This article was originally published in a journal published by
Elsevier, and the attached copy is provided by Elsevier for the
author’s benefit and for the benefit of the author’s institution, for
non-commercial research and educational use including without
limitation use in instruction at your institution, sending it to specific
colleagues that you know, and providing a copy to your institution’s
administrator.
All other uses, reproduction and distribution, including without
limitation commercial reprints, selling or licensing copies or access,
or posting on open internet sites, your personal or institution’s
website or repository, are prohibited. For exceptions, permission
may be sought for such use through Elsevier’s permissions site at:
http://www.elsevier.com/locate/permissionusematerial
Author's Personal Copy
Clinical Biomechanics 25 (2010) 103–109
Contents lists available at ScienceDirect
Clinical Biomechanics
journal homepage: www.elsevier.com/locate/clinbiomech
Torso muscle EMG profile differences between patients of back pain and control
Shrawan Kumar a,*, Narsimha Prasad b
a
b
Physical Medicine Institute, University of North Texas Health Science Center, Fort Worth, TX 76107, United States
Department of Mathematical and Statistical Sciences, Faculty of Sciences, University of Alberta, Edmonton, Alberta, Canada
a r t i c l e
i n f o
Article history:
Received 30 July 2009
Accepted 27 October 2009
Keywords:
Chronic pain
EMG
Low back pain
Classification
Identification
Diagnosis
a b s t r a c t
Background: Electrophysiological criteria that identify and characterize low back pain can lead to better
understanding of the afliction and possibly aid in its treatment.
Method: Nineteen male and 22 female subjects with chronic back pain, without lumbar radiculopathy;
and 30 male and 33 female control subjects with no history of low back pain in the last 12 months, were
recruited into the study. All subjects flexed, extended, laterally flexed, flexed anterolaterally and
extended posterolaterally isometrically to 20% and 100% of their maximal voluntary contraction
(MVC). Additionally, patients were asked to do these activities to their pain threshold levels and control
subjects to 60% maximum voluntary contraction. Surface electromyograms (EMG) were recorded from
lumbar erectores spinae, external obliques and rectus abdominis bilaterally. The electromyogram was
subjected to magnitude, Fast Fourier Transform, and wavelet analyses. The median frequency and frequency bands were calculated with their power. The wavelet decomposition was done and a logistic discriminate analysis was carried out to classify patients and normal controls.
Findings: The normalized peak electromyograms of patients were significantly greater than controls
(P < 0.01). The muscle conduction velocity was not disturbed by pain. Significant differences were found
in total power between patients and controls (P < 0.01). The analysis correctly classified patients and controls 65% and 98% of the time, respectively at 20% MVC, 95.1% (patients) and 86.8% (controls) at pain
threshold/60% MVC, and 74.3% (patients) and 86.4% (controls) at pain tolerance/MVC (P < 0.05).
Interpretation: The surface electromyography can be used in discriminating chronic low back pain
patients and controls. This would be an objective test over and above other subjective tests, such as pain
provocation.
Ó 2009 Elsevier Ltd. All rights reserved.
1. Introduction
Since the underlying etiology of the idiopathic low back pain
and ensuing physiological changes remain elusive, the interventions remain nonspecific with variable results. The predominant
mechanism of musculoskeletal injury causing pain would be due
to injury to these structural components of the low back (Kumar,
2001). An electrophysiological technique is likely to be primarily
based on objective rather than subjective factors. The chronic
low back pain patients showed significantly increased EMG activity
in the swing phase of the gait, a phase where lumbar muscles are
normally silent (Arendt-Nielsen et al., 1996). This change correlated significantly with the intensity of the back pain. The authors
concluded that the musculoskeletal pain modulates motor performance during gait, probably via reflex pathways. Experimentally
induced pain as well as a fear of pain has subtle effects on the erector spinae EMG activity during walking (Lamoth et al., 2004). These
* Corresponding author.
E-mail address: [email protected] (S. Kumar).
0268-0033/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.clinbiomech.2009.10.013
authors concluded that the altered gait observed in the low back
pain patients was probably a complex, evolved consequence of
lasting pain rather than an immediate effect. Postural aberration
with the chronic low back pain has also been reported (Christie
et al., 1995), likely due to differential muscle activation.
In a sample of 30 chronic low back pain patients and 30 control
subjects muscle activity mean values were threefold higher in
chronic low back pain patients than controls in a static exertions,
and two times greater in dynamic exertions (P < 0.01) (Ambroz
et al., 2000). Finneran et al. (2003) have reported that a large-array
surface electromyography in low back pain yielded a multi-focal
and asymmetrical pattern of EMG activity. Based on this observation the authors concluded that such a method will be useful in
evaluating patients with low back pain. Despite the foregoing studies, the changes in EMG due to pain have not been characterized
adequately. Understanding the EMG behavior has been tackled
by creating experimental pain through injection of noxious substances, and studying the aberrations in EMG characteristics. Some
studies have concluded that the experimental pain does not alter
the spectral parameters, motor unit properties or the conduction
Author's Personal Copy
104
S. Kumar, N. Prasad / Clinical Biomechanics 25 (2010) 103–109
velocity significantly (Birch et al., 2000; Farina et al., 2005). Others
indicate an alteration in motor unit recruitment due to pain, especially in the time and frequency domain parameters (Lundblad
et al., 1998; Hodges and Richardson, 1999). A change in timedomain parameters enabling 97% accurate classification of chronic
cervical pain patients has been reported by Kumar et al. (2007).
Several studies have indicated a discernible and significant increase in the EMG amplitude due to experimental or pathological
pain (Sohn et al., 2004; Geisser et al., 2004; Ervilha et al., 2004,
Graven-Nielsen et al., 1997).
This project was designed to investigate time and frequencydomain characteristics of torso muscle EMG from patients of
established muscle pathophysiology in low back pain cases, and
compare them with normal controls. Such a tool may allow physical diagnosis and perhaps to discern the severity of the problem.
Furthermore, it may also serve as a tool to measure efficacy of various pain treatment modalities.
2. Methods
Fig. 1. The experimental set-up.
2.1. Sample
vertical telescopic 15 cm wide rectangular metal tube welded to
a thick iron plate rigidly bolted in the floor (Fig. 1). The 12 cm wide
inner concentric tube could be raised or lowered and securely
locked into place. Experimental subjects were fitted with a nonstretchable Velcro belt, mounted perpendicular to the tubing by
means of an aircraft cable with an intervening load cell (Interface
Technology Inc., Walnut, CA; Model I-250 with force range of
1000 N). Thus, any force exerted on the Velcro belt was registered
on the load cell, which was fixed in its mechanical path and remained in line with the cable at all times during all tests.
Nineteen male (mean age 56 years) and 22 female patients
(means age 50 years) of chronic low back pain were recruited from
a neurology EMG clinic. The exclusion criterion was lumbar radiculopathy which was ruled out based on EMG and neuroimaging
studies. The mean height and weight of male patients was
176 cm (8 cm) and 91 kg (16 kg). The females, on the other hand,
had a mean height of 163 cm (8 cm) and mean weight of 174 kg
(11 kg). There were 30 male controls (mean age 30 years) with
height and weight of 175 cm (9 cm) and 74 kg (11 kg.). The control
subjects, who did not have low back pain, were recruited from general population. In addition, there were 33 female control subjects
(mean age 33 years) with a mean height of 162 cm (7 cm) and
mean weight of 59 kg (10 kg).
2.2. Subject preparation
After obtaining informed consent and suitable skin preparation
(which included shaving, where needed, and rubbing skin with
alcohol–acctone mixture to improve conductivity) Delsys bipolar
active knife edge surface electrodes with a fixed inter-electrode
distance (10 mm) were applied to the lumbar erectores spinae, at
the level of the third lumbar spinae process, external oblique
(10 cm below the costal margin and 14 cm lateral to umbilicus)
and rectus abdominis (6 cm above the umbilicus) bilaterally in
both patients and control subjects. These electrodes had an on-site
pre-amplification of 10 times. The signals were fed to an amplifier
for an additional gain of 100 times. Such prepared subjects were
seated in the experimental set up for testing.
2.3. Testing device
The testing device consisted of an adjustable chair, sliding platform, and floor mounted strength measuring device (Fig. 1). The
chair consisted of a molded plastic seat mounted on an iron platform with telescopic metal legs fixed to the base plate. The back
rest and seat were fitted with a Velcro restraint system to stabilize
lower extremities. The chair with its base plate was pivoted in the
center and had casters on the periphery for circular motion with
stability. The circular plate was graduated in intervals of 5° and
holed through which bolts were placed to match the holes in the
sliding board of the platform. Two bolts were placed at opposite
ends and tightened for a rigid fixation of the chair in the desired
position for the subject. The force resisting device consisted of a
2.4. Tasks
After signing the informed consent form, subjects were measured and weighed and their ages recorded. Finally, they were
seated and stabilized in an erect and upright posture. Previously
prepared subjects were seated in the experimental set up for torso
strength testing. The subjects were asked to exert in isometric flexion, extension, left lateral flexion, left antero-lateral flexion, and
left postero-lateral extension. The sequence of these exertions
was fully randomized. Prior to the start of the trial, the subjects
were instructed to bring their contraction to the subjectively determined target level in the first two seconds and hold it there for another three seconds. At this time the trial was terminated. The
patients were asked to exert to their 20% maximum voluntary contraction (MVC), pain threshold, and pain tolerance levels. The patients were instructed to stop exertion as soon as they reached
their self assessed subjective 20% maximum voluntary contraction
(MVC) and as soon as the pain was evoked (for pain threshold exertion but to continue to exert as hard as they could for pain tolerance level. This level of contraction among low back pain
patients was also assigned as their MVC. The control subjects were
asked to exert to their 20% MVC, 60% MVC and maximum voluntary contraction. The exertion levels of 20% and 60% of MVC were
subjectively assessed by the subjects. All exertions were required
to be five seconds long. While the subjects exerted force, the
EMG of all six muscles and force of exertion were sampled at 2 kHz.
3. Data analysis
3.1. EMG
The raw EMG signals were band pass filtered with low cutoff
frequency of 20 Hz and high cutoff frequency of 450 Hz. The signals
in this frequency band were pre-amplified and EMG activities were
Author's Personal Copy
S. Kumar, N. Prasad / Clinical Biomechanics 25 (2010) 103–109
marked and selected for Fast Fourier Transform analysis. These
activities were isometric and data segments were stationary, as
such chosen for such analysis. However, a test of the stationarity
of the data was carried out by calculating autocorrelation before
proceeding with the rest of the analysis. The spectral data analysis
of each muscle, in each of the 15 activities for the trunk muscles,
was done separately for patients of back pain and control subjects.
From the spectral analysis, the lower and upper 3 dB frequencies
and the bandwidth was extracted for patients and control subjects.
Each of these parameters was calculated for 20% MVC (patients and
controls); pain threshold (for patients), 60% MVC for controls; and,
pain tolerance (for patients) and MVC (for controls). The median
frequency of each of the torso muscle, in each of the activities of
the patients and controls were extracted. Similarly, mean power
frequency, total power and peak power were calculated for the
above mentioned muscles from the spectral analysis.
Descriptive statistics of the variables of EMG magnitude, normalized peak EMG, median frequency, mean power frequency, total power, peak power, frequency at peak power, time to onset and
time to peak EMG were calculated. Each of these variables was also
subjected to a one-way analysis of variance. This ANOVA was carried out to determine any significant difference between patients
and controls.
3.2. Time and frequency domain analyses
The EMG signals were first rectified and subjected to wavelet
analysis to obtain eight levels of decomposition of signals for each
subject under each condition and for each muscle. Daubechies
wavelet transformation was used for this purpose. The eight levels
or scales, of the raw signals decomposition, were confined to following frequency bands:
1.
2.
3.
4.
5.
6.
7.
8.
0–7.8125 Hz.
7.8125–15.625 Hz.
15.625–31.25 Hz.
31.25–62.5 Hz.
62.5–125 Hz.
125–250 Hz.
250–500 Hz.
500–1000 Hz.
At each of the above eight scales the following were computed
as frequency features.
Eight RMS values (RMS1, RMS2, . . ., RMS8).
Mean (m), dispersion (d), skewness (s) and kurtosis (k).
105
to ‘Low Back Pain Group’ if predicted probability from the model
was greater than or equal to 0.5, otherwise the subject was classified to ‘Control Group.’ Based on onefold (delete one) and crossvalidation approaches miss-classification errors were computed
to evaluate the performance of the proposed classification scheme.
With the selected features from the logistic regression analysis
we also performed linear discriminate classification to examine the
discrimination power. Two-sigma bar plots on linear discriminate
scores were plotted separately for ‘Control’ and ‘Low Back Pain’
groups to demonstrate the appropriateness of the proposed feature
selections in classification of subjects into two groups.
4. Results
4.1. Strength
The body weight normalized means and standard deviations of
all strength of patient and control samples for 20% MVC, 60% MVC
(control)/pain threshold (patients), and MVC (control) and pain tolerance (patients) for flexion, left antero-lateral flexion, left lateral
flexion, left postero-lateral extension and extension are presented
in Table 1. Low back pain patient’s pain threshold strength values
were between 18% and 23% of their body weight for males and
between 14% and 21% for females. Pain tolerance values of the patients were similar to those of the maximum voluntary contraction
efforts of the control subjects except in flexion where patients
were 5% lower than control subjects. The ANOVA revealed that
there was no significant difference between the body weight normalized strength between the groups. However, there were significant difference between patients and controls for flexion in 60%/
pain threshold and MVC contraction efforts (P < 0.001).
4.2. EMG
4.2.1. EMG magnitude and pattern
The normalized peak EMG scores in activities of flexion, left
antero-lateral flexion, lateral flexion, left postero-lateral extension,
and extension activities of patients were higher than those of controls for both genders at all levels of contractions. The data for patient pain threshold and 60% contraction of normal controls is
presented in Table 2. Also, these peak EMG scores progressively increased with the levels of contractions for all subjects. However,
the general magnitude relationship as indicated above was maintained at every level of contraction. There was no obvious difference between the groups in pattern of the recorded EMG.
4.3. Spectral parameters
For time-domain features, rectified EMG signals were subjected
to fourth-order autoregressive analysis using the ‘arfit.m’ in Matlab
using the fourth-order autoregressive model.
From this fourth-order autoregressive model, autocorrelations
(rxx1, rxx2, rxx3, rxx4) and autoregression coefficients (phi1,
phi2, phi3, phi4) up to order four were extracted to describe
time-domain features, namely dependency of the signals. All these
signals were also subjected to Wavelet Analysis and Time Series
Analysis to obtain time and frequency domain features on all subjects (patients and control).
Logistic regression analysis was applied to identify best frequency-domain and time-domain features in classifying subjects
to ‘control’ and ‘Low Back Pain Group’ by testing group as a binary
response variable and all above mentioned frequency-domain and
time-domain measures as predictors. Separate logistic models
were fitted for three exertion levels. After identifying the ‘best
model’ with minimum number of predictors, classification procedure was formed based on the ‘best model’ by classifying a subject
The median frequencies of various muscles were found to be
different. For the left external oblique, the median frequency ranged between 81 Hz and 157 Hz for both male and female and patients and controls. Similarly, the ranges for the right external
oblique, left erectors spinae, right erectors spinae, left and right
rectus abdominis were between 72 and 87, 82 and 130, 92 and
156, 95 and 159 and 100 and 172, respectively. The median frequency of individual muscles also varied slightly between the five
different activities of flexion, left antero-lateral flexion, left lateral
flexion, left postero-lateral extension and extension. The ANOVA
revealed sporadic differences between patients and controls in
the median frequency of the torso muscles and also showed
sporadic significant differences between patients and controls in
different activities. Generally, the external obliques and right erectors spinae had the highest amount of total power in their signals,
but did not discriminate between patients and controls consistently. When the total EMG bandwidth was divided into 10%
Author's Personal Copy
106
S. Kumar, N. Prasad / Clinical Biomechanics 25 (2010) 103–109
Table 1
Body weight normalized peak strength of subjects in different experimental conditions.
Subject type
Male
Female
Exertion levels
20% MVC
Threshold/60% MVC
Tolerance/MVC
Norm peak force, w/w (%)
Norm peak force, w/w (%)
Norm peak, w/w (%)
Mean
Mean
Std. deviation
Std. deviation
Mean
Std. deviation
Patient
Flexion
Left antero
Left lateral
Left postero
Extension
4.56
4.71
5.05
6.05
8.64
1.19
1.25
1.66
3.60
4.18
17.95
17.02
17.43
17.23
22.91
9.93
5.80
5.59
6.37
10.59
19.97
20.17
21.20
21.77
29.08
6.47
5.49
6.33
6.76
11.96
Control
Flexion
Left antero
Left lateral
Left postero
Extension
5.91
5.19
5.67
5.74
6.81
2.99
2.04
2.86
2.28
3.94
10.68
11.44
12.77
12.88
14.27
2.47
3.24
2.98
4.43
4.21
24.84
22.30
23.59
25.54
30.45
12.81
8.74
9.21
11.43
20.24
Patient
Flexion
Left antero
Left lateral
Left postero
Extension
9.28
6.31
5.38
9.84
9.30
5.42
1.67
1.23
10.85
8.84
14.23
16.18
15.64
16.67
21.14
6.69
7.15
4.88
6.66
9.41
20.01
21.69
19.39
22.77
25.32
7.81
9.10
5.46
9.43
9.89
Control
Flexion
Left antero
Left lateral
Left postero
Extension
7.44
5.67
5.21
5.47
6.60
4.96
2.08
2.10
3.21
3.34
11.79
12.34
12.30
12.52
13.35
6.32
5.03
3.97
4.40
4.98
26.32
25.11
22.32
22.60
28.23
11.57
10.22
9.22
12.88
14.50
Table 2
Normalized peak EMG score of torso muscles of patients and controls in pain threshold/60% maximum voluntary contraction (% of MVC).
Subject type
Male
Female
Flexion
Left antero
Left lateral
Left postero
Extension
Normalized peak (%)
Normalized peak (%)
Normalized peak (%)
Normalized peak (%)
Normalized peak (%)
Mean
Std. deviation
Mean
Std. deviation
Mean
Std. deviation
Mean
Std. deviation
Mean
Std. deviation
Patient
LEO
LES
LRA
REO
RES
RRA
244.81
33.06
94.32
160.19
41.14
115.54
205.11
15.49
26.22
133.98
32.28
64.30
150.60
43.90
79.96
245.00
23.64
50.07
87.52
28.80
40.96
224.01
12.40
30.54
97.19
67.99
73.24
184.43
30.30
77.21
37.01
41.45
38.10
213.10
23.79
64.52
44.19
77.51
94.42
139.99
38.75
65.58
14.77
30.69
93.01
126.86
26.98
61.32
62.03
90.87
59.72
74.18
80.47
54.93
37.48
31.96
34.27
78.53
23.92
46.35
Control
LEO
LES
LRA
REO
RES
RRA
134.76
17.18
83.89
148.72
51.41
61.69
51.07
8.31
19.11
47.53
47.63
11.75
87.12
29.14
95.91
106.23
17.15
76.41
11.41
10.94
70.53
84.74
7.47
46.60
80.89
15.82
74.60
43.95
30.46
47.14
1.94
5.45
42.67
31.45
19.42
37.53
32.36
38.92
35.08
41.76
42.02
24.22
6.20
15.54
34.95
43.52
14.89
25.18
24.69
70.95
24.11
34.38
64.86
20.84
10.60
15.17
26.80
37.41
12.43
22.33
Patient
LEO
LES
LRA
REO
RES
RRA
236.91
51.26
84.45
212.24
73.90
68.85
210.84
37.27
44.04
172.55
59.03
27.52
165.35
53.00
84.72
198.61
52.04
55.65
86.88
25.22
31.26
134.26
24.69
29.94
94.31
76.60
58.22
137.02
60.55
79.34
22.25
50.22
35.89
110.59
40.30
123.27
69.12
103.62
67.93
61.44
71.36
44.15
53.44
44.73
59.69
27.91
30.03
26.36
157.56
96.29
50.54
72.75
95.96
42.27
256.36
23.94
42.38
77.49
20.60
29.05
Control
LEO
LES
LRA
REO
RES
RRA
41.69
24.16
78.81
35.05
13.04
37.12
9.23
28.31
7.25
5.11
6.40
9.54
23.71
42.33
49.66
28.00
45.31
51.24
9.23
32.39
36.65
9.57
59.07
16.23
63.23
86.67
74.70
28.33
26.35
34.95
27.77
39.11
40.31
15.45
29.00
39.07
32.06
39.11
33.49
21.61
15.14
30.46
15.89
15.60
25.82
13.83
5.52
24.02
29.66
99.17
47.94
24.67
65.44
29.92
16.60
1.18
36.45
10.84
7.82
19.66
intervals and the power contained in those bandwidths were compared between patients and controls, activity specific differences
emerged. In flexion, the right rectus abdominis in pain tolerance/
MVC contraction showed significant differences in each of the 10
bands (P < 0.01) and the right erectors spinae showed difference
significant in 6 out of 10 bands. Others were significant in fewer
bands (P < 0.05). In left postero-lateral extension for the pain tolerance/MVC contractions, the left erectors spinae were significant in
all bands (P < 0.05). In extension, the left and right erectors spinae
were found to be a discriminating factor only in pain tolerance/
MVC contraction (P < 0.01) in all 10 bands.
5. Classification results
At 20% MVC exertion level the logistic discrimination resulted
in classification of low back patients correctly 64.9% of the times
by both re-substitution and cross-validated methods (Table 3).
However, the control subjects were correctly classified 97.7% of
Author's Personal Copy
107
S. Kumar, N. Prasad / Clinical Biomechanics 25 (2010) 103–109
Table 3
Classification results for different exertion levels between patients and normal controls.
Method
Effort
Subject type
Re-substitution
20% MVC
Cross-validated
20% MVC
Pain threshold and 60% MVC
Subject type
Re-substitution
PT/60% MVC
Cross-validated
PT/60% MVC
Pain tolerance and MVC
Subject Type
Re-substitution
100% MVC
Cross-validated
100% MVC
Patient
Control
Patient
Control
Predicted group membership (per count of total)
Patient
Patient
Control
Patient
Control
Predicted group membership (per count of total)
Patient
Patient
Control
Patient
Control
Discriminate Scores = 1.021984*clesphi1+0.
354863*elraphi4
Exertion Levels: Threshold/60% MVC for Back Pain Subj. Group
1.25
1.00
0.75
0.50
0.25
Patient
Control
Discriminate Score = 1.029184*1st order auto-regression coefficient for the Muscle
Lumbar Erectores Spinae in Lateral Flexion + 0.354863*4th order auto-regression
coefficient for the muscle Left Rectus Abdominis in the Extension direction.
Fig. 2. 95% Confidence Intervals for mean discriminate scores at 60% exertion level
for back pain patients and control subject groups.
the times, again by both methods. At pain, threshold/60% MVC levels of contraction the classification of patients increased to 95.1% of
the times but that of controls declined to 86.8% correct classification (Table 3). At the pain tolerance/MVC levels of contraction
the patients were correctly classified 74.3% and controls 86.4% of
the time using the reconstitution method. With cross-validated
method the correct classification of patients declined further to
71.4% but that of control subjects remained unchanged (Table 3).
Fig. 2 shows 95% confidence intervals for mean discriminate
scores at pain threshold/60% MVC. Depending on the variable chosen, the values for patients and control subjects could be higher or
lower but difference between them was clear.
6. Discussion
The torso is a mechanical structure which has both, strength
and flexibility. It has been suggested that the biomechanics of
the spinal movement is affected by abnormal patterns of muscle
activity which could result in mechanically induced pain. Abnormal patterns of muscle activity during spinal movement may also
Predicted group membership (per count of total)
Total patient
Patient
Control
64.9
2.3
64.9
2.3
Total patient
Control
95.1
13.2
95.1
13.2
Total patient
Control
74.3
13.6
71.4
13.6
35.1
97.7
35.1
97.7
100
100
100
100
4.9
86.8
4.9
86.8
100
100
100
100
25.7
86.4
28.6
86.4
100
100
100
100
predispose the spine to an unstable state therefore result in abnormal loading, causing neuromuscular dysfunction resulting in pain
(Panjabi, 1992; Cholewicki and McGill, 1996). Asymmetry of the
muscle activity can occur and thereby decrease or increase inactivity compared to normal symmetrical response. The decreased
activity is attributed to the reflex inhibition and the increased
activity has been attributed to muscle spasm which would prevent
painful movement (Price et al., 1948).
The EMG of the six torso muscles for flexion and extension demonstrated a significantly different pattern (Fig. 3) and magnitude of
activity (Table 2) between patients and controls. In this patient
sample, given the lower strength, high EMG amplitude indicates
to the fact that there is hypersensitivity. This has been assigned
as muscle spasm which would prevent painful movements (Price
et al., 1948). If this pattern was present in only pain threshold contractions, the magnitude of force could explain some of this difference. However, given the fact that in the MVC/pain tolerance
contractions the pattern of EMG remained unchanged indicated a
clear presence of increased EMG activity.
Significant differences in normalized peak EMG were reported
both in pain threshold/60% MVC and pain tolerance/MVC contractions. However, there were more muscles which were showing significant differences in the pain threshold/60% MVC contraction. In
these, the left erector spinae and external oblique’s in both males
and females were significantly different in patients as compared
to controls for all five contractions (P < 0.01) (flexion, left anterolateral, left lateral flexion, left postero-lateral extension and extension). In addition, all muscle’s normalized peak EMG were
significantly different between patients and controls for flexion
and antero-lateral flexion (P < 0.03). The left rectus abdominis and
right external oblique’s were also significantly different for males
for all activities (P < 0.02) and many in females (P < 0.001). It would
clearly appear that the role of the erectors spinae and external oblique’s were the major extensor and flexor muscles and they had to
balance the mechanics of the torso as agonists, synergists or antagonists. In extensor activities, as one would expect, that the erectors
spinae muscles were significantly different in patients as compared
to normals. It would, therefore, appear that when pain threshold
contractions are compared with the submaximal 60% MVC contractions of normal controls, there is a significant difference and that can
be used as a useful classifier. This suggested twitch force may be a
passive compensation mechanism to maintain constant force output in painful muscles. It has been demonstrated that in repeated
contraction of shoulder muscles, there was a shift in the mean frequency of the surface EMG which was greater in patients than in
Author's Personal Copy
108
S. Kumar, N. Prasad / Clinical Biomechanics 25 (2010) 103–109
Back: Flexion 60%MVC/Threshold 95% CI
Patient
LES(µV)
RES(µV)
400
300
200
100
0
-100
LRA(µV)
400
300
200
100
0
-100
400
300
200
100
0
-100
RRA(µV)
RRA(µV)
400
300
200
100
0
-100
400
300
200
100
0
-100
LEO(µV)
400
300
200
100
0
-100
0
20
40
60
80
100
400
300
200
100
0
-100
REO(µV)
RES(µV)
LEO(µV)
400
300
200
100
0
-100
LRA(µV)
400
300
200
100
0
-100
REO(µV)
LES(µV)
Control
400
300
200
100
0
-100
400
300
200
100
0
-100
400
300
200
100
0
-100
0
Percentage of Task Cycle
20
40
60
80
100
Percentage of Task Cycle
Fig. 3. Magnitude and pattern of EMG traces obtained from patients and controls in a pain threshold/60% MVC contraction during trunk flexion.
controls (Sohn et al., 2004). No consistent significant differences between the mean frequency or the median frequency of the torso
muscles, between patients and controls, indicates that they had
similar spectral characteristics and insignificant differences in the
conduction velocity of the muscles (Pullman et al., 2000). However,
when a comparison in total power in different 10% bands was conducted (especially for pain tolerance/MVC), the rectus abdominis
and external oblique’s showed significant difference between patients and controls in many of the flexor activities. The erectors spinae similarly, showed significant differences in extensor activity for
all bands. These differences can have some value as classifier.
However, the time and frequency domain analyses subsequent
to wavelet analysis and logistic regression discrimination model
were deemed most useful in classifying patients and controls more
accurately and reliably. It is clear that the classification accuracy of
low back pain patients, obtained here, is lower than that of the cervical pain patients as published by Kumar et al. (2007) using the
same methodology. It is possible that the complexity of the lower
back pain and its motor pattern may have some obscuring effect. It
may also be possible that inclusion of several muscles in multiple
activities may have played an interactive role. In future studies
carefully selected homogenous patient groups, muscles, and muscle specific activities may produce single discriminative variable
with even better results with higher percentages being accurately
discriminated.
7. Limitations
The study has a few limitations. First, though the patients met the
inclusion and exclusion criteria, the sample is likely to have had different etiology of low back pain, which could not be determined for
the study. Furthermore, the age matching between patients and controls could not be achieved, though no significant differences were
found in strength of exertion between the two groups except in flexion at pain threshold for patients and 60% MVC for controls. Another
limitation of the study is that surface electrodes were used for EMG
recording. Every care was taken to avoid any cross talk but, surface
EMG being volume conducted could carry signals generated from
neighboring muscles. However, these activities would have been
synergistic. Additionally, the pick up from inserted electrodes would
have been from a very small volume of tissue and may not have reflected the general state of activity of these muscles.
8. Conclusions
(1) The strength of exertion between control and patients were
not significantly different at 20% MVC and MVC. However,
there was a significant difference in the strength of the
two groups during flexion at 60% MVC (controls) and pain
threshold (patients) (P < 0.001).
Author's Personal Copy
S. Kumar, N. Prasad / Clinical Biomechanics 25 (2010) 103–109
(2) The EMG scores in patients in all activities were greater than
those of controls at all levels of contraction.
(3) The median frequencies of different torso muscles were different and they also varied with activities within the same
muscles.
(4) A frequency banding into 10 percentile groups demonstrated significant differences in rectus abdominis (P <
0.01) and erector spinae (P < 0.05) in several bands between
patients and controls.
(5) Logistic discrimination at 20% MVC classified LBP patients
64.9% of times and controls 97.7% of times correctly. At pain
threshold the patients were correctly classified 95.1% of
times but the controls at 60% MVC were classified 86.8% of
times correctly. At pain tolerance/MVC levels patients and
controls were classified 74.3% and 86.4% of times correctly
respectively.
Acknowledgment
The financial support of Natural Science and Engineering Research Council of Canada is gratefully acknowledged. The assistance of Mr. Yogesh Narayan in data collection and Dr. Z. Siddiqi
in patient referral is gratefully acknowledged.
References
Ambroz, C., Scott, A., Ambroz, A., Talbott, E.O., 2000. Chronic low back pain
assessment using surface electromyography. J. Occup. Med. 42, 60–69.
Arendt-Nielsen, L., Graven-Nielsen, T., Svarrer, H., Svensson, P., 1996. The influence
of low back pain on muscle activity and coordination during gait: a clinical and
experimental study. Pain 64, 231–240.
Birch, L., Christiansen, H., Arendt-Nielsen, L., Graven-Nielsen, T., Sogaard, K., 2000.
The refers of experimental muscle pain on motor unit activity during low-level
contraction. Eur. J. Appl. Physiol. 83, 200–206.
Cholewicki, J., McGill, S.M., 1996. The mechanical of stability of the in vivo lumbar
spine: implications for injury and chronis low back pain. Clin. Biomech. 11, 1–15.
109
Christie, H., Kumar, S., Warren, S., 1995. Postural aberrations among low back pain
patients. Arch. Phys. Med. Rehabil. 76, 218–224.
Ervilha, U.F., Arendt-Nielsen, L., Duarte, M., Graven-Nielsen, T., 2004. Effect of load
level muscle pain intensity in the motor control of elbow flexion movements.
Eur. J. Appl. Physiol. 92, 168–175.
Farina, D., Arendt-Nielsen, L., Graven-Nielsen, T., 2005. Experimental muscle pain
reduces initial motor unit discharge rates during sustained submaximal
contractions. J. Appl. Physiol. 98, 999–1005.
Finneran, M.T., Mazanec, D., Marsolais, M.E., Marsolais, E.B., Pease, W.S., 2003.
Large-array surface electromyography in low back pain: a pilot study. Spine 28,
1447–1454.
Geisser, M.E., Haig, A.J., Wallbom, A.S., Wiggert, E.A., 2004. Pain-related fear, lumbar
flexion and dynamic EMG among persons with chronic musculoskeletal low
back pain. Clin. J. Pain 20, 61–69.
Graven-Nielsen, T., McArdle, A., Phoenix, J., Ardendt-Nielsen, L., Jensen, T.S., Jackson,
M.J., Edwards, R.H., 1997. In vivo model of muscle pain: quantification of
intramuscular chemical, electrical, and pressure changes associated with salineinduced muscle pain in humans. Pain 69, 137–143.
Hodges, P.W., Richardson, C.A., 1999. Altered trunk muscle recruitment in people
with low back pain with upper limb movement at different speeds. Arch. Phys.
Med. Rehabil. 80, 1005–1012.
Kumar, S., 2001. Theories of musculoskeletal injury causation. Ergonomics 44,
17–47.
Kumar, S., Narayars, Y., Prasad, N., Shuaib, A., Siddiqui, 2007. Cervical electromyogram
profile differences between patients of week pain and control. Spine 32, E246–
E253.
Lamoth, C.J., Daffertshofer, A., Meijer, O.G., Lorimer Moseley, G., Wuisman, P.I., Beek,
P.J., 2004. Effects of experimentally induced pain and fear of pain on trunk
coordination and back muscle activity during walking. Clin. Biomech. 19,
551–563.
Lundblad, I., Elert, J., Gerdle, B., 1998. Worsening of the neck and shoulder
complaints in humans are correlated with frequency parameters of
electromyogram recorded one year-earlier. Eur. J. Appl. Physiol. 79, 7–16.
Panjabi, M.M., 1992. The stability system of the spine. Part I. Function, dysfunction,
adaption, and enhancement. J. Spinal Disord. 5, 383–389.
Price, J.P., Clare, M.H., Ewehardt, F.H., 1948. The studies in low backache with
persistent muscle spasm. Arch. Phys. Med. 29, 703–709.
Pullman, S.L., Goodwin, D.S., Marquinez, A.I., Tabbal, S., Rubin, M., 2000. Clinical
utility of surface EMG. Report of the therapeutics and technology assessment
subcommittee of the american academy of neurology. Neurology 55,
171–177.
Sohn, M.K., Graven-Nielsen, T., Arendt-Nielsen, L., Svenson, P., 2004. Effects of
experimental muscle pain on mechanical properties of single motor units in
human masseter. Clin. Neurophysiol. 115, 76–84.