Content uploaded by Nihat Özgören
Author content
All content in this area was uploaded by Nihat Özgören on Nov 01, 2017
Content may be subject to copyright.
COMPARISON OF THREE EMG-BASED MUSCLE FATIGUE ASSESSMENTS IN
DYNAMIC CONTRACTIONS
Özgören1 1
1Biomechanics Research Group, Department of Exercise and Sport Sciences,
Faculty of Sport Sciences, Hacettepe University, Ankara, Turkey
The aim of this study was to (1) focus on the number of peaks methods (NoP) as an
alternative method for electromyography (EMG) analysis and to (2) compare the
performance of this method with two traditional EMG parameters, mean frequency (MNF)
and median frequency (MDF) in assessment of muscle fatigue during cyclic dynamic
contractions. A total of 6 participants performed 50 repeated, maximal concentric isokinetic
muscle actions of the left leg extensors at a velocity of 180 deg/s using an isokinetic
dynamometer. Quantification of the EMG activity of the muscles at each repetition was made
by using three methods. NoP values were found to be positively correlated with the peak
torque as well as MNF, and MDF. The study showed that the NoP method which has a
simple implementation, can be used for assessing the level of muscle fatigue.
KEYWORDS: Electromyography, isokinetic, dynamic contractions, muscle force, fatigue
INTRODUCTION: Dynamic strength, endurance and fatigue of human skeletal muscle is often
measured by using isokinetic dynamometers. In isokinetic testing, peak torque (PT) is one of the
prominent biomechanical variables which is used to evaluate the muscle performance. Besides,
the muscle activity which is observed as electromyography (EMG) generally utilized to quantify
the developed force by a muscle. The most popular frequency-domain features of surface EMG
are median frequency (MDF) and mean frequency (MNF) of the power spectral density which
are frequently used for assessment of muscle fatigue. In the previous study (Özgören and
, number of peaks (NoP) method, was used for the quantification of EMG activity
during cyclic dynamic contractions where the level of muscle fatigue was assessed based on
NoP results. Similar methods were previously used (Dayan, Spulber, Eftekhar, Georgiou,
Bergmann and McGregor, 2012; Gabriel, 2000; Calder, Gabriel and McLean, 2009; Gabriel,
Christie, Inglis and Kamen, 2011; Gabriel, Lester, Lenhardt and Cambridge, 2007) in a number
of studies where researchers focused on the change in various EMG signal spike parameters.
Reliability of the NoP method was considered to be discussed by taking the results of similar
frequency based analysis methods into account. From this point of view, the aim of this study
was to compare the performance of NoP method with MNF and MDF in assessment of muscle
fatigue from EMG signals during cyclic dynamic contractions for three leg extensor muscles,
rectus femoris (RF), vastus medialis (VM) and vastus lateralis (VL).
METHODS: Five male and one female healthy volunteers (age: 25 ± 2.5 years, height: 1.79 ±
0.01 m, bodyweight: 76.8 ± 10.7 kg) participated in this study. Each participant performed 50
repeated maximal concentric isokinetic concentric knee extension of the left leg from 90° of
flexion to 0° at a velocity of 180 deg/s using an isokinetic dynamometer (Cybex-Humac Norm,
U.S.). EMG signals were collected from the left RF, VM, and VL muscles throughout testing with
surface electrodes of a Bagnoli 8-channel desktop system (Delsys Inc., U.S.). The EMG signals
were sampled with a frequency of 1 kHz and amplified with a gain of 1000. A reference
electrode was placed over the right iliac crest.
EMG signal and torque data processing was then performed using custom written codes in
MATLAB (MathWorks Inc., U.S.). EMG signals were digitally zero-phase filtered by a 3rd order
Butterworth band-pass filter (20 to 250 Hz). The duration of each concentric knee extension was
500 ms since the velocity of the dynamometer was 180 deg/s. Thus the EMG signal groups of
50 concentric isokinetic muscle action were automatically detected after the first signal of the
first group was selected. The power-density spectrum was obtained using the fast Fourier
transform (FFT) technique. Then MNF and MDF of the power spectrum were computed from the
EMG signal group for each concentric phase of the contraction cycle. The peaks in 50 signal
groups were detected and the mean value of these peaks was calculated after full wave
rectification of the EMG data. For quantification of the EMG activity, the number of peak values
greater than the mean value was calculated for each signal group 2016).
MNF, MDF and NoP data for all muscles, and PT data of each subject were modelled using a
single-term exponential fit ( ) with %95 confidence bounds and parameter b was used
as fatigue index. Total of 60 fatigue indices (3 indices from three methods for a single muscle of
one participant, 1 index from torque data for each participant) were obtained. The fatigue
indices were statistically tested using Friedman’s nonparametric test in order to compare the
effect of used analysis method on the fatigue indices of the muscles. Wilcoxon signed-rank test
was used for comparing the fatigue index samples in case of any possible difference. The
relationship between PT and MDF, MNF and NoP of the muscles were analysed using linear
regression analysis.
RESULTS: Mean MNF, MDF, and NoP for RF, VM, and VL exhibited a decreasing trend along
with the PT as it is shown in Figure 1. There were positive significant correlations between PT
and MNF, MDF, and NoP for the three muscles of all participants. The correlations among MNF,
MDF and NoP were also positive (Table 1).
Table 1.
Correlation coefficients (R) between mean EMG variables obtained from three methods and the
peak torque.
Muscle
Variable
s
RF
VM
VL
MNF – PT 0.9699
†
0.8166
†
0.9657
†
MDF – PT 0.9721
†
0.7915
†
0.9595
†
NoP – PT 0.9217
†
0.9672
†
0.9372
†
MNF – MDF 0.9952
†
0.9688
†
0.9795
†
MNF – NoP 0.8976
†
0.8311
†
0.8795
†
MDF – NoP 0.9104
†
0.7902
†
0.8754
†
Note that †indicates significant correlations.
The fatigue indices for PT, MNF, MDF, and NoP of all muscles were negative for all participants
as it is clear from Figure 1b, 1d and 1f which indicated a decrease in all variables throughout
testing. According to the Friedman’s test results, there was no significant difference between the
fatigue indices of RF, VM, and VL when a particular method was concerned (p-values for
methods; MNF: 0.31, MDF: 0.60, NoP: 0.11). Obtained p-values for fatigue indices of each
muscle from MNF, MDF, and NoP were 0.001, 0.002, and 0.007 respectively which
demonstrated that the effect of at least one method on the fatigue indices was significant.
Results of Wilcoxon test showed that this difference derived from only NoP of VM muscle.
Figure 1. Change in PT and (a) MNF, (c) MDF, (e) NoP for three muscles through the test.
DISCUSSION: The data provided in Table 1 showed that NoP was positively correlated with
MNF and MDF for each muscle. A significant correlation between MDF and peak counting in
EMG signals for isometric contractions of RF and VL was previously shown by Dayan et al.
(2012). This study demonstrated that a positive correlation exists between MDF, MNF and NoP
during cyclic dynamic contractions of RF, VM, and VL. A nonlinear relationship was described
between EMG and torque production using exponential functions in this study while linear
(Shinohara, Kouzaki, Yoshihisa and Fukunaga, 1998) and nonlinear (Watanabe and Akima,
2006) relationships have been discussed previously. All fatigue indices were found to be
negative which indicated a decline in MDF, MNF, and NoP along with the PT as it can be clearly
seen in Figure 1b, 1d and 1f. When the difference in fatigue indices of VM originating from NoP
method, and the stronger correlation between NoP of VM and PT (Table 1) is considered, it is
suggested that NoP may be a method which is more sensitive to frequency content of the EMG
signals in dynamic fatiguing exercises since VM could have a variation in fibre composition
different than RF and VL. Moreover, NoP is a threshold dependent method such that the
threshold is set based on the mean peak value of the complete EMG data set. So that if the VM
muscle has a greater type I fibre composition, a lower peak frequency and a peak amplitude
less than the threshold value is likely to be acquired in EMG signals of the VM muscle.
CONCLUSION: This study showed that NoP method can be utilized to analyse the EMG activity
of RF, VM and VL muscles during dynamic concentric contractions. Further, the NoP method
which has a simple implementation, can be used for assessing the level of muscle fatigue since
it agrees with the traditional features, MNF, and MDF. Consequently, NoP method would rather
be used for interpreting the level of muscle fatigue since MNF and MDF of the power spectral
density require intensive computations.
REFERENCES
Calder, K. M., Gabriel, D. A. & McLean, L. (2009). Differences in EMG spike shape between individuals
with and without non-specific arm pain. Journal of Neuroscience Methods. 178(1):148-156. doi:
10.1016/j.jneumeth.2008.11.015
Dayan, O., Spulber, I., Eftekhar, A., Georgiou, P., Bergmann, J. & McGregor, A. (2012). Applying EMG
spike and peak counting for a real-time muscle fatigue monitoring system. In proceedings: IEEE
Biomedical Circuits and Systems Conference (BioCAS), pp:41-44, 28-30 November, Hsinchu, Taiwan.
doi: 10.1109/BioCAS.2012.6418474
Gabriel, D. A. (2000). Reliability of SEMG spike parameters during concentric contractions.
Electroencephalography and Clinical Neurophysiology. 40(7): 423–430.
Gabriel, D. A., Lester, S. M., Lenhardt, S. A. & Cambridge, E. D. J. (2007). Analysis of surface EMG spike
shape across different levels of isometric force. Journal of Neuroscience Methods. 159(1):146-152.
doi:10.1016/j.jneumeth.2006.07.004
Gabriel, D. A., Christie, A., Inglis, J. G. & Kamen, G. (2011). Experimental and modelling investigation of
surface EMG spike analysis. Medical Engineering & Physics. 33(4):427-437. doi:
doi:10.1016/j.medengphy.2010.11.010
-repetition isokinetic fatigue test. In
proceedings: 14th International Sport Sciences Congress, pp: 338, (ISBN: 978-605-83126-1-6) 1-4
November, Antalya, Turkey.
Shinohara M, Kouzaki M, Yoshihisa T, and Fukunaga T. (1998). Mechanomyogram from the different
heads of the quadriceps muscle during incremental knee extension. European Journal of Applied
Physiology, 78:289– 295.
Watanabe K. and Akima H. (2006). Normalized EMG to normalized torque relationship of vastus
intermedius muscle during isometric knee extension. European Journal of Applied Physiology, 106:665-
673.
