emg spectral characteristics of spinal muscles during...

Journal of Electromyography and Kinesiology 9 (1999) 21–37

EMG spectral characteristics of spinal muscles during isometricaxial rotation

Shrawan Kumar*, Yogesh NarayanDepartment of Physical Therapy, University of Alberta, 3–75 Corbett Hall, Edmonton, Alberta, T6G 2G4, Canada

Received 13 November 1997; received in revised form 23 February 1998; accepted 18 March 1998

Abstract

The objective of this study was to determine the frequency profile, median frequency (MF) and mean power frequency (MPF)of trunk muscles in an isometric graded maximal voluntary contraction (MVC) in isometric axial trunk rotation from a neutralupright seated posture. Twelve young healthy subjects (seven males, five females) were instrumented with surface electrodes ontheir external obliques, internal obliques, rectus abdominis, pectoralis, latissimus dorsi and erector spinae at T10 and L3 levelsbilaterally. These subjects were stabilized in seated posture in an axial rotation tester (AROT) and asked to perform a gradedisometric contraction of their maximal value to both right and left directions from a neutral posture within a period of 10 s. EMGfrom all 14 channels were sampled at 1 kHz at 10% intervals of MVC from 10% to MVC. These samples were subjected to fastFourier transform analysis. The frequency profile plots demonstrated the power of muscles involved in agonistic and antagonisticactivity. However, the frequency composition showed little difference between them. The MF was higher in agonists of the samemuscle. The MPF was always higher than MF. Both values were generally insignificantly different between different levels ofcontraction. However, with increasing level of contraction there was increase in power. 1999 Elsevier Science Ltd. All rightsreserved.

Keywords:Trunk muscles; Spectral parameters; EMG; Frequency profile

1. Introduction

Frequency-domain parameters of myoelectric signalsreveal aspects of motor function that are not possible todiscern by the use of amplitude of the same signals. Itis for this reason that the frequency characteristics ofseveral muscles have been studied and reported in theliterature [1–6]. These studies have revealed that thepower spectral characteristics or their relationship withindependent factors such as level of contraction are notuniform across muscles. Different muscles seem to havedifferent pattern as well as band width, which also arereported to be affected by the proportion of differentfibre types in the muscle in question.

Several studies have reported that the mean power fre-quency (MPF) of the power spectrum increases with thelevel of contraction as it rises from a fractional value to

* Corresponding author. Tel.:1 1-403-492-5979; fax:1 1-403-492-1626.

1050-6411/99/$ - see front matter 1999 Elsevier Science Ltd. All rights reserved.PII: S1050-6411 (98)00016-9

the maximum voluntary contraction (MVC) [7–10]. Asimilar pattern of behaviour is reported for median fre-quency as well [11]. Solomonow et al. [6], in an experi-ment using orderly stimulation of cat gastrocnemius,found a linear increase in the median frequency. On thecontrary, other studies have shown that there was norelationship between the median frequency and the mag-nitude of contraction [12–14]. Similarly, it has beenreported the MPF of the power spectrum was unrelatedto the magnitude of force [4,15]. However, Bilodeau etal. [1], in their experiment with triceps brachii andanconeus with graded contractions at the levels of 10,20, 40, 60, 80 and 100% MVC, found that the MPF ofthe anconeus increased significantly up to 60% of MVC.The authors reported no significant change in MPF ofthe triceps brachii across different levels of contraction.They later reported that the ramp and step contractionsaffected the MPF and MF differently in triceps brachiiand anconeous, although there was no statistically sig-nificant difference between them [2].

Yet another factor that is reported to have a significant

22 S. Kumar, Y. Narayan/Journal of Electromyography and Kinesiology 9 (1999) 21–37

effect on the spectral parameters of the EMG signals andits response to contraction level is the proportion of fibretypes present in any given muscle. Fast twitch muscletype (Type II) is known to be phasic fibre responding tothe need for higher force production. Therefore, theseare recruited when higher tension is evoked [16]. Theirgradual recruitment with increasing force level [17] issuggested to be the reason for the increase in MPF withincreasing level of contraction as observed and reportedby some workers [8,18]. The MPF of the EMG signalsis related to the fibre composition of the muscle [19,20].However, triceps brachii (65% fast twitch) and anconeus(65% slow twitch) responded in a manner exactlyopposite [1] to that expected from the foregoing logicwith increasing force level. This creates difficulty inexplaining the results. The authors suggested that a dif-ferential thickness of skin and intervening tissuesbetween the electrodes and the muscles would have haddifferent filtering characteristics, manifesting this result.Whether the foregoing was the cause of the effectsobserved remains to be established beyond doubt.Another study reported that up to 10% of variation inMPF was random variation [21]. Even this, however,does not explain the reported results of the foregoingstudy.

Most of the studies conducted and reported have beenon muscles of the upper and lower extremities, presum-ably due to the ease of access. In view of the fact thateven in the extremities different muscles show differentspectral characteristics and varying behaviour due toforce of contraction and fibre type composition, there islittle which can be extrapolated to the muscles of thetrunk. Since the motor behaviour of these muscles mayhave a role in and explain aspects of back injuries andpain, a spectral profiling of spinal muscles becomes use-ful. Since trunk twisting has been reported by severalauthors to be highly associated with back injuries [22–24,38], it was chosen as the activity for the study. It iswith this purpose that the spectral characteristics of theerector spinae, latissimus dorsi, external obliques,internal obliques, rectus abdominis and pectoralis majorwere studied and are reported here.

The objective of this study was to describe the fre-quency profile of trunk muscles in a standardized gradualand graded isometric maximal voluntary contraction intrunk axial rotation from a neutral posture. Furthermore,another aim was to determine the median and meanpower frequencies of these trunk muscles and comparethem at different levels of contraction between agonistsand antagonists.

2. Materials and methods

2.1. Subjects

Data from 12 normal young and asymptomatic sub-jects was recorded. The experimental sample consisted

of seven males (mean age 25.1 yr, standard deviation5.3 yr; mean weight 69.1 kg, standard deviation 5.8 kg;and mean height 176.2 cm, standard deviation 5.2 cm)and five females (mean age 21.6 yr, standard deviation2.8 yr; mean weight 57.9 kg, standard deviation 11.4 kg;and mean height 166.1 cm, standard deviation 9.9 cm).These subjects were screened for neuromuscular andmusculoskeletal disorders, and any spinal or abdominalsurgery. The subjects were informed about the objectivesand procedures of the study and they signed an infor-med consent.

2.2. Equipment

2.2.1. Axial rotation tester (AROT)The axial rotation tester (AROT) was a device speci-

ally designed and fabricated to obtain reliable, repeatableand standardized axial twisting of the human trunk, pre-venting any flexion or extension. As axial rotation is amotion coupled with lateral flexion, the device permittedfree lateral flexion thus providing a floating axis forrotation. Because the device tests subjects in seated pos-ture, it also eliminates any contribution of the hips andlower extremities in twisting of the trunk. The device isdescribed in its entirety by Kumar [25]. Briefly, it isdesigned to stabilize lower extremities hip down in aseated posture and the shoulders such that the verticalspinal axis is aligned to the device’s rotation axis. Suchstabilization allows free twisting motion to occur onlyin the thoracolumbar region with accompanying lateralflexion. The AROT was equipped with a load cell anda precision potentiometer to provide continuous outputof the twisting force applied and the angular motion ach-ieved. The output of the load cell was divided into two;one was fed to the data collection system and the otherwas fed to a digital display device for feedback to thesubject.

2.2.2. EMG systemThe EMG system consisted of surface electrodes,

electrode cables, preamplifiers and amplifiers. Silver–sil-ver chloride circular surface electrodes of 1 cm diameterand recessed pregelled elements (HP 144445) were usedwith inter-electrode distance of 2 cm. These electrodeswere connected to 16-channel, fully isolated, low-noiseamplifiers. These amplifiers had low non-linearity, highcommon mode rejection ratio (130 dB) and a wide band-width (25 MHz). These preamplifiers fed to a low-power, high-accuracy instrumentation amplifier designedfor signal conditioning and amplification. The amplifiersystem was run off an internal charged battery. Theamplifier had AC-coupled inputs with a single-pole RCbandpass filter with a low cut-off frequency at 8 Hz andhigh cut-off at 500 Hz. The preamplifiers and amplifierswere built by Measurement Systems, Inc., Ann Arbor,Michigan.

23S. Kumar, Y. Narayan/Journal of Electromyography and Kinesiology 9 (1999) 21–37

2.2.3. Controller and A/D boardThe outputs of the AROT load cell, precision poten-

tiometer and EMG amplifiers were fed to a MetraByteDAS 20 A/D board. These signals were sampled at1 kHz. The sampled signals were stored in the hard diskof a 486 computer with a tape backup (Colorado Mem-ory Systems Inc.) for further analysis and interpretation.

2.3. Experimental procedures

The subjects were weighed and measured for theirheight. Their age was also recorded. These subjects wereinstrumented with 14 pairs of disposable, pregelled, sur-face electrodes (HP 144445) at an inter-electrode dis-tance of 2 cm after suitable preparation of the skin withan alcohol–acetone mixture. These electrodes wereplaced on erector spinae levelled with spinous processesof T10 and L3 vertebrae bilaterally, 4 cm lateral to thetips of the spinous processes. Surface electrodes werealso applied to the left and right latissimus dorsi. On theventral side, surface electrodes were applied bilaterallyto the pectoralis major, rectus abdominis, externaloblique and the internal oblique (in the area of externaloblique aponeuroses to minimize overlap with it). Aground electrode was applied to anterosuperior iliacspine.

Prepared subjects were seated in the chair of the axialrotation tester. The seat was adjusted for height so thatthe subjects were seated comfortably resting their feet,with the knee at 90° angle. The seat was then alignedwith the axial rotation tester harness, which was loweredon the subjects’ shoulders and fastened. The subjectswere stabilized in this upright neutral posture, seatedposition hip down, by using four velcro straps at the hip,distal thigh, proximal shin and ankle.

The circular disc above the shoulder harness wasattached to an immovable object by means of an airplanecable with the load cell in its path but leaving no slack.The subjects were then asked to attempt maximal volun-tary contraction (MVC) in torso rotation by applying aforce through their shoulders on the harness, which waslocked in position through the airplane cable and main-tained the isometric condition. The rotation wasattempted such that force gradually increased up to MVCwithin a period of 10 s. Since the trial was started afterthe data acquisition began and was terminated before theend of the sampling period, the total duration of contrac-tion lasted 5 to 7 s from start to finish. In order to obtaingraded contraction the subjects were provided with vis-ual feedback. The subjects were required to practise thisactivity several times before the day of the experimentfor them to become familiar with the process and be ableto produce consistent torque with no motion of the trunk.

2.4. Data acquisition

Data were acquired using a custom-designed modularsoftware for the project. It allowed input of subject dataand created data files. Subsequently it acquired the dataaccording to predetermined variables (e.g., samplingrate, duration, etc.). All 14 channels were sampled at1 kHz for a window of 10 s with the entire trial lasting5 to 7 s. From the latter, segments of 0.4 s (0.2 s beforereaching the desired force and 0.2 s after the force wasreached) were obtained for data extraction, frequencyprofiling and further analyses.

2.5. Data analysis

Previously collected data were loaded into the com-puter memory. The EMG samples of all 14 channels(erector spinae at T12 and L3 levels, latissimus dorsi, pec-toralis major, rectus abdominis, external and internalobliques bilaterally) were processed in the time domainat levels of 10%, 20%, 30%, 40%, 50%, 60%, 70%,80%, 90% and 100% contraction. The power spectrawere calculated from the raw signals by means ofWelch’s method [26]. The method involves sectioningthe record and averaging modified periodograms of thesections. Thus the sampled signals were divided intothree equal segments of 256 points in length, processedthrough a type 1 Welch window and then subjected topower spectral analysis. The window had a taperedshape which attenuated the end points; therefore, a frac-tional overlap of 0.715 was used to marginally recoverthe samples at the end points. The overlapping alsoallowed the variance in the estimation to be reduced,while maintaining a desired spectral resolution anddependency between segments. For each of the samplingperiods, the 401 points were divided into three equidis-tant sliding and overlapping segments of 256 points.From each segment the average value was subtracted forDC removal and then a Welch window applied to it.Subsequently the power spectrum of the segment wascalculated. Since Fast Fourier transfrom (FFT) assumessignal stationarity, its non-stationarity was also checked.For the final power spectrum the average of the threesegment spectra was taken. The latter was then smoothedwith linear polynomial smoothing using seven point seg-ments and repeating once. From these power spectra thefrequency profile of the muscle was obtained.

The frequency values for each of the channels at eachof the task grades were plotted against the power spec-tral density.

3. Results

3.1. Torque

In their maximal voluntary contraction for axialrotation to the left and right, the male sample generated


Tab

le1

Mea

nto

rque

(Nm

)an

dan

gula

rde

viat

ion

(deg

.)fr

omth

ene

utra

lpo

stur

edu

ring

max

imal

and

grad

edax

ial

rota

tion

Var

iabl

eG

ende

rD

irect

ion

Gra

des

ofco

ntra

ctio

n

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

MS

DM

SD

MS

DM

SD

MS

DM

SD

MS

DM

SD

MS

DM

SD

Tor

que

Mal

e0→

left

22.4

10.3

31.2

11.8

41.8

14.3

53.1

17.0

58.6

23.3

70.9

30.8

82.9

34.8

97.5

42.4

111.

947

.112

2.7

52.7

(Nm

)0→

right

21.5

5.9

28.5

6.7

38.2

11.7

48.3

14.9

54.1

17.8

68.0

19.6

78.7

22.7

92.3

31.0

104.

734

.011

5.4

37.0

Fem

ale

0→le

ft11

.03.

714

.14.

117

.45.

922

.19.

728

.111

.234

.911

.940

.113

.746

.115

.852

.218

.658

.720

.30→

right

10.5

3.7

14.5

4.5

21.3

6.3

26.3

8.7

33.7

9.4

40.1

10.7

46.0

12.2

53.6

15.3

60.8

17.1

67.9

18.6

Ang

ular

Mal

e0→

left

2.2

2.3

1.5

1.9

0.9

1.7

0.3

1.5

0.3

1.32

0.2

1.2

20.

61.

22

1.0

1.4

21.

61.

72

2.1

1.9

devi

atio

n0→

right

0.7

1.7

1.3

1.2

2.1

1.2

2.8

1.5

2.8

2.2

3.3

2.3

3.8

2.7

4.2

3.2

4.6

3.5

5.1

3.7

(deg

.)F

emal

e0→

left

3.4

2.9

2.9

2.9

2.6

3.0

2.3

3.1

1.5

2.7

0.8

2.0

0.4

1.8

0.0

1.720.

21.

62

0.5

1.6

0→rig

ht2

0.2

1.9

0.3

1.5

0.8

1.7

1.1

1.7

1.6

1.4

1.9

1.4

2.2

1.4

2.6

1.5

2.8

1.4

3.5

1.5

M—

mea

n;S

D—

stan

dard

devi

atio

n.


Fig. 1. Power spectra of left and right external obliques in a gradual contraction to maximal voluntary isometric contraction in attempted rightwardaxial rotation of the trunk from a neutral posture.

a mean torque of 122.7 N m and 115.4 N m, respect-ively. The female sample, on the other hand, generatedmean MVC torques of 58.7 N m and 67.9 N m, respect-ively. Thus females generated torques which were closeto half of those of males. The details are presented in

Table 1. The deviation of the torso in the axial planewas less than 5°, thus staying very close to the neutralposition.


Fig. 2. Power spectra of left and right internal obliques in a gradual contraction to maximal voluntary isometric contraction in an attemptedrightward axial rotation of the trunk from a neutral posture.

3.2. Frequency profile

The frequency–power plots of the 14 muscles of asample subject for graded levels of contraction arepresented in Figs. 1–7. The power of primary agonistmuscles was always considerably higher than those ofthe respective antagonistic muscles. This was obvious by

comparing contralateral external obliques with ipsilateralexternal obliques; and ipsilateral latissimus dorsi,internal obliques and erector spinae with their respectivecontralateral counterparts. These plots also clearly dem-onstrate little change in the median frequency withincreasing grades of contraction, except an increase intheir power.


Fig. 3. Power spectra of left and right rectus abdominis in a gradual contraction to maximal voluntary isometric contraction in an attemptedrightward rotation of the trunk from a neutral posture.

In antagonistic internal oblique, past 75% MVC, therewere two frequencies with greater power than the rest.It should also be pointed out that, for this muscle, powerwas quite low. For ipsilateral rectus abdominis suchdeviation started much sooner, perhaps around 30%MVC. The ipsilateral pectoralis had very little power buta wide range of frequencies from 0 to 500 Hz seemed tobe present. Contrary to this, the contralateral pectoralis

demonstrated high power level and a considerably nar-rower band of frequencies, the majority of these under150 Hz. In erector spinae, the ipsilateral muscles demon-strated a wider frequency band (8 –250 Hz comparedwith 8 –150 Hz in contralateral muscles) as well ashigher power (30 nW) than contralateral (1.2 nW). Theerector spinae at thoracic level had similar power as thatat the lumbar level (30 nW).


Fig. 4. Power spectra of left and right pectoralis in a gradual contraction to maximal voluntary isometric contraction in an attempted rightwardrotation of the trunk from a neutral posture.

3.3. Mean median frequency

The mean median frequencies of the 14 muscles(external and internal obliques, rectus abdominis, pec-toralis major, latissimus dorsi, erector spinae at 10th tho-racic and 3rd lumbar vertebral levels bilaterally) arepresented in Tables 2 and 3 for all grade levels andmaximum voluntary contraction, for both left and right

axial rotations, and for both male and female samples.The mean median frequencies were always higher inagonist muscles as compared with antagonists and stabil-izers at all levels of contraction (for example, at MVC,the mean median frequencies (Hz) were: contralateralexternal obliques 75 vs. 51, ipsilateral latissimus dorsi76 vs. 42, internal oblique 61 vs. 52, erector spinae atL3 55 vs. 46 and erector spinae at T10 65 vs. 52 among


Fig. 5. Power spectra of left and right latissimus dorsi in a gradual contraction to maximal voluntary isometric contraction in an attemptedrightward rotation of the trunk from a neutral posture.

males, and similar values among females, see Tables 2and 3). Generally, different grades of contractions didnot evoke significantly different median frequency(Tables 2 and 3). For example, the median frequencies(Hz) at 10% and MVC were 64 and 75, 52 and 61, 41and 51, 39 and 52, 72 and 76, 33 and 42, 50 and 55, 33and 46, 55 and 65, 33 and 52 for left external obliques,

left internal obliques, right external obliques, rightinternal obliques, left latissimus dorsi, right latissimusdorsi, left erector spinae at L3, right erector spinae at L3,left erector spinae at T10 and right erector spinae at T10,respectively. In erector spinae the median frequency ofthe EMG signals of the ipsilateral and contralateralmuscles were generally within 15 Hz of each other, rang-


Fig. 6. Power spectra of left and right erector spinae at T10 in a gradual contraction to maximal voluntary isometric contraction in an attemptedrightward rotation of the trunk from a neutral posture.

ing between 44 and 68 Hz. The pectoralis muscle, how-ever, had more variable median frequency in the range of48–108 Hz and 30–60 Hz for ipsilateral and contralateralmuscles, respectively, among males. In females, theywere even more variable, ranging between 40 and113 Hz for ipsilateral muscles and between 38 and 61 Hz

for contralateral muscle. Among female subjects the rec-tus abdominis was found to be more variable than amongmale subjects (Tables 2 and 3).


Fig. 7. Power spectra of left and right erector spinae at L3 in a gradual contraction to maximal voluntary isometric contraction in an attemptedrightward rotation of the trunk from a neutral posture.

3.4. Mean power frequency

The mean power frequency for all 14 muscles forgraded and maximal voluntary contraction during axialrotation to left and right are presented in Tables 4 and5 for male and female samples, respectively. The meanpower frequencies for all muscles in both genders werefound to be approximately 20 Hz higher than the corre-

sponding median frequency. The pattern of the meanpower frequency was similar to that observed for medianfrequency, being higher for the agonists compared withthe antagonists except for erector spinae at thoracic aswell as lumbar levels (for instance, in neutral to leftwardisometric axial rotation among males, contralateral ver-sus ipsilateral external obliques—99 vs. 60, 111 vs. 58,107 vs. 67, and 100 vs. 58 Hz for 25%, 50%, 75% and


Table 2Mean median frequencies (Hz) of individual muscles during maximal and graded contractions in left and right axial rotations among males

Muscles 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

M SD M SD M SD M SD M SD M SD M SD M SD M SD M SD

Neutral to left grade of contraction (% MVC)LEO 64 31 66 29 69 28 82 24 74 25 70 22 71 18 71 19 74 25 75 23LIO 52 15 55 15 60 14 63 15 61 18 65 16 69 13 69 13 68 18 61 11REO 41 10 41 10 43 10 47 9 50 12 51 13 50 13 51 11 51 13 51 9RIO 39 21 45 26 44 27 44 15 33 14 42 12 46 16 46 18 54 16 52 14LRA 46 34 53 38 64 36 75 39 69 27 82 23 84 29 85 33 74 36 80 34RRA 32 11 37 14 51 29 57 17 56 25 63 19 67 21 73 23 70 26 80 17LP 48 57 49 59 52 64 63 71 54 56 76 59 89 47 108 61 81 66 78 46RP 45 16 44 16 44 16 52 23 46 14 47 14 46 8 46 6 50 10 60 16LLD 72 20 78 19 83 17 85 15 79 26 83 28 83 24 80 19 74 21 76 15RLD 33 8 37 7 37 7 37 5 37 7 38 4 39 4 47 9 40 8 42 8LL3 50 11 52 11 52 8 54 9 50 9 47 7 45 8 49 11 55 14 55 14RL3 33 9 39 10 45 14 47 13 45 12 46 6 50 13 56 13 49 7 46 18LT10 55 15 56 13 58 13 60 13 63 16 66 17 68 16 73 19 70 20 65 15RT10 33 11 39 10 45 21 45 8 44 13 57 15 51 11 57 6 49 12 52 20Neutral to right grade of contraction (% MVC)LEO 42 9 44 8 45 7 46 7 44 7 46 11 50 10 54 12 46 11 50 11LIO 32 19 33 19 31 17 39 15 47 20 49 18 52 20 50 23 50 24 52 9REO 52 21 60 17 73 20 80 25 71 14 65 14 68 24 85 34 80 26 68 24RIO 52 32 55 27 59 25 61 24 63 20 69 28 64 14 74 17 76 12 79 15LRA 45 30 41 29 47 24 63 14 71 17 59 24 72 19 63 19 71 26 62 31RRA 58 38 55 45 51 28 68 36 75 22 71 32 84 35 84 35 81 23 88 38LP 38 23 30 10 35 11 48 23 48 26 45 27 53 15 52 13 51 16 59 12RP 56 36 63 37 64 38 86 51 68 23 82 64 100 70 107 74 93 64 87 59LLD 36 4 36 4 37 4 40 5 40 4 38 4 39 6 40 5 42 4 40 8RLD 89 33 88 23 98 29 92 6 83 19 84 14 83 10 86 11 86 12 88 10LL3 46 18 41 14 37 11 47 11 45 15 40 13 51 14 55 24 47 10 49 10RL3 48 12 49 10 47 10 49 11 46 10 49 11 52 10 49 10 52 8 60 19LT10 52 34 43 17 46 18 56 21 55 20 49 18 54 14 57 21 46 8 44 10RT10 51 11 52 14 55 16 61 14 59 14 64 16 64 15 65 14 66 13 68 12

M—mean;SD—standard deviation; LEO—left external oblique; LIO—left internal oblique; REO—right external oblique; RIO—right internaloblique; LRA—left rectus abdominis; RRA—right rectus abdominis; LP—left pectoralis; RP—right pectoralis; LLD—left latissimus dorsi; RLD—right latissimus dorsi; LL3—left erector spinae at L3; RL3—right erector spinae at L3; LT10—left erector spinae at T10; RT10—right erector spinaeat T10.

MVC respectively; ipsilateral versus contralateral latis-simus dorsi—87 vs. 59, 96 vs. 59, 98 vs. 65, and 90vs. 57 Hz for 25%, 50%, 75% and MVC respectively;ipsilateral versus contralateral internal obliques—76 vs.52, 78 vs. 47, 83 vs. 57, and 80 vs. 67 Hz for 25%,50%, 75% and MVC respectively; and ipsilateral versuscontralateral erector spinae at thoracic level—73 vs. 76,80 vs. 79, 88 vs. 90, and 78 vs. 83 Hz for 25%, 50%,75% and MVC respectively). Similar patterns werefound for other conditions (neutral to right) amongmales, and both activities among females (Tables 4 and5). However, between the grades of contractions, themean power frequency was more variable than the meanmedian frequency with a spread of approximately 10 Hzexcept in pectoralis, where the maximum spread was33 Hz in males and 63 Hz in females.

4. Discussion

Axial rotation is an asymmetric and mechanicallycomplex activity which involves some muscles in agon-istic role (contralateral external oblique, ipsilateralinternal oblique, latissimus dorsi and erector spinae) andtheir contralateral counterparts in antagonistic role [27].However, significant and strong EMG muscle activity incontralateral muscles has been reported as well [28]. Asignificant activity among muscles such as erector spinaeand rectus abdominis, which do not have their fibres ori-ented in the direction of motion, could possibly be serv-ing the function of stabilizing the spine. Because of theincongruence of the fibre orientation with the axis ofmotion, their contraction cannot be contributingefficiently to the spinal mechanics even in a stabilizingrole. However, ipsilateral external obliques and contrala-


Table 3Mean median frequencies (Hz) of individual muscles during maximal and graded contractions in left and right axial rotations among females

Muscles 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%




teral internal obliques have their fibre orientation betteraligned in the direction of the motion and thereby areoptimally suited to act as antagonists in eccentric modeof contraction. Hence, in order to provide spinal stab-ility, the mechanically inefficient contraction of non-aligned muscles is likely to be at a significantly higherlevel than what would be necessary had they beenaligned. It is, therefore, surmized that these muscles willalso be contracting at a high level, evoking significantrecruitment of their motor units in addition to eccentriccontraction of the antagonistic obliques.

Spectral parameters, e.g., median frequency and meanpower frequency, are dependent on the conduction velo-city [29–31] and the conduction velocity increases withthe diameter of fibres. Thus it stands to reason that as amuscle shortens in contraction, it will increase its diam-eter, thereby increasing conduction velocity [4,32]. Theincreased conduction velocity will manifest itself inhigher values of median frequency and mean power fre-

quency. Such was the case in the results obtained in thisexperiment for the agonistic muscles. However, theimpact of the conduction velocity alone on the MF andMPF is not expected to be large [32]. A contracted andshortened muscle requires higher rates of stimuli to pro-duce high tension compared with a longer muscle [33].In addition, to produce and/or maintain a force, anadditional excitation is required [34,35]. Also in a short-ened agonist muscle one will need a greater number ofactive muscle fibres to generate the desired force due tothe maximized cross-bridging between actin and myosinfilaments. Finally, the firing rate is dependent on themechanical properties of the muscle [36]. In a contractedmuscle there is a decrease in dynamic stiffness [33] anddamping [37]. Such changes would require considerablyhigher firing rates to generate and produce a force. Allof the foregoing phenomena are expected to have a com-bined effect on the median frequency and mean powerfrequency.


Table 4Mean power frequency (Hz) of individual muscles during maximal and graded contractions in left and right axial rotations among males

Muscles 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%




It would appear that in an asymmetric isometric con-traction, while the agonist muscles contract, the antagon-istic and stabilizer muscles will also be contracting. Aconsiderably lower value for MF and MPF among non-aligned antagonistic and stabilizer would tend to indicatethat the force production and accompanying level of con-traction would probably be significantly lower. However,reports of strong antagonistic muscle activity in axialrotation tend to confuse the picture. The latter especiallytrue when it is not possible to differentiate between theforces produced by agonists and antagonists separately.It is, however, likely that the antagonist oblique musclesmay undergo a small amount of lengthening whileundergoing eccentric contraction and may not be respon-sible for the production of similar magnitudes of forcesas agonsits. However, since these are likely to providea significant component of the antagonistic force, theytoo demonstrated high MF and MPF.

In a study of power spectra of elbow extensors com-

paring ramp and step isometric contractions, the authorsreported a relatively gradual and steady increase in theMF with the level of ramp contraction, and rather stableand similar values for various levels of contraction forthe step contraction [2]. In the experiment reported herethe trend of the values did not correspond with those ofBiladeau et al. [2]. In most cases (erector spinae at bothlevels, latissimus dorsi and external obliques bilaterally)the MF had insignificantly different values for all levelsof contraction. In some abdominal muscles (generallyinternal oblique) the value of MF for 25% level of con-traction was lower than the rest of the other values. How-ever, a clearcut trend did not emerge. These stable valuesof MF for a ramp contraction in trunk muscles areclearly different from what was reported in elbowmuscles. The significance of such an observationremains unclear. However, the trunk, being a large partof the body controlled by spinal muscles which are gen-erally responsible for gross motion, may have large


Table 5Mean power frequency (Hz) of individual muscles during maximal and graded contractions in left and right axial rotations among females

Muscles 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%




motor unit territory. Due to the inefficient mechanicalmileu and need for a stronger contraction, motor unitrecruitment may not be as important a process in gener-ation of greater force as the firing rate. Furthermore, spi-nal muscles do have a significant proportion of type Ipostural muscle fibres, which also reduces the need aswell the chance for increasing recruitment to be a domi-nant process in force production. A more plausibleexplanation for the wide variation in MF and MPF ofthe pectoralis muscle lies in variable contraction of themuscle. As these muscles were not one of the primarymovers they varied their contraction during the courseof contraction, phasing in and out at perhaps strategictime as may have been perceived by other contractingmuscles for producing the required force magnitude.

5. Conclusions

This study reports the frequency profile of the trunkmuscles during a standardized flexion/extension free

axial rotation. The latter demonstrates that with increas-ing force the frequency composition changes little butthe magnitude of power increases considerably.

A consistent significant difference in the magnitudesof median frequency and mean power frequency wasfound. Therefore, in axial rotation, for the trunk musclesthe median frequency and mean power frequency shouldnot be equated. Both the median and mean power fre-quencies for the agonist muscles were significantlyhigher than those found for antagonistic muscles, exceptfor ipsilateral external oblique and contralateral internalobliques. Orientation of these muscles was in the direc-tion of motion and they were optimally positioned for asignificant antagonistic activity, thereby demonstratinga high level of MF and MPF. With the isometric trialconditions the results indicate that the agonistic muscleshad undergone shortening, producing more force thanthe antagonistic muscle, which may have lengthenedslightly. However, this was not the case for ipsilateralexternal obliques and contralateral internal obliques,which may have been contracting eccentrically.


The spectral parameters of the prime movers and stab-ilizers demonstrated a steady pattern, whereas the sec-ondary muscles such as pectoralis were variable.

References

[1] Bilodeau M, Arsenault AB, Gravel D. The influence of anincrease in the level of force on the EMG power spectrum ofelbow extensors. Eur J Appl Physiol 1990;61:461–6.

[2] Bilodeau M, Arsenault AB, Gravel D, Bourbonnais D. EMGpower spectra of elbow extensors during ramp and step isometriccontractions. Eur J Appl Physiol 1991;63:24–8.

[3] DeLuca CJ, Sabbahi MA, Roy SH. Median frequency of themyoelectric signal. Effect of hand dominance. Eur J Appl Physiol1986;55:457–64.

[4] Rosenburg R, Seidel H. Electromyography of lumbar erector spi-nae muscles influence of posture, interelectrode distance, strengthand fatigue. Eur J Appl Physiol 1989;59:104–14.

[5] Roy SH, Deluca CJ, Schneider J. Effects of electrode locationson myoelectric conduction velocity and median frequency esti-mates. J Appl Physiol 1986;61:1510–7.

[6] Solomonow M, Baten C, Smit J, Baratta R, Hermens H, D’Am-brosia R, Shoji H. Electromyogram power spectrafrequenciesassociated with motor unit recruitment strategies. J Appl Physiol1990;68:1177–85.

[7] Gerdle B, Eriksson NE, Hagberg C. Changes in the surface elec-tromyogram during increasing isometric shoulder forwardflexions. Eur J Appl Physiol 1988;57:404–8.

[8] Hagberg M, Ericson BE. Myoelectric power spectrum depen-dence on muscular contraction level of elbow flexor. Eur J ApplPhysiol 1982;48:147–56.

[9] Komi PV, Viitasalo JHT. Signal characteristics of EMG at differ-ent levels of muscle tension. Acta Physiol Scand 1976;96:267–76.

[10] Kwatny E, Thomas DH, Kwatny HG. An application of signalprocessing techniques to the study of myoelectric signals. IEEETrans Biomed Eng 1970;17:303–13.

[11] Kranz H, Williams AM, Cassell J, Caddy DJ, Silberstein RB.Factors determining the frequency content of the electromyog-ram. J Appl Physiol Respir Environ Exerc Physiol 1983;55:392–9.

[12] Petrofsky JS, Lind AR. Frequency analysis of the surface electro-myogram during sustained isometric contractions. Eur J ApplPhysiol 1980;43:173–82.

[13] Petrofsky JS, Lind AR. The influence of temperature on theamplitude and frequency components of the EMG during briefand sustained isometric contractions. Eur J Appl Physiol1980;44:189–200.

[14] Petrofsky JS, Glaser RM, Phillips CA, Lind AR, Williams C.Evaluation of the amplitude and frequency components of thesurface EMG as an index of muscle fatigue. Ergonomics1982;25:213–23.

[15] Tarkka IM. Power spectrum of electromyography in arm and legmuscles during isometric contractions and fatigue. J Sports Med1984;24:189–94.

[16] Goldberg LJ, Derfler B. Relationship among recruitment order,spike amplitude, and twitch tension of single motor units inhuman masseter muscle. J Neurophysiol 1977;40:879–90.

[17] Sypert GW, Munson JB. Basis of segmental motor, control:motoneuron size or motor unit type? Neurosurgery 1981;8:608–21.

[18] Moritani T, Muro M. Motor unit activity and surface electromy-ogram power spectrum during increasing force of contraction.Eur J Appl Physiol 1987;56:260–5.

[19] Gerdel B, Eriksson NE, Brundin L, Edstrom M. Surface EMGrecordings during maximum static shoulder forward flexion indifferent positions. Eur J Appl Physiol 1988;57:415–9.

[20] Gerdle B, Eriksson NE, Brundin L. The behaviour of the meanpower frequency of the surface electromyogram in biceps brachiiwith increasing force and during fatigue. Electromyogr Clin Neu-rophysiol 1990;30:483–9.

[21] Oberg T, Sandsjo L, Kadefors R. Variability of the EMG meanpower frequency: a study of the trapezius muscle. J ElectromyogrKinesiol 1991;1(4):237–43.

[22] Frymoyer JW, Pope MH, Clements JH, Wilder DG, McPhearsonB, Ashikaga T. Risk factors in low back pain: an epidemiologicstudy. J Bone Joint Surgery 1983;65a:213–8.

[23] Frymoyer JW, Pope MH, Costanza MC, Rosen JC, Goggen JF,Wilder DG. Epidemiologic studies of low back pain. Spine1980;5:419–23.

[24] Manning DP, Mitchell RG, Blanchfield LP. Body movements andevents contributing to accidental and non-accidental back injur-ies. Spine 1984;9:734–49.

[25] Kumar S. Isolated planar trunk strength and mobility measure-ment for the normal and impaired backs: part 1—the devices. IntJ Ind Erg 1996;17:81–90.

[26] Welch PD. The use of fast Fourier transform for the estimationof power spectra: a method based on time averaging over short,modified periodograms. IEEE Trans on Audio and Electroacoust-ics 1967;15:70–3.

[27] Kumar S, Narayan Y, Zedka M. An electromyographic study ofunresisted trunk rotation with normal velocity among healthy sub-jects. Spine 1996;21(13):1500–12.

[28] Pope MH, Andersson GBJ, Broman H, Svensonn M, ZetterbergC. Electromyographic studies of the lumbar trunk musculatureduring the development of axial torques. J Ortho Res1986;4:288–97.

[29] Bazzy AR, Korten JB, Haddad GG. Increase in electromyogramlow-frequency power in nonfatigued contracting skeletal muscle.J Appl Physiol 1986;61(3):1012–7.

[30] DeLuca CJ. Myoelectric manifestations of localized muscularfatigue in humans. Crit Rev Biomed Eng 1984;11:251–79.

[31] Morimoto S. Effects of length change in muscle fibres on conduc-tion velocity in human motor units. Japanese J Physiol1986;36:773–82.

[32] Sadoyama T, Masuda T. Changes of the average muscle fibreconduction velocity during a varying force contraction. Electroen-ceph Clin Neurophysiol 1987;67:495–7.

[33] Houk JC, Rymer WZ. Neural control of muscle length and ten-sion. In: Brooks VB, editor. Handbook of physiology, vol. 2, part1. Baltimore (MA): Waverley Press, 1981:257–323.

[34] Agarwal GC, Gottlieb GL. Mathematical modeling and simul-ation of the postural control loop. 1. CRC Crit Rev Biomed Eng1982;8:93–134.

[35] Solomonow M, Baratta R, Zhou BH, Shoji H, D’Ambrosia RD.The EMG-force model of electrically stimulated muscle: depen-dence on control strategy and predominant fiber composition.IEEE Trans Biomed Eng BME 1987;34:692–703.

[36] Freund HJ. Motor unit and muscle activity in voluntary motorcontrol. Physiol Rev 1983;63:387–436.

[37] Saziorski WM, Aruin AS, Selujanow WN. Biomechanik desmenschlichen Bewegungsapparates. Berlin: Sportverlag, 1984.

[38] Schaffer H. Back injuries associated with lifting. Bulletin 2144.Washington (DC): US Department of Labor, Bureau of Statistics,1982:1–20.


Shrawan Kumar is currently a Professor inphysical therapy in the Faculty of RehabilitationMedicine and in the Division of Neuroscience,Faculty of Medicine. He joined the Faculty ofRehabilitation Medicine in 1977 and rose to therank of Full Professor in 1982. Dr Kumar holdsBSc (biology and chemistry) and MSc (zoology)degrees from the University of Allahabad, India,and a PhD (human biology) degree from theUniversity of Surrey, U.K. Following his PhDhe did his post-doctoral work at Trinity College,Dublin, in engineering, and worked as a

Research Associate at the University of Toronto in the Department ofPhysical Medicine and Rehabilitation. For his life-time work, Dr Kumarwas recognized by the University of Surrey, U.K. by the award of a DScdegree in 1994. Dr Kumar was invited as a Visiting Professor for the year1983–84 at the University of Michigan, Department of Industrial Engin-eering. He was a McCalla Professor 1984–85.

Dr Kumar has over 200 scientific peer-reviewed publications, and worksin the area of musculoskeletal injury causation/prevention with specialemphasis on low-back pain. He has edited/authored seven books/monographs. He currently holds a grant from NSERC. His work has been

supported in the past, in addition to the above, by MRC, WCB and NRC.He has supervised or is supervising 10 MSc students, three PhD students,and two post-doctoral students. He is Editor of theInternational Journalof Industrial Ergonomics, Consulting Editor ofErgonomics, Advisory Edi-tor of Spine, and Assistant Editor of theTransactions of RehabilitationEngineering. He serves as a reviewer for several other international peer-reviewed journals. He also acts as a grant reviewer for NSERC, MRC,Alberta Occupational Health and Safety, and BC Research.

Yogesh Narayan obtained his BSc inelectrical/electronics engineering from the Uni-versity of Alberta, Canada. He specialized indigital signal processing and microprocessor-based systems design. After graduating, heworked on research projects in biomedicalengineering, at the Grey Nuns Hospital, Edmon-ton, Alberta, Canada. Currently, he is a ResearchAssistant for Dr Shrawan Kumar at the Univer-sity of Alberta.

emg spectral characteristics of spinal muscles during...

Documents