Influence of interelectrode distance and force level on the spectral parameters of surface electromyographic recordings from the lumbar muscles

Real-Time Forecasting of sEMG Features for Trunk Muscle Fatigue Using Machine Learning

OBJECTIVE

Several features of the surface electromyography (sEMG) signal are related to muscle activity and fatigue. However, the time-evolution of these features are non-stationary and vary between subjects. The aim of this study is to investigate the use of adaptive algorithms to forecast sEMG feature of the trunk muscles.

METHODS

Shallow models and a deep convolutional neural network (CNN) were used to simultaneously learn and forecast 5 common sEMG features in real-time to provide tailored predictions. This was investigated for: up to a 25 second horizon; for 14 different muscles in the trunk; across 13 healthy subjects; while they were performing various exercises.

RESULTS

The CNN was able to forecast 25 seconds ahead of time, with 6.88% mean absolute percentage error and 3.72% standard deviation of absolute percentage error, across all the features. Moreover, the CNN outperforms the best shallow model in terms of a figure of merit combining accuracy and precision by at least 30% for all the 5 features.

CONCLUSION

Even though the sEMG features are non-stationary and vary between subjects, adaptive learning and forecasting, especially using CNNs, can provide accurate and precise forecasts across a range of physical activities.

SIGNIFICANCE

The proposed models provide the groundwork for a wearable device which can forecast muscle fatigue in the trunk, so as to potentially prevent low back pain. Additionally, the explicit real-time forecasting of sEMG features provides a general model which can be applied to many applications of muscle activity monitoring, which helps practitioners and physiotherapists improve therapy.

The influence of confounding factors on the relationship between muscle contraction level and MF and MPF values of EMG signal: a review

The purpose of this article is to gather results of studies on the relationship between median frequency (MF) and mean power frequency (MPF) and the level of muscle contraction, and to use those results to discuss the differences in the trends according to factors related to measurement technique and subject. Twenty-one studies with 63 cases for upper limb muscles and nine studies with 31 cases for lower limb muscles were analysed. Most results showed an increase in parameters with an increased level of muscle contraction, only some studies showed a decrease. The influence on parameters of the level of muscle contraction and factors such as subjects, type of contraction, muscle length and electrodes was analysed for each muscle. It was concluded that when analysing the influence of different factors on MF and MPF, because those factors interact they should be considered together, not separately.

Influência da distância intereletrodos e da cadência de movimento no domínio da frequência do sinal de EMG de superfície

A eletromiografia de superfície (sEMG), apesar de amplamente utilizada em investigações biomecânicas, ainda apresenta inúmeros questionamentos sobre a influência das distâncias intereletrodos (DIE) na morfologia do sinal, principalmente em contrações isotônicas. Logo, muitos dos trabalhos desenvolvidos ainda se limitam ao âmbito do laboratório de pesquisa, onde é possível estabelecer maior controle nos protocolos de registro e análise, o que não é comumente observado na prática clínico-desportiva. Dessa forma, o objetivo do estudo foi examinar os efeitos de dois protocolos de colocação de eletrodos e a realização de contrações isotônicas no domínio da frequência do sinal de sEMG. Quinze sujeitos do sexo masculino (idade: 22,8 ± 3,5 anos), todos destros, realizaram contrações dinâmicas do bíceps braquial direito com carga estimada em 20% da contração voluntária máxima em três diferentes cadências (30, 45, 60bpm). Os sinais de sEMG foram registrados por meio de dois canais, cujas DIEs foram de 4,2 e 13cm, respectivamente. A avaliação dos sinais de sEMG foi baseada na frequência mediana do espectro de potencial do sinal, calculado via transformada rápida de Fourier. A DIE e a cadência foram definidas como fatores (ANOVA two-way; α = 0,05). Não foram observadas diferenças estatísticas e qualquer interação entre ambos os fatores nas três cadências (P > 0,05). Sugere-se que, independentemente da distância utilizada entre os eletrodos, uma investigação no domínio da frequência do sinal de sEMG em tarefas dinâmicas seja evitada, mesmo a partir de DIEs reduzidas, como é sugerido pela literatura, dado que variações no torque e no comprimento muscular podem corromper o sinal e, portanto, sua interpretação.

Task dependency in back muscle fatigue--correlations between two test methods

BACKGROUND

Various test methods which engage the back muscles in different tasks have been used in studies of back muscle fatigue with electromyography. The present objective was to study task dependency in lumbar muscle fatigue by comparing two test methods.

METHODS

In this cross-sectional study, 22 healthy subjects performed a seated (45s) and a prone test (to the limit of endurance) of back muscle fatigue in randomised order. Fatigue of the lumbar muscles was assessed using electromyography spectral variables and ratings of back muscle fatigue (Borg scale). Linear regression of the median frequency during contraction, and conventional statistical tests of group differences and correlations were used.

FINDINGS

Significant differences (P<0.001) between the seated and the prone test were found for the initial median frequency, the slope, the median frequency decrease during the whole contraction, and for the ratings. However, correlation coefficients between the seated and the prone test were low for the median frequency decrease (r=0.42), absent for the slopes of median frequency (r=-0.08), higher for the Borg ratings (r(s)=0.51; P<0.05) and highest for the initial median frequency (r=0.69; P<0.05). Within each test, correlations between the Borg ratings and the electromyography variables were essentially absent (r<0.19).

INTERPRETATION

Electromyography variables assessed in one type of task in a fatiguing test may not be valid for other types of fatiguing tasks, for example in daily life work situations. Thus task dependency has to be considered when using surface electromyography in determining lumbar muscle fatigue. Ratings of fatigue, however, seem to be less task dependent than the electromyography variables.

EMG sensor location: Does it influence the ability to detect differences in muscle contraction conditions?

Even though it is well known that electromyography (EMG) characteristics are influenced by electrode placement it is common to use a single pair of sensors per muscle for EMG. This study was designed to determine if the ability to distinguish between contraction conditions was influenced by sensor location. Subjects (n = 10; 27+/-5.3 years; 82+/-13.4 kg; 178+/-7.1 cm) completed six elbow flexor conditions: three isometric contraction intensities (100% maximum effort, 80%, 50%) and three isotonic contraction intensities (heavy weight, 80% and 50% of the weight). Three pairs of electrodes were placed centrally, medially and laterally on the biceps brachii belly in line with the muscle fibers. Isometric contractions were held for 5s, with the middle 3 s analyzed. Isotonic exercises included five repetitions of elbow flexion-extension, with the middle three repetitions analyzed. Average EMG (EMG(AVG)), root mean square EMG (EMG(RMS)) and mean power frequency (MPF) were calculated for each extracted data set. Dependent variables were analyzed using 2 (contraction type) x 3 (intensity) repeated measures ANOVAs per sensor. EMG(AVG) was influenced by the interaction between contraction type and intensity for all sensors (p < 0.05). EMG(RMS) as well as MPF were influenced by the interaction between contraction type and intensity for the lateral and central leads (p < 0.05) but not the medial leads (p > 0.05). Different conclusions could have been reached from the same experiment due to different sensor locations. These differences were primarily related to comparing contraction types (i.e., isotonic vs. isometric).

On the errors in assessment of severity of involuntary movements using surface EMG

Surface electromyogram (SEMG) is a useful tool to depict involuntary movements, but evaluation of the intensity of such movements with SEMG over multiple recording instances requires awareness of factors influencing quantified SEMG signals. We investigated the differences in the amplitude of SEMGs due to electrode displacement in isometric voluntary contraction in the upper arm, forearm and lower leg in 8 healthy men. The SEMGs of gross muscle activity simultaneously recorded with 4 electrode pairs from the agonist and antagonist sides in 3 displacement conditions with respect to parallel position, interelectrode distance, and rotation were compared. The amount of EMG integration (equivalent to the average SEMG amplitude) of each electrode pair was compared to the reference electrode pair with interelectrode distance of 40 mm placed on the center of the tested muscles. The average EMG difference ratios ranged 1.1-2.2%/mm in parallel shift, 1.0-1.9%/mm in distance shift, and 0.3-0.6%/degree in rotation shift. Displacement error of electrodes in separate recording instances should be reduced using anatomical landmarks, when SEMG is applied as a quantitative method to evaluate change in the states of involuntary movements.

The effects of interelectrode distance on electromyographic amplitude and mean power frequency during isokinetic and isometric muscle actions of the biceps brachii

The purpose of this study was to examine the effects of interelectrode distance (IED) on the absolute and normalized electromyographic (EMG) amplitude and mean power frequency (MPF) versus isokinetic and isometric torque relationships for the biceps brachii muscle. Ten adults [mean+/-SD age=22.0+/-3.4 years] performed submaximal to maximal, isokinetic and isometric muscle actions of the dominant forearm flexors. Following determination of isokinetic peak torque (PT) and the isometric maximum voluntary contraction (MVC), the subjects performed randomly ordered, submaximal step muscle actions in 10% increments from 10% to 90% PT and MVC. Surface EMG signals were recorded simultaneously from bipolar electrode arrangements placed over the biceps brachii muscle with IEDs of 20, 40, and 60mm. Absolute and normalized EMG amplitude (muVrms and %max) increased linearly with torque during the isokinetic and isometric muscle actions (r(2) range=0.988-0.998), but there were no significant changes for absolute or normalized EMG MPF (Hz or %max) from 10% to 100% PT and MVC. In some cases, there were significant (p<0.05) differences among the three IED arrangements for absolute EMG amplitude and MPF values, but not for the normalized values. These findings suggested that for the biceps brachii muscle, IEDs between 20 and 60mm resulted in similar patterns for the EMG amplitude or MPF versus dynamic and isometric torque relationships. Furthermore, unlike the absolute EMG amplitude and MPF values, the normalized EMG data were not influenced by changes in IED between 20 and 60mm. Thus, normalized EMG data can be compared among previous studies that have utilized different IED arrangements.

Intra-rater reliability of electromyographic recordings and subjective evaluation of neck muscle fatigue among helicopter pilots

The aim was to evaluate the reliability of a method of measuring neck muscle fatigue among helicopter pilots.

METHOD

Surface EMG from three areas in the neck region, bilaterally, was recorded among 10 male helicopter pilots while they were performing isometric contractions in flexion and extension for 45 s, sustaining a force representing 75% of maximum strength in a seated position. Perceived fatigue was rated using the Borg CR-10 scale. The test was repeated twice the first day and then two additional times with one-week intervals. Variables analyzed were the slope of the median frequency change, the normalized slope, and the ratings after 15, 30 and 45 s; and also the initial median frequency (IMDF). The intra-class correlation (ICC) and the measurement error (S(w)), intra- and inter-day were calculated statistically.

RESULTS

The best reliability for the slope was found for the 45 s intra-day analysis taking all measurements into account (ICC 0.65-0.83). The reliability after 30 s was poorer but still acceptable (ICC 0.52-0.71). For the subjective ratings, the highest reliability was found after 30 s inter-day (ICC 0.86-0.88). IMDF showed generally high reliability for the intra-day analyses (ICC 0.63-0.80).

CONCLUSION

The method is reliable for use in further research. Since performing a contraction of 75% of maximum was quite strenuous, we recommend that the protocol be shortened to 30 s.

Activation characteristics of shoulder muscles during maximal and submaximal efforts

Determination of individual maximum voluntary contraction (MVC) force is used as the gold standard for normalising surface EMG (SEMG) data. Assuming a linear amplitude-force relationship, individual strain levels are defined according to percentage rates of the measured MVC levels. The purpose of the study was to investigate if the assumed force-strain relationship can be applied without qualification. Therefore, healthy volunteers (nine men, ten women) were investigated during isometric exercises of shoulder muscles at MVC and 50% levels. Tasks were performed at four different angular positions in frontal, sagittal and horizontal planes. In each plane, both possible force directions were investigated. SEMG was taken simultaneously from 13 muscles of the shoulder and upper arms from both sides of the body. At a force level of 50% MVC, SEMG amplitude levels were compared to the expected 50% level. Differences in muscular co-ordination patterns were also determined. During retroversion and horizontal flexion, amplitude levels significantly remained at levels below 50%. This was seen for all the muscles investigated, independent of relative contribution to force production. During horizontal extension and abduction, the main force-producing muscles showed amplitude levels that significantly exceeded the expected 50% level. Co-ordination patterns differed significantly between MVC and submaximal conditions for anteversion, retroversion and horizontal extension. Specifically, four shoulder muscles showed higher proportions at the 50% level compared to MVC. Therefore, certain percentage rates of MVC force levels exhibit quite different strain rates, as identified by SEMG. Depending on force direction, differences in co-ordination patterns exist between MVC and submaximal test conditions. Both findings have implications for therapeutic and training applications.

Assessment of low back muscle fatigue by surface EMG signal analysis: methodological aspects

This paper focuses on methodological issues related to surface electromyographic (EMG) signal detection from the low back muscles. In particular, we analysed (1) the characteristics (in terms of propagating components) of the signals detected from these muscles; (2) the effect of electrode location on the variables extracted from surface EMG; (3) the effect of the inter-electrode distance (IED) on the same variables; (4) the possibility of assessing fatigue during high and very low force level contractions. To address these issues, we detected single differential surface EMG signals by arrays of eight electrodes from six locations on the two sides of the spine, at the levels of the first (L1), the second (L2), and the fifth (L5) lumbar vertebra. In total, 42 surface EMG channels were acquired at the same time during both high and low force, short and long duration contractions. The main results were: (1) signal quality is poor with predominance of non-travelling components; (2) as a consequence of point (1), in the majority of the cases it is not possible to reliably estimate muscle fiber conduction velocity; (3) despite the poor signal quality, it was possible to distinguish the fatigue properties of the investigated muscles and the fatigability at different contraction levels; (4) IED affects the sensitivity of surface EMG variables to electrode location and large IEDs are suggested when spectral and amplitude analysis is performed; (5) the sensitivity of surface EMG variables to changes in electrode location is on average larger than for other muscles with less complex architecture; (6) IED influences amplitude initial values and slopes, and spectral variable initial values; (7) normalized slopes for both amplitude and spectral variables are not affected by IED and, thus, are suggested for fatigue analysis at different postures or during movement, when IED may change in different conditions (in case of separated electrodes); (8) the surface EMG technique at the global level of amplitude and spectral analysis cannot be used to characterize fatigue properties of low back muscles during very low level, long duration contractions since in these cases the non-stable MU pool has a major influence on the EMG variables. These considerations clarify issues only partially investigated in past studies. The limitations indicated above are important and should be carefully discussed when presenting surface EMG results as a means for low back muscle assessment in clinical practice.

Lumbar muscle fatigue and recovery in patients with long-term low-back trouble--electromyography and health-related factors

OBJECTIVE

The aim was to explore the validity and reliability of EMG for assessing lumbar muscle fatigue.

DESIGN

Patients with long-term low-back trouble (n=57) were compared to a healthy reference group (n=55). Back muscle fatigue and recovery were studied in relation to health-related factors.

BACKGROUND

EMG spectral variables are important tools in the assessment of patients with low-back trouble. The influence of disability on these variables needs further investigation.

METHODS

EMG from the lower back muscles was recorded during a 45 s trunk extension at 80% of maximal voluntary contraction torque and during recovery. Disability was studied using questionnaires.

RESULTS

The reliability was high for maximal voluntary contraction torque and EMG initial median frequency, lower for the median frequency slope, and insufficient for median frequency recovery half-time. The patients had lower maximal voluntary contraction torque, higher initial median frequency at L5 level, flatter slope, and longer recovery half-time than the healthy subjects did. However, for subjects with significantly negative slope, indicating fatigue, there was no significant difference in slope between patients and healthy subjects, while, for subjects without such fatigue, patients showed significantly flatter slopes at L5. The sensitivity/specificity of the test was 86%/78%. The most significant variables selected with logistic regression were maximal voluntary contraction torque and initial median frequency at L5. Patients without significantly negative slopes during contraction and/or not exponential-like EMG recovery scored worse on several items concerning disability and self-efficacy.

CONCLUSIONS

EMG spectral variables in combination with torque might be used for classification. For patients with long-term low-back trouble, the ability to fatigue the lumbar muscles sufficiently to obtain a significantly negative slope during an 80% maximal voluntary contraction may be a sign of better functioning.

RELEVANCE

The ability to fatigue the back muscles during a test requiring a high force output might be achieved with back muscle training focused on increasing strength and self-efficacy.