect of eccentric exercise on the ability to exert steady submaximal forces with muscles that cross the elbow joint. Eight subjects performed two tasks requiring isometric contraction of the right elbow flexors: a maximum voluntary contraction (MVC) and a constant-force task at four submaximal target forces (5, 20, 35, 50% MVC) while electromyography (EMG) was recorded from elbow flexor and extensor muscles. These tasks were performed before, after, and 24 h after a period of eccentric (fatigue and muscle damage) or concentric exercise (fatigue only). MVC force declined after eccentric exercise (45% decline) and remained depressed 24 h later (24%), whereas the reduced force after concentric exercise (22%) fully recovered the following day. EMG amplitude during the submaximal contractions increased in all elbow flexor muscles after eccentric exercise, with the greatest change in the biceps brachii at low forces (3-4 times larger at 5 and 20% MVC) and in the brachialis muscle at moderate forces (2 times larger at 35 and 50% MVC). Eccentric exercise resulted in a twofold increase in coactivation of the triceps brachii muscle during all submaximal contractions. Force fluctuations were larger after eccentric exercise, particularly at low forces (3-4 times larger at 5% MVC, 2 times larger at 50% MVC), with a twofold increase in physiological tremor at 8-12 Hz. These data indicate that eccentric exercise results in impaired motor control and altered neural drive to elbow flexor muscles, particularly at low forces, suggesting altered motor unit activation after eccentric exercise.     

Functional Corticospinal Projections from Human Supplementary Motor Area Revealed by Corticomuscular Coherence during Precise Grip Force Control

The purpose of the present study was to investigate whether corticospinal projections from human supplementary motor area (SMA) are functional during precise force control with the precision grip (thumb-index opposition). Since beta band corticomuscular coherence (CMC) is well-accepted to reflect efferent corticospinal transmission, we analyzed the beta band CMC obtained with simultaneous recording of electroencephalographic (EEG) and electromyographic (EMG) signals. Subjects performed a bimanual precise visuomotor force tracking task by applying isometric low grip forces with their right hand precision grip on a custom device with strain gauges. Concurrently, they held the device with their left hand precision grip, producing similar grip forces but without any precision constraints, to relieve the right hand. Some subjects also participated in a unimanual control condition in which they performed the task with only the right hand precision grip while the device was held by a mechanical grip. We analyzed whole scalp topographies of beta band CMC between 64 EEG channels and 4 EMG intrinsic hand muscles, 2 for each hand. To compare the different topographies, we performed non-parametric statistical tests based on spatio-spectral clustering. For the right hand, we obtained significant beta band CMC over the contralateral M1 region as well as over the SMA region during static force contraction periods. For the left hand, however, beta band CMC was only found over the contralateral M1. By comparing unimanual and bimanual conditions for right hand muscles, no significant difference was found on beta band CMC over M1 and SMA. We conclude that the beta band CMC found over SMA for right hand muscles results from the precision constraints and not from the bimanual aspect of the task. The result of the present study strongly suggests that the corticospinal projections from human SMA become functional when high precision force control is required.     

Phonomyogram and electromyogram relationships with isometric force reinvestigated in man

Contracting muscle generates sounds which can be recorded easily by means of a microphone. To determine if a phomomyogram (PMG) can be used to monitor muscle force, a comparison was made between simultaneous recordings of PMG and monopolar electromyogram (EMG) from the isometrically contracting biceps brachii muscle and the external flexion force. Locations of the monopolar electrode and microphone were identified in relation to the motor point. Whatever the recording site, PMG amplitude was proportional to EMG amplitude and both showed a quadratic relationship to muscle force. Changes in the PMG spectrum with force were similar to those in EMG, i.e. the mean power frequency increased up to about 30% maximal voluntary contraction and then reached a plateau. Despite a slightly higher variability, PMG was shown to be a valid index of muscular isometric force. At the same force, the amplitude of both PMG and EMG was lower in the prone than in the supine position of the hand. This result indicated a selective recording of biceps brachii muscle activity.     

Alpha Band Cortico-Muscular Coherence Occurs in Healthy Individuals during Mechanically-Induced Tremor

The present work aimed at investigating the effects of mechanically amplified tremor on cortico-muscular coherence (CMC) in the alpha band. The study of CMC in this specific band is of particular interest because this coherence is usually absent in healthy individuals and it is an aberrant feature in patients affected by pathological tremors; understanding its mechanisms is therefore important. Thirteen healthy volunteers (23±4 years) performed elbow flexor sustained contractions both against a spring load and in isometric conditions at 20% of maximal voluntary isometric contraction (MVC). Spring stiffness was selected to induce instability in the stretch reflex servo loop. 64 EEG channels, surface EMG from the biceps brachii muscle and force were simultaneously recorded. Contractions against the spring resulted in greater fluctuations of the force signal and EMG amplitude compared to isometric conditions (p<.05). During isometric contractions CMC was systematically found in the beta band and sporadically observed in the alpha band. However, during the contractions against the spring load, CMC in the alpha band was observed in 12 out of 13 volunteers. Partial directed coherence (PDC) revealed an increased information flow in the EMG to EEG direction in the alpha band (p<.05). Therefore, coherence in the alpha band between the sensory-motor cortex and the biceps brachii muscle can be systematically induced in healthy individuals by mechanically amplifying tremor. The increased information flow in the EMG to EEG direction may reflect enhanced afferent activity from the muscle spindles. These results may contribute to the understanding of the presence of alpha band CMC in tremor related pathologies by suggesting that the origin of this phenomenon may not only be at cortical level but may also be affected by spinal circuit loops.     

A comparison of optimisation methods and knee joint degrees of freedom on muscle force predictions during single-leg hop landings

The aim of this paper was to compare the effect of different optimisation methods and different knee joint degrees of freedom (DOF) on muscle force predictions during a single legged hop. Nineteen subjects performed single-legged hopping manoeuvres and subject-specific musculoskeletal models were developed to predict muscle forces during the movement. Muscle forces were predicted using static optimisation (SO) and computed muscle control (CMC) methods using either 1 or 3 DOF knee joint models. All sagittal and transverse plane joint angles calculated using inverse kinematics or CMC in a 1 DOF or 3 DOF knee were well-matched (RMS error<3°). Biarticular muscles (hamstrings, rectus femoris and gastrocnemius) showed more differences in muscle force profiles when comparing between the different muscle prediction approaches where these muscles showed larger time delays for many of the comparisons. The muscle force magnitudes of vasti, gluteus maximus and gluteus medius were not greatly influenced by the choice of muscle force prediction method with low normalised root mean squared errors (<48%) observed in most comparisons. We conclude that SO and CMC can be used to predict lower-limb muscle co-contraction during hopping movements. However, care must be taken in interpreting the magnitude of force predicted in the biarticular muscles and the soleus, especially when using a 1 DOF knee. Despite this limitation, given that SO is a more robust and computationally efficient method for predicting muscle forces than CMC, we suggest that SO can be used in conjunction with musculoskeletal models that have a 1 or 3 DOF knee joint to study the relative differences and the role of muscles during hopping activities in future studies.     

An open source lower limb model: Hip joint validation

Musculoskeletal lower limb models have been shown to be able to predict hip contact forces (HCFs) that are comparable to in vivo measurements obtained from instrumented prostheses. However, the muscle recruitment predicted by these models does not necessarily compare well to measured electromyographic (EMG) signals. In order to verify if it is possible to accurately estimate HCFs from muscle force patterns consistent with EMG measurements, a lower limb model based on a published anatomical dataset (Klein Horsman et al., 2007. Clinical Biomechanics. 22, 239-247) has been implemented in the open source software OpenSim. A cycle-to-cycle hip joint validation was conducted against HCFs recorded during gait and stair climbing trials of four arthroplasty patients (Bergmann et al., 2001. Journal of Biomechanics. 34, 859-871). Hip joint muscle tensions were estimated by minimizing a polynomial function of the muscle forces. The resulting muscle activation patterns obtained by assessing multiple powers of the objective function were compared against EMG profiles from the literature. Calculated HCFs denoted a tendency to monotonically increase their magnitude when raising the power of the objective function; the best estimation obtained from muscle forces consistent with experimental EMG profiles was found when a quadratic objective function was minimized (average overestimation at experimental peak frame: 10.1% for walking, 7.8% for stair climbing). The lower limb model can produce appropriate balanced sets of muscle forces and joint contact forces that can be used in a range of applications requiring accurate quantification of both. The developed model is available at the website https://simtk.org/home/low_limb_london.     

Total trunk muscle force and spinal compression are lower in asymmetric moments as compared to pure extension moments.

The aim of the present study was to test the assumption that asymmetric trunk loading requires a higher total muscle force and consequently entails a higher compression forces on the spine as compared to symmetric loading. When the trunk musculature is modelled in sufficient detail, optimisation shows that there is no mechanical necessity for an increase in total muscle force (or compression force) with task asymmetry. A physiologically based optimisation does also not predict an increase in total muscle force or spinal loading with asymmetry. EMG data on 14 trunk muscles collected in eight subjects showed antagonistic coactivity to be present in both conditions. However, estimates of total muscle force based on the EMG were lower when producing an asymmetric moment. In conclusion, producing an asymmetric moment appears to cause slightly lower forces on the lumbosacral joint as compared to a symmetric moment. Only lateral shear forces increase with asymmetry but these remain well below failure levels.     

Less is more: high pass filtering, to remove up to 99% of the surface EMG signal power, improves EMG-based biceps brachii muscle force estimates.

It is generally assumed that raw surface EMG (sEMG) should be high pass filtered with cutoffs of 10-30 Hz to remove motion artifact before subsequent processing to estimate muscle force. The purpose of the current study was to explore the benefits of filtering out much of the raw sEMG signal when attempting to estimate accurate muscle forces. Twenty-five subjects were studied as they performed rapid static, anisotonic contractions of the biceps brachii. Biceps force was estimated (as a percentage of maximum) based on forces recorded at the wrist. An iterative approach was used to process the sEMG from the biceps brachii, using progressively greater high pass cutoff frequencies (20-440 Hz in steps of 30 Hz) with first and sixth order filters, as well as signal whitening, to determine the effects on the accuracy of EMG-based biceps force estimates. The results indicate that removing up to 99% of the raw sEMG signal power resulted in significant and substantial improvements in biceps force estimates. These findings challenge previous assumptions that the raw sEMG signal power between about 20 and 500 Hz should used when estimating muscle force. For the purposes of force prediction, it appears that a much smaller, high band of sEMG frequencies may be associated with force and the remainder of the spectrum has little relevance for force estimation.     

Methodological aspects of SEMG recordings for force estimation – A tutorial and review

Insight into the magnitude of muscle forces is important in biomechanics research, for example because muscle forces are the main determinants of joint loading. Unfortunately muscle forces cannot be calculated directly and can only be measured using invasive procedures. Therefore, estimates of muscle force based on surface EMG measurements are frequently used. This review discusses the problems associated with surface EMG in muscle force estimation and the solutions that novel methodological developments provide to this problem. First, some basic aspects of muscle activity and EMG are reviewed and related to EMG amplitude estimation. The main methodological issues in EMG amplitude estimation are precision and representativeness. Lack of precision arises directly from the stochastic nature of the EMG signal as the summation of a series of randomly occurring polyphasic motor unit potentials and the resulting random constructive and destructive (phase cancellation) superimpositions. Representativeness is an issue due the structural and functional heterogeneity of muscles. Novel methods, i.e. multi-channel monopolar EMG and high-pass filtering or whitening of conventional bipolar EMG allow substantially less variable estimates of the EMG amplitude and yield better estimates of muscle force by (1) reducing effects of phase cancellation, and (2) adequate representation of the heterogeneous activity of motor units within a muscle. With such methods, highly accurate predictions of force, even of the minute force fluctuations that occur during an isometric and isotonic contraction have been achieved. For dynamic contractions, EMG-based force estimates are confounded by the effects of muscle length and contraction velocity on force producing capacity. These contractions require EMG amplitude estimates to be combined with modeling of muscle contraction dynamics to achieve valid force predictions.     

Effects of mechanical assistance on muscle activity and motor performance during isometric elbow flexion

Mechanical assistance on joint movement is generally beneficial; however, its effects on cooperative performance and muscle activity needs to be further explored. This study examined how motor performance and muscle activity are altered when mechanical assistance is provided during isometric force control of ramp-down and hold phases. Thirteen right-handed participants (age: 24.7 ± 1.8 years) performed trajectory tracking tasks. Participants were asked to maintain the reference magnitude of 47 N (REF) during isometric elbow flexion. The force was released to a step-down magnitude of either 75% REF or 50% REF and maintained, with and without mechanical assistance. The ramp-down durations of force release were set to 0.5, 2.5, or 5.0 s. Throughout the experiment, we measured the following: (1) the force output using load cells to compute force variability and overshoot ratio; (2) peak perturbation on the elbow movement using an accelerometer; (3) the surface electromyography (sEMG) from biceps brachii and triceps brachii muscles; and (4) EMG oscillation from the biceps brachii muscle in the bandwidth of 15–45 Hz. Our results indicated that mechanical assistance, which involved greater peak perturbation, demonstrated lower force variability than non-assistance (p < 0.01), while EMG oscillation in the biceps brachii muscle from 15 to 45 Hz was increased (p < 0.05). These findings imply that if assistive force is provided during isometric force control, the central nervous system actively tries to stabilize motor performance by controlling specific motor unit activity in the agonist muscle.     

Dynamic joint and muscle forces during knee isokinetic exercise

Isokinetic exercise has been commonly used in knee rehabilitation, conditioning and research in the past two decades. Although many investigators have used various experimental and theoretical approaches to study the muscle and joint force involved in isokinetic knee extension and flexion exercises, only a few of these studies have actually distinguished between the tibiofemoral joint forces and muscle forces. Therefore, the objective of this study was to specify, via an eletromyography(EMG)-driven muscle force model of the knee, the magnitude of the tibiofemoral joint and muscle forces acting during isokinetic knee extension and flexion exercises. Fifteen subjects ranging from 21 to 36 years of age volunteered to participate in this study. A Kin Com exercise machine (Chattecx Corporation, Chattanooga, TN, U.S.A.) was used as the loading device. An EMG-driven muscle force model was used to predict muscle forces, and a biomechanical model was used to analyze two knee joint constraint forces; compression and shear force. The methods used in this study were shown to be valid and reliable (r > 0.84 andp < 0.05). The effects on the tibiofemoral joint force during knee isokinetic exercises were compared with several functional activities that were investigated by earlier researchers. The muscle forces generated during knee isokinetic exercise were also obtained. Based on the findings obtained in this study, several therapeutic justifications for knee rehabilitation are proposed.     

Effects of EMG processing on biomechanical models of muscle joint systems: sensitivity of trunk muscle moments, spinal forces, and stability

Biomechanical models are in use to estimate parameters such as contact forces and stability at various joints. In one class of these models, surface electromyography (EMG) is used to address the problem of mechanical indeterminacy such that individual muscle activation patterns are accounted for. Unfortunately, because of the stochastical properties of EMG signals, EMG based estimates of muscle force suffer from substantial estimation errors. Recent studies have shown that improvements in muscle force estimation can be achieved through adequate EMG processing, specifically whitening and high-pass (HP) filtering of the signals. The aim of this paper is to determine the effect of such processing on outcomes of a biomechanical model of the lumbosacral joint and surrounding musculature. Goodness of fit of estimated muscle moments to net moments and also estimated joint stability significantly increased with increasing cut-off frequencies in HP filtering, whereas no effect on joint contact forces was found. Whitening resulted in moment estimations comparable to those obtained from optimal HP filtering with cut-off frequencies over 250 Hz. Moreover, compared to HP filtering, whitening led to a further increase in estimated joint-stability. Based on theoretical models and on our experimental results, we hypothesize that the processing leads to an increase in pick-up area. This then would explain the improvements from a better balance between deep and superficial motor unit contributions to the signal.     

Experimental muscle pain increases mechanomyographic signal activity during sub-maximal isometric contractions.

This study was designed to investigate the local effect of experimental muscle pain on the MMG and the surface EMG during a range of sub-maximal isometric contractions. Muscle pain was induced by injections of hypertonic saline into the biceps brachii muscle in 12 subjects. Injections of isotonic saline served as a control. Pain intensity and location, MMG and surface EMG from the biceps brachii were assessed during static isometric (0%, 10%, 30%, 50% and, 70% of the maximal voluntary contraction) and ramp isometric (0-50% of the maximal voluntary contraction) elbow flexions. MMG and surface EMG signals were analyzed in the time and frequency domain. Experimentally induced muscle pain induced an increase in root mean square values of the MMG signal while no changes were observed in the surface EMG. Most likely this increase reflects changes in the mechanical contractile properties of the muscle and indicates compensatory mechanisms, i.e. decreased firing rate and increased twitch force to maintain a constant force output in presence of experimental muscle pain. Under well-controlled conditions, MMG recordings may be more sensitive than surface EMG recordings and clinically useful for detecting non-invasively increased muscle mechanical contributions during muscle pain conditions.     

Towards optimal multi-channel EMG electrode configurations in muscle force estimation: a high density EMG study.

Surface EMG is an important tool in biomechanics, kinesiology and neurophysiology. In neurophysiology the concept of high-density EMG (HD-EMG), using two dimensional electrode grids, was developed for the measurement of spatiotemporal activation patterns of the underlying muscle and its motor units (MU). The aim of this paper was to determine, with the aid of a HD-EMG grid, the relative importance of a number of electrode sensor configurations for optimizing muscle force estimation. Sensor configurations are distinguished in two categories. The first category concerns dimensions: the size of a single electrode and the inter electrode distance (IED). The second category concerns the sensor's spatial distribution: the total area from which signals are obtained (collection surface) and the number of electrodes per cm(2) (collection density). Eleven subjects performed isometric arm extensions at three elbow angles and three contraction levels. Surface-EMG from the triceps brachii muscle and the external force at the wrist were measured. Compared to a single conventional bipolar electrode pair, the force estimation quality improved by about 30% when using HD-EMG. Among the sensor configurations, the collection surface alone appeared to be responsible for the major part of the EMG based force estimation quality by improving it with 25%.     

Quantitative assessment of upper limb muscle fatigue depending on the conditions of repetitive task load

The aim of this study was to discriminate fatigue of upper limb muscles depending on the external load, through the development and analysis of a muscle fatigue index. Muscle fatigue is expressed by a fatigue index based on an amplitude parameter (calculated in the time domain) and a fatigue index based on a frequency parameter (a parameter calculated in the frequency domain). The fatigue index involves a regression function that describes changes in the EMG signal parameter, time elapsing before muscle fatigue and the probability of specific trends in changes in EMG parameters for the population under study.

The experimental study covered a group of 10 young men. During the study, they exerted force at a specific level and for a specific time in 12 load variants. During the study, EMG signals from four muscles of the upper limb were recorded (trapezius pars descendents, biceps brachii caput breve, extensor carpi radialis brevis, flexor carpi ulnaris). For each variant and for each examined muscles, the value of the fatigue index was calculated. Values of that index quantitatively expressed fatigue of a specific muscle in a specific load variant.

A statistical analysis indicated variation in the fatigue of the biceps brachii caput breve, extensor carpi radialis brevis, and flexor carpi ulnaris muscles depending on the external load (load variant) according to the task performed with the upper limb.

The study demonstrated usefulness of the fatigue index in expressing quantitatively muscle fatigue and in discriminating muscle fatigue depending on the external load.     

Electromyographic measures of muscle activation and changes in muscle architecture of human elbow flexors during fatiguing contractions.

The study compared changes in intramuscular and surface recordings of EMG amplitude with ultrasound measures of muscle architecture of the elbow flexors during a submaximal isometric contraction. Ten subjects performed a fatiguing contraction to task failure at 20% of maximal voluntary contraction force. EMG activity was recorded in biceps brachii, brachialis, and brachioradialis muscles using intramuscular and surface electrodes. The rates of increase in the amplitude of the surface EMG for the long and short heads of biceps brachii and brachioradialis were greater than those for the intramuscular recordings measured at different depths. The amplitude of the intramuscular recordings from three muscles increased at a similar rate (P = 0.13), as did the amplitude of the three surface recordings from two muscles (P = 0.83). The increases in brachialis thickness (27.7 +/- 5.7 to 30.9 +/- 3.5 mm; P < 0.05) and pennation angle (10.9 +/- 3.5 to 16.5 +/- 4.8 degrees ; P = 0.003) were not associated with the increase in intramuscular EMG amplitude (P > 0.58). The increase in brachioradialis thickness (22.8 +/- 4.8 to 25.5 +/- 3.4 mm; P = 0.0075) was associated with the increase in the amplitude for one of two intramuscular EMG signals (P = 0.007, r = 0.79). The time to failure was more strongly associated with the rate of increase in the amplitude of the surface EMG than that for the intramuscular EMG, which suggests that the surface measurement provides a more appropriate measure of the change in muscle activation during a fatiguing contraction.