- Email: [email protected]
Frequency response of spine extensors during rapid isometric contractions: effects of muscle length and tension
Journal of Electromyography and Kinesiology 8 (1998) 227–232
Frequency response of spine extensors during rapid isometric contractions: effects of muscle length and tension Lisa C. Brereton, Stuart M. McGill
*
Occupational Biomechanics and Safety Laboratories, Department of Kinesiology, Faculty of Applied Health Sciences, University of Waterloo, Waterloo, Ontario, Canada
Abstract During muscle contraction, electrical activity necessarily precedes force output, yet models that utilize processed electromyograms sometimes predict force as preceding EMG under rapid ballistic loading conditions. The purpose of this study was to define the frequency response transfer function of the upper and lower erector spinae musculature, at different lengths and tensions, using rectified, low pass filtered EMG. This would enable accurate estimates of force from the processed electromyogram, specifically during impulsive contractions. Abdominal and erector spinae EMG were measured in synchrony with impulsive low back moments in five men. EMG signals were rectified and low pass filtered repeatedly with cut-off frequencies from 1 to 3 Hz at 0.5 Hz increments in order to quantify the frequency response. It was found that EMG signals processed through a simple, Butterworth low pass filter could not produce the measured force output without an additional time shift. These shifts were quantified by cross-correlating EMG and force with increments of 1 ms. In order to define the transfer function of EMG to force, optimal cut-off frequencies were selected two ways: quantitatively by searching for maximum cross correlations coefficients, and qualitatively. Results indicated that the frequency response of both the upper and lower erector spinae can be modelled with a cut-off frequency between 2 and 2.5 Hz and that these values are not significantly modulated by changes in muscle length or tension. 1998 Elsevier Science Ltd. All rights reserved. Keywords: Electromyogram; Frequency response; Impulse; Lumbar spine
1. Introduction During contraction in vivo, the electrical activity within muscle necessarily precedes force output irrespective of the nature of the contraction condition. Several models have utilized the electromyogram (EMG) together with the modulating effects of instantaneous length and velocity of contraction to predict muscle forces (in the ankle, Hof and VandenBerg [1]; in the low back, McGill and Norman [2]), and with the direct measurement of tendon force in the cat [3]. In our work on the spine, estimated muscle forces, predicted in part from low pass filtered EMG, closely conform to measured torque during slow controlled movements, yet paradoxically, while under rapid ballistic type loading conditions, predicted muscle forces have been observed to
* Corresponding author. Tel.: + 1-519-885-1211; Ext. 6761; Fax: + 1-519-746-6776; E-mail: [email protected] 1050-6411/98/$19.00 1998 Elsevier Science Ltd. All rights reserved. PII: S 1 0 5 0 - 6 4 1 1 ( 9 8 ) 0 0 0 0 9 - 1
precede the electrical activity (unpublished results). These observations motivated this study to more closely examine the mechanical coupling of low pass filtered EMG to force output during impulsive contraction conditions. Although there exists a variety of ways to process EMG, a biologically justifiable process remains attractive—namely using a single low pass filter because the cut-off frequency of the Butterworth filter can be tuned to match the frequency response of the muscle [4]. The intention of a single pass of the filter is to produce a phase lag to mimic the electromechanical delay seen in vivo. The critical issue is to define the frequency response of different muscle contractions throughout a range of instantaneous lengths and velocities to facilitate prediction of force— specifically during ballistic contractions. The frequency response of some muscles has been well documented under isometric conditions. MilnerBrown et al. [5] found the mean natural frequency of 12 motor units within the first dorsal interosseous muscle
228
L.C. Brereton, S.M. McGill / Journal of Electromyography and Kinesiology 8 (1998) 227–232
to equal 2.4 Hz (range 1.4–5 Hz) under voluntary contraction. They concluded that a critically damped, linear, second order system with a cut-off of 3 Hz optimally satisfied the transfer function of this muscle. Similar findings have been reported for the triceps [6] and soleus [7] muscles documenting a cut-off frequency of 1.7–1.9 and 2 Hz respectively. Buchthal and Schmalbruch [8] found the contraction response time of muscle, measured from the onset to peak of contraction, was 64 ms for histologically typed fast twitch fibres and 120 ms for slow twitch fibres, translating into a cut-off frequency of 3.9 and 2.1 Hz respectively. While it is more challenging to quantify muscle frequency response under increasingly dynamic conditions, work performed by Norman et al. [3] provided convincing evidence that rectified low pass filtered EMG signals are strongly correlated with directly measured tendon force during dynamic movement, even without accounting for muscle length and contraction velocity. Their results support the choice of using a low pass filtered EMG to mimic force output. The purpose of this study was to characterize the frequency response of the erector spinae musculature to enable estimates of force from the electromyogram particularly during ballistic contraction. The mechanical coupling of EMG to force is sensitive to both chemical and mechanical muscle properties, together with inertial properties of the affected body segments. Furthermore, the elastic stretch and stiffness of muscle are modulated by muscle length and tension respectively; therefore, the effect of muscle length, contraction effort, and muscle type (thoracic vs lumbar erector spinae) on the frequency response were examined. The specific goal was to obtain a cut-off frequency to characterize the full range of dynamic response of these muscles.
2. Methods 2.1. Data collection Five male subjects (age: mean 26.6 yr, range 23–31, height: mean 180 cm, range 175–185, weight: mean 71.4 kg, range 67–75) were recruited to perform rapid, isometric contractions of the back musculature, with the back placed against a load cell while seated in a restraining jig. EMG of spine extensor and abdominal wall musculature (to facilitate analysis of co-contraction) were measured together with force output from the back against the load cell. All subjects were in good health and screened for disabling low back pain. The experimental procedure was approved by the Office of Human Research, University of Waterloo. Eight pairs of disposable surface electrodes (Ag– AgCl; Graphic controls, Gananoque ON) were applied to skin that had been shaved, cleaned with a 1:1 water–
ethanol solution and lightly abraded until pink in appearance. Electrodes were placed with a centre to centre distance of 3 cm bi-laterally over the belly of the following muscles: rectus abdominis, 3 cm lateral to the umbilicus; external oblique, approximately 15 cm lateral to the umbilicus; thoracic erector spinae, 5 cm lateral to the T9 spinous process; and the lumbar erector spinae, 3 cm lateral to the L3 spinous process. Raw EMG signals were pre-filtered, producing a band width of 10–500 Hz, and amplified with a differential amp (CMRR > 90 db at 60 Hz, input impedance > 10 M⍀ above 1 Hz) in accordance with ISEK guidelines [9]. All EMG signals were A/D converted at 1024 Hz on line and processed digitally. Angular displacement of the lumbar spine about the flexion–extension axis was measured to control for muscle length using the 3 SPACE ISOTRAK (Polhemus Navigation Sciences, McDonnell Douglas Electronics) and A/D converted at a frequency of 20.5 Hz for the duration of the trials. Maximum lumbar range of motion (ROM) was measured for each subject between upright standing and forward flexion. To isolate spine motion and fixate the pelvis, a seated restraining jig was used [10] (Fig. 1). Subjects were seated in a semi-kneeling posture with a hip angle that approximated elastic equilibrium about the hip joint [11]. The pelvis was fixated with a rigid clamp and secured by an adjustable belt about the hips. A uniaxial load cell was mounted behind the restraining jig and was adjusted to make contact between the scapulae above the thoracic erector spinae electrodes to measure the trunk forces upon spine extension. Prior to data analysis, all EMG signals were normalized to the electrical activity measured during maximal effort isometric contractions (100% MVC). MVC’s were performed according to the protocol outlined in McGill [12]. Briefly, abdominal MVC’s were measured in a bent-knee sit-up posture on top of a padded test bench with the feet restrained by a strap. Hands were placed behind the head and the trunk formed an angle with the horizontal of approximately 30°. An assistant provided a matching resistance to the shoulders during a maximum sit-up effort. Erector spinae MVC’s were measured while the subject leaned over the side of the test bench, aligning the anterior superior iliac spines with the edge of the bench, maintaining a neutral state of lordosis. The feet were restrained as a matching resistance was provided on the upper back during a maximum extension effort. Bias (or average value) of each channel was determined from the raw EMG. Maximal EMG signals during the MVC trials were determined after full wave rectification and low pass filtering (cut-off of 2.5 Hz) of the raw EMG signal.
L.C. Brereton, S.M. McGill / Journal of Electromyography and Kinesiology 8 (1998) 227–232
229
100% ROM and maximum force achieved in each position and organized into bins based on the normalized spine posture and force amplitude. These bins corresponded to 0–12°, 12–30°, 30–55° for posture and ⬍ 55, 55–75, > 75% for effort.
2.3. Data analysis
Fig. 1. Subject in upright position in restraint jig with hips and spine in approximate elastic equilibrium. Forces were measured with an LVDT (arrow).
2.2. Tasks Subjects were instructed to perform rapid isometric contractions of the back extensor musculature while minimizing abdominal co-contraction. Indeed some subjects needed to learn this motor skill. EMG and force were recorded for 4 s; although the full cycle of contraction and relaxation as measured from the force output, typically lasted less than 1 s. Each task was repeated twice in three positions (modulating muscle length) and at three efforts (modulating muscle tension), for a total of 18 trials per subject. By rotating the pelvis between trials, three lordotic postures, or three muscle lengths of the erector spinae were achieved. These three positions, were monitored during each trial by the 3-SPACE, and displayed to the subject for feedback. Each position corresponded to approximately 0, 20 and 40° of lumbar spine flexion. These angles covered the greatest possible ROM while being restrained in the testing apparatus. For each posture, three levels of effort were performed: maximum, moderate and light. Post-hoc, spine position and effort of each trial were respectively normalized to
Raw EMG signals were rectified and filtered repeatedly with varying low pass filter cut-off frequencies from 1 to 3 Hz at 0.5 Hz increments to mimic the frequency response of the in vivo muscle– torso system. At each cut-off frequency, an average EMG signal was derived from the normalized right and left filtered EMG for both the upper and lower erector spinae muscles. A classic challenge to the biomechanist remains how to compare two time histories, such as force and low pass filtered EMG. The cut-off frequency that best aligned extensor EMG with force was selected in two ways. Cut-off frequencies were selected qualitatively by comparing graphs of force output with averaged EMG for the lower and upper erector spinae muscles filtered at each of the five cut-off frequencies. The priority in selecting a cutoff frequency was to align the rising slope of the EMG with that of the force, negating both the alignment of the downward slope of EMG and force due to the presence of possible co-contractions, and phase shifts between peak EMG and peak force. Cut-off frequencies were also determined by cross correlating force with each EMG channel (right, left and weighted average for upper and lower erector spinae muscles) over a 2 s window commencing half a second before the onset of force. Then, cross correlation coefficients were computed while the force was phase shifted ± 100 ms at 1 ms intervals in order to determine the full transfer function. The maximum cross correlation coefficients and corresponding phase shift values were recorded for each channel of EMG at each of five frequencies. The maximum cross correlation coefficient value of the resultant five values from the average EMG channel for both the upper and lower erector spinaes were selected to represent the optimal cut-off frequency for that trial. All five subjects performed maximal efforts at the three positions. The effect of contraction effort was only examined in two subjects since the lower level efforts of three subjects were so low they were within the noise level of the instrumentation. An ANOVA with repeated measures on subject, effort, position, muscle type (thoracic vs lumbar erector spinae) and method (crosscorrelation vs qualitative) was performed to assess the effect of varying conditions on the selection of optimal cut-off frequency. Statistical significance could not be assumed for the effect of contraction effort as only two subjects completed all three levels of effort.
230
L.C. Brereton, S.M. McGill / Journal of Electromyography and Kinesiology 8 (1998) 227–232
Fig. 2. Histogram of cut-off frequencies averaged across muscle (lower and upper erector spinae) and method (cross-correlation and qualitative measures) for maximum efforts (N = 112).
3. Results Seventy percent of optimal cut-off frequencies from all trials and subjects resided between 2 and 2.5 Hz (Fig. 2). A typical trial processed at each of the five cut-off frequencies indicated that the optimal cut-off frequency lay near 2–2.5 Hz (Fig. 3). The filtered EMG data were presented as scaled to zero and peak force for a cut-off of 2 Hz to facilitate qualitative comparison between the force and EMG curves. The peak activation (% MVC) at which subjects were contracting varied, but generally was around 45% for the maximum trial, 30% for the moderate effort trials and 15% for the light trials although some efforts were very low as to not allow analysis. EMG data is presented as an average of the bilateral signal for upper and lower erector spinae muscles
Fig. 4. Phasic response of the right, left and average EMG signal from the lower (a) and upper (b) erector spinae muscles for one maximum effort in the upright position.
(Fig. 4). Although amplitude differences existed between right and left EMG channels, differences in the phasic relationship between sides were minimal; therefore, the average EMG was representative of the phasic response of the muscle. Results from the ANOVA, revealed there were no significant effects of muscle length (position), or muscle type (lower vs. upper erector spinae) on the optimal cutoff frequency (␣ = 0.05). It also appeared that there were no effects of effort, although this could not be statistically determined. Means and standard deviations of selected cut-off frequencies by muscle, effort and position, collapsed across subject and method highlight the similarities across conditions (Table 1). Table 1 Mean and standard deviation of cut-off frequency by positions, efforts and muscle type, collapsed across subjects and method Mean
Fig. 3. Relationship between force output and cut-off frequency of maximum effort trial in upright position for lower erector spinae (a) and upper erector spinae (b) of a single subject. Force and EMG peaks scaled to align at 2 Hz. EMG scaled to zero percent force at beginning of trial.
Position 0°–12° 12°–30° 30°–50° Effort Maximum Moderate Light Muscle Lower ES Upper ES
SD
2.21 2.18 2.19
0.57 0.36 0.49
2.03 2.16 2.16
0.50 0.59 0.53
2.13 2.25
0.52 0.45
L.C. Brereton, S.M. McGill / Journal of Electromyography and Kinesiology 8 (1998) 227–232
Although the statistical results did reveal a significant difference (P = 0.04) between the two methods used to determine optimal cut-off frequency (cross-correlation and qualitative measures), closer inspection revealed that the difference between the cut-off frequency selected by each method fell within 0.5 Hz for most trials. Abdominal EMG was collected to ensure that co-contraction was not affecting the biological coupling between extensor EMG and force. In some cases, subjects were able to silence the activity of the abdominal musculature during back extensor contraction to less than 5% MVC. The majority of trials showed co-contraction between 5–25% MVC, in phase with back extensor contraction (Fig. 5).
4. Discussion It would appear from these results that the cut-off frequency for erector spinae musculature lies between 2 and
Fig. 5. Phasic relationship between erectors and abdominals during upright maximal contraction. Top graph highlights abdominals contracting in phase with erectors. Bottom graph illustrates no contraction of abdominals during extension contraction.
231
2.5 Hz for isometric, impulsive contractions. This cut-off frequency appears to be independent of muscle length, contraction effort, or anatomical regions within the erector spinae musculature. This is an important finding for understanding the dynamics of lower back function, and encouraging for spine modelers and those interested in using EMG to estimate force across different muscle lengths. The decision to base the selection of cut-off frequency on the average of the normalized bi-lateral EMG may be viewed as a limitation; however, post-hoc analysis revealed only subtle differences in EMG collected from the right and left side of each muscle. In the majority of cases, the selected cut-off frequency would have been identical whether selected using the left, right or averaged EMG signal. Although subjects were instructed to relax all muscles prior to and following the impulsive contraction, results indicated that it was necessary for some subjects to lightly contract their back musculature in order to maintain a given posture. This occurred primarily during an upright posture, but was also observed to a smaller extent during moderately slouched, and fully slouched positions. As this study was primarily concerned with the phasic response between EMG and force during ballistic contraction, any EMG stemming from muscle tension required to maintain a given posture prior to the onset of the impulsive contraction was normalized to zero. Although this scaling process would significantly affect the amplitude relationship between force and EMG, it should not alter the phasic relationship between the two. An attempt was made to characterize the frequency response of the extensor muscles from measurements of the intact torso system. Naturally occurring concurrent abdominal contraction could have potentially altered the mechanical coupling of the system, thereby causing significant error in the selection of optimal cut-off frequency. Qualitative analysis revealed that abdominal cocontraction was ballistic and in phase with back extensors. Although this would have altered the amplitude of the extensor EMG, the frequency response of the extensor muscles should not have been manipulated. Also, it appeared that generating a maximal rapid, isometric back extension without simultaneous abdominal co-contraction while being restrained in a jig was a significant motor challenge for all of the subjects. This may help to explain the subjects’ inability to elicit no greater than 50% MVC during maximal effort trials. Statistical analysis did reveal a significant difference in cut-off frequencies selected by cross-correlation and qualitative measures. However, when taking into consideration the variability inherent in biological systems, the margin of error between cut-off frequencies selected by the two methods should not be erroneously interpreted as biologically significant. Rather, variability of 0.5 Hz in the selection of cut-off frequencies was con-
232
L.C. Brereton, S.M. McGill / Journal of Electromyography and Kinesiology 8 (1998) 227–232
sidered acceptable. It should be noted that the window over which cross-correlations were performed began at approximately 0.5 s before the onset of force and terminated 2 s later, regardless of the duration of contraction which was invariably shorter. This method was selected to simplify data processing; however, it is recognized that any extraneous movement without concurrent force output, occurring during this 2 s window could have affected the cross-correlation value, and consequently the selected cut-off frequency. Frequently, the correlation values between two or three cut-off frequencies were within 1% of each other. Due to factors that may affect cross correlation values, it would be inappropriate to assume that the selected cut-off frequency for each trial was an absolute. We attempted to substantiate our results by comparing cross correlations measures to qualitative measures. Although this second methodology may not have been ideal, the qualitative results supported the cross correlation results within an acceptable margin of error. Our findings compare well with those presented Milner-Brown et al. [5], Buchthal and Schmalbruch [8], and Coggshall and Bekey [6] who reported cut-off frequencies for isometric contractions in the range of 1.7– 3 Hz. Our results also compared favourably with the work of Norman et al. [3] as muscle length was not a factor when correlating force to EMG. We encourage researchers who utilize RMS processing procedures for smoothing EMG to compare their results with a second order system to determine if RMS smoothing generates physiologically based results. In conclusion, it appears that the cut-off frequency response of the low back musculature during rapid isometric contractions is between 2 and 2.5 Hz indicating that the dynamic characteristics of the muscle can be captured by appropriately treated EMG.
References [1] Hof AL, VandenBerg JW. EMG to force processing I: an electrical analogue of the hill muscle model. J Biomech 1971;14(11):747–58. [2] McGill SM, Norman RW. Partitioning of the L4-L5 dynamic moment into disc, ligamentous and muscular components during lifting. Spine 1988;11(7):666–77. [3] Norman R, Gregor R, Dowling J. The prediction of cat tendon force form EMG in dynamic muscular contractions. In: Proceedings from the Fifth Biennial Conference Symposium, Canadian Society of Biomechanics, 1988:120–1. [4] Winter DA. Biomechanics and motor control of human movement, 2nd ed, ch. 2. New York: John Wiley and Sons, 1990. [5] Milner-Brown HS, Stein RB, Yemm R. The contractile properties of human motor units during voluntary isometric contractions. J Physiol 1973;228:285–306. [6] Coggshall JC, Bekey GA. EMG-force dynamics in human skeletal muscle. Med Biol Eng 1970;8:265–70. [7] Bawa P, Stein RB. Frequency response of human soleus muscle. J Neurophysiol 1976;39(4):788–93. [8] Buchthal F, Schmalbruch H. Contraction times and fibre types in intact human muscle. Acta Physiol Scand 1970;79:435–52. [9] Standards for reporting EMG data. Journal of Electromyography and Kinesiology 1996;6(1):III–IV. [10] Sutarno CG, McGill SM. Isovelocity investigation of the lengthening behaviour of the erector spinae muscle. Eur J Appl Physiol 1995;70:146–53. [11] Keegan JJ. Alterations of the lumbar curve related to posture and standing. J Bone Joint Surg 1953;35A:589. [12] McGill SM. Electromyographic activity of the abdominal and low back musculature during the generation of isometric and dynamic axial trunk torque: implications for lumbar mechanics. J Orthop Res 1991;9:91–103. Lisa C. Brereton received a B.Sc. and B. PHE degree (1996) from Queen’s University, Kingston, Ontario. She is presently a Masters student at the University of Waterloo, Waterloo, Ontario. Ms Brereton is the current student executive representative for the Canadian Society of Biomechanics. Her research focuses on how the spine becomes injured.
Acknowledgements Financial support for this research by the Natural Sciences and Engineering Research Council of Canada is gratefully acknowledged. Special thanks is extended to Vanessa Yingling for her contribution to the initial stages of this research.
Stuart M. McGill received a Ph.D. degree from the University of Waterloo in 1986. He is a professor of spine biomechanics in the Department of Kinesiology at the University of Waterloo where his research interests focus on how the spine works and becomes injured, and on developing better strategies for prevention and rehabilitation of low back pain. He is married, has two children and enjoys fishing.
Frequency response of spine extensors during rapid isometric contractions: effects of muscle length and tension Lisa C. Brereton, Stuart M. McGill
*
Occupational Biomechanics and Safety Laboratories, Department of Kinesiology, Faculty of Applied Health Sciences, University of Waterloo, Waterloo, Ontario, Canada
Abstract During muscle contraction, electrical activity necessarily precedes force output, yet models that utilize processed electromyograms sometimes predict force as preceding EMG under rapid ballistic loading conditions. The purpose of this study was to define the frequency response transfer function of the upper and lower erector spinae musculature, at different lengths and tensions, using rectified, low pass filtered EMG. This would enable accurate estimates of force from the processed electromyogram, specifically during impulsive contractions. Abdominal and erector spinae EMG were measured in synchrony with impulsive low back moments in five men. EMG signals were rectified and low pass filtered repeatedly with cut-off frequencies from 1 to 3 Hz at 0.5 Hz increments in order to quantify the frequency response. It was found that EMG signals processed through a simple, Butterworth low pass filter could not produce the measured force output without an additional time shift. These shifts were quantified by cross-correlating EMG and force with increments of 1 ms. In order to define the transfer function of EMG to force, optimal cut-off frequencies were selected two ways: quantitatively by searching for maximum cross correlations coefficients, and qualitatively. Results indicated that the frequency response of both the upper and lower erector spinae can be modelled with a cut-off frequency between 2 and 2.5 Hz and that these values are not significantly modulated by changes in muscle length or tension. 1998 Elsevier Science Ltd. All rights reserved. Keywords: Electromyogram; Frequency response; Impulse; Lumbar spine
1. Introduction During contraction in vivo, the electrical activity within muscle necessarily precedes force output irrespective of the nature of the contraction condition. Several models have utilized the electromyogram (EMG) together with the modulating effects of instantaneous length and velocity of contraction to predict muscle forces (in the ankle, Hof and VandenBerg [1]; in the low back, McGill and Norman [2]), and with the direct measurement of tendon force in the cat [3]. In our work on the spine, estimated muscle forces, predicted in part from low pass filtered EMG, closely conform to measured torque during slow controlled movements, yet paradoxically, while under rapid ballistic type loading conditions, predicted muscle forces have been observed to
* Corresponding author. Tel.: + 1-519-885-1211; Ext. 6761; Fax: + 1-519-746-6776; E-mail: [email protected] 1050-6411/98/$19.00 1998 Elsevier Science Ltd. All rights reserved. PII: S 1 0 5 0 - 6 4 1 1 ( 9 8 ) 0 0 0 0 9 - 1
precede the electrical activity (unpublished results). These observations motivated this study to more closely examine the mechanical coupling of low pass filtered EMG to force output during impulsive contraction conditions. Although there exists a variety of ways to process EMG, a biologically justifiable process remains attractive—namely using a single low pass filter because the cut-off frequency of the Butterworth filter can be tuned to match the frequency response of the muscle [4]. The intention of a single pass of the filter is to produce a phase lag to mimic the electromechanical delay seen in vivo. The critical issue is to define the frequency response of different muscle contractions throughout a range of instantaneous lengths and velocities to facilitate prediction of force— specifically during ballistic contractions. The frequency response of some muscles has been well documented under isometric conditions. MilnerBrown et al. [5] found the mean natural frequency of 12 motor units within the first dorsal interosseous muscle
228
L.C. Brereton, S.M. McGill / Journal of Electromyography and Kinesiology 8 (1998) 227–232
to equal 2.4 Hz (range 1.4–5 Hz) under voluntary contraction. They concluded that a critically damped, linear, second order system with a cut-off of 3 Hz optimally satisfied the transfer function of this muscle. Similar findings have been reported for the triceps [6] and soleus [7] muscles documenting a cut-off frequency of 1.7–1.9 and 2 Hz respectively. Buchthal and Schmalbruch [8] found the contraction response time of muscle, measured from the onset to peak of contraction, was 64 ms for histologically typed fast twitch fibres and 120 ms for slow twitch fibres, translating into a cut-off frequency of 3.9 and 2.1 Hz respectively. While it is more challenging to quantify muscle frequency response under increasingly dynamic conditions, work performed by Norman et al. [3] provided convincing evidence that rectified low pass filtered EMG signals are strongly correlated with directly measured tendon force during dynamic movement, even without accounting for muscle length and contraction velocity. Their results support the choice of using a low pass filtered EMG to mimic force output. The purpose of this study was to characterize the frequency response of the erector spinae musculature to enable estimates of force from the electromyogram particularly during ballistic contraction. The mechanical coupling of EMG to force is sensitive to both chemical and mechanical muscle properties, together with inertial properties of the affected body segments. Furthermore, the elastic stretch and stiffness of muscle are modulated by muscle length and tension respectively; therefore, the effect of muscle length, contraction effort, and muscle type (thoracic vs lumbar erector spinae) on the frequency response were examined. The specific goal was to obtain a cut-off frequency to characterize the full range of dynamic response of these muscles.
2. Methods 2.1. Data collection Five male subjects (age: mean 26.6 yr, range 23–31, height: mean 180 cm, range 175–185, weight: mean 71.4 kg, range 67–75) were recruited to perform rapid, isometric contractions of the back musculature, with the back placed against a load cell while seated in a restraining jig. EMG of spine extensor and abdominal wall musculature (to facilitate analysis of co-contraction) were measured together with force output from the back against the load cell. All subjects were in good health and screened for disabling low back pain. The experimental procedure was approved by the Office of Human Research, University of Waterloo. Eight pairs of disposable surface electrodes (Ag– AgCl; Graphic controls, Gananoque ON) were applied to skin that had been shaved, cleaned with a 1:1 water–
ethanol solution and lightly abraded until pink in appearance. Electrodes were placed with a centre to centre distance of 3 cm bi-laterally over the belly of the following muscles: rectus abdominis, 3 cm lateral to the umbilicus; external oblique, approximately 15 cm lateral to the umbilicus; thoracic erector spinae, 5 cm lateral to the T9 spinous process; and the lumbar erector spinae, 3 cm lateral to the L3 spinous process. Raw EMG signals were pre-filtered, producing a band width of 10–500 Hz, and amplified with a differential amp (CMRR > 90 db at 60 Hz, input impedance > 10 M⍀ above 1 Hz) in accordance with ISEK guidelines [9]. All EMG signals were A/D converted at 1024 Hz on line and processed digitally. Angular displacement of the lumbar spine about the flexion–extension axis was measured to control for muscle length using the 3 SPACE ISOTRAK (Polhemus Navigation Sciences, McDonnell Douglas Electronics) and A/D converted at a frequency of 20.5 Hz for the duration of the trials. Maximum lumbar range of motion (ROM) was measured for each subject between upright standing and forward flexion. To isolate spine motion and fixate the pelvis, a seated restraining jig was used [10] (Fig. 1). Subjects were seated in a semi-kneeling posture with a hip angle that approximated elastic equilibrium about the hip joint [11]. The pelvis was fixated with a rigid clamp and secured by an adjustable belt about the hips. A uniaxial load cell was mounted behind the restraining jig and was adjusted to make contact between the scapulae above the thoracic erector spinae electrodes to measure the trunk forces upon spine extension. Prior to data analysis, all EMG signals were normalized to the electrical activity measured during maximal effort isometric contractions (100% MVC). MVC’s were performed according to the protocol outlined in McGill [12]. Briefly, abdominal MVC’s were measured in a bent-knee sit-up posture on top of a padded test bench with the feet restrained by a strap. Hands were placed behind the head and the trunk formed an angle with the horizontal of approximately 30°. An assistant provided a matching resistance to the shoulders during a maximum sit-up effort. Erector spinae MVC’s were measured while the subject leaned over the side of the test bench, aligning the anterior superior iliac spines with the edge of the bench, maintaining a neutral state of lordosis. The feet were restrained as a matching resistance was provided on the upper back during a maximum extension effort. Bias (or average value) of each channel was determined from the raw EMG. Maximal EMG signals during the MVC trials were determined after full wave rectification and low pass filtering (cut-off of 2.5 Hz) of the raw EMG signal.
L.C. Brereton, S.M. McGill / Journal of Electromyography and Kinesiology 8 (1998) 227–232
229
100% ROM and maximum force achieved in each position and organized into bins based on the normalized spine posture and force amplitude. These bins corresponded to 0–12°, 12–30°, 30–55° for posture and ⬍ 55, 55–75, > 75% for effort.
2.3. Data analysis
Fig. 1. Subject in upright position in restraint jig with hips and spine in approximate elastic equilibrium. Forces were measured with an LVDT (arrow).
2.2. Tasks Subjects were instructed to perform rapid isometric contractions of the back extensor musculature while minimizing abdominal co-contraction. Indeed some subjects needed to learn this motor skill. EMG and force were recorded for 4 s; although the full cycle of contraction and relaxation as measured from the force output, typically lasted less than 1 s. Each task was repeated twice in three positions (modulating muscle length) and at three efforts (modulating muscle tension), for a total of 18 trials per subject. By rotating the pelvis between trials, three lordotic postures, or three muscle lengths of the erector spinae were achieved. These three positions, were monitored during each trial by the 3-SPACE, and displayed to the subject for feedback. Each position corresponded to approximately 0, 20 and 40° of lumbar spine flexion. These angles covered the greatest possible ROM while being restrained in the testing apparatus. For each posture, three levels of effort were performed: maximum, moderate and light. Post-hoc, spine position and effort of each trial were respectively normalized to
Raw EMG signals were rectified and filtered repeatedly with varying low pass filter cut-off frequencies from 1 to 3 Hz at 0.5 Hz increments to mimic the frequency response of the in vivo muscle– torso system. At each cut-off frequency, an average EMG signal was derived from the normalized right and left filtered EMG for both the upper and lower erector spinae muscles. A classic challenge to the biomechanist remains how to compare two time histories, such as force and low pass filtered EMG. The cut-off frequency that best aligned extensor EMG with force was selected in two ways. Cut-off frequencies were selected qualitatively by comparing graphs of force output with averaged EMG for the lower and upper erector spinae muscles filtered at each of the five cut-off frequencies. The priority in selecting a cutoff frequency was to align the rising slope of the EMG with that of the force, negating both the alignment of the downward slope of EMG and force due to the presence of possible co-contractions, and phase shifts between peak EMG and peak force. Cut-off frequencies were also determined by cross correlating force with each EMG channel (right, left and weighted average for upper and lower erector spinae muscles) over a 2 s window commencing half a second before the onset of force. Then, cross correlation coefficients were computed while the force was phase shifted ± 100 ms at 1 ms intervals in order to determine the full transfer function. The maximum cross correlation coefficients and corresponding phase shift values were recorded for each channel of EMG at each of five frequencies. The maximum cross correlation coefficient value of the resultant five values from the average EMG channel for both the upper and lower erector spinaes were selected to represent the optimal cut-off frequency for that trial. All five subjects performed maximal efforts at the three positions. The effect of contraction effort was only examined in two subjects since the lower level efforts of three subjects were so low they were within the noise level of the instrumentation. An ANOVA with repeated measures on subject, effort, position, muscle type (thoracic vs lumbar erector spinae) and method (crosscorrelation vs qualitative) was performed to assess the effect of varying conditions on the selection of optimal cut-off frequency. Statistical significance could not be assumed for the effect of contraction effort as only two subjects completed all three levels of effort.
230
L.C. Brereton, S.M. McGill / Journal of Electromyography and Kinesiology 8 (1998) 227–232
Fig. 2. Histogram of cut-off frequencies averaged across muscle (lower and upper erector spinae) and method (cross-correlation and qualitative measures) for maximum efforts (N = 112).
3. Results Seventy percent of optimal cut-off frequencies from all trials and subjects resided between 2 and 2.5 Hz (Fig. 2). A typical trial processed at each of the five cut-off frequencies indicated that the optimal cut-off frequency lay near 2–2.5 Hz (Fig. 3). The filtered EMG data were presented as scaled to zero and peak force for a cut-off of 2 Hz to facilitate qualitative comparison between the force and EMG curves. The peak activation (% MVC) at which subjects were contracting varied, but generally was around 45% for the maximum trial, 30% for the moderate effort trials and 15% for the light trials although some efforts were very low as to not allow analysis. EMG data is presented as an average of the bilateral signal for upper and lower erector spinae muscles
Fig. 4. Phasic response of the right, left and average EMG signal from the lower (a) and upper (b) erector spinae muscles for one maximum effort in the upright position.
(Fig. 4). Although amplitude differences existed between right and left EMG channels, differences in the phasic relationship between sides were minimal; therefore, the average EMG was representative of the phasic response of the muscle. Results from the ANOVA, revealed there were no significant effects of muscle length (position), or muscle type (lower vs. upper erector spinae) on the optimal cutoff frequency (␣ = 0.05). It also appeared that there were no effects of effort, although this could not be statistically determined. Means and standard deviations of selected cut-off frequencies by muscle, effort and position, collapsed across subject and method highlight the similarities across conditions (Table 1). Table 1 Mean and standard deviation of cut-off frequency by positions, efforts and muscle type, collapsed across subjects and method Mean
Fig. 3. Relationship between force output and cut-off frequency of maximum effort trial in upright position for lower erector spinae (a) and upper erector spinae (b) of a single subject. Force and EMG peaks scaled to align at 2 Hz. EMG scaled to zero percent force at beginning of trial.
Position 0°–12° 12°–30° 30°–50° Effort Maximum Moderate Light Muscle Lower ES Upper ES
SD
2.21 2.18 2.19
0.57 0.36 0.49
2.03 2.16 2.16
0.50 0.59 0.53
2.13 2.25
0.52 0.45
L.C. Brereton, S.M. McGill / Journal of Electromyography and Kinesiology 8 (1998) 227–232
Although the statistical results did reveal a significant difference (P = 0.04) between the two methods used to determine optimal cut-off frequency (cross-correlation and qualitative measures), closer inspection revealed that the difference between the cut-off frequency selected by each method fell within 0.5 Hz for most trials. Abdominal EMG was collected to ensure that co-contraction was not affecting the biological coupling between extensor EMG and force. In some cases, subjects were able to silence the activity of the abdominal musculature during back extensor contraction to less than 5% MVC. The majority of trials showed co-contraction between 5–25% MVC, in phase with back extensor contraction (Fig. 5).
4. Discussion It would appear from these results that the cut-off frequency for erector spinae musculature lies between 2 and
Fig. 5. Phasic relationship between erectors and abdominals during upright maximal contraction. Top graph highlights abdominals contracting in phase with erectors. Bottom graph illustrates no contraction of abdominals during extension contraction.
231
2.5 Hz for isometric, impulsive contractions. This cut-off frequency appears to be independent of muscle length, contraction effort, or anatomical regions within the erector spinae musculature. This is an important finding for understanding the dynamics of lower back function, and encouraging for spine modelers and those interested in using EMG to estimate force across different muscle lengths. The decision to base the selection of cut-off frequency on the average of the normalized bi-lateral EMG may be viewed as a limitation; however, post-hoc analysis revealed only subtle differences in EMG collected from the right and left side of each muscle. In the majority of cases, the selected cut-off frequency would have been identical whether selected using the left, right or averaged EMG signal. Although subjects were instructed to relax all muscles prior to and following the impulsive contraction, results indicated that it was necessary for some subjects to lightly contract their back musculature in order to maintain a given posture. This occurred primarily during an upright posture, but was also observed to a smaller extent during moderately slouched, and fully slouched positions. As this study was primarily concerned with the phasic response between EMG and force during ballistic contraction, any EMG stemming from muscle tension required to maintain a given posture prior to the onset of the impulsive contraction was normalized to zero. Although this scaling process would significantly affect the amplitude relationship between force and EMG, it should not alter the phasic relationship between the two. An attempt was made to characterize the frequency response of the extensor muscles from measurements of the intact torso system. Naturally occurring concurrent abdominal contraction could have potentially altered the mechanical coupling of the system, thereby causing significant error in the selection of optimal cut-off frequency. Qualitative analysis revealed that abdominal cocontraction was ballistic and in phase with back extensors. Although this would have altered the amplitude of the extensor EMG, the frequency response of the extensor muscles should not have been manipulated. Also, it appeared that generating a maximal rapid, isometric back extension without simultaneous abdominal co-contraction while being restrained in a jig was a significant motor challenge for all of the subjects. This may help to explain the subjects’ inability to elicit no greater than 50% MVC during maximal effort trials. Statistical analysis did reveal a significant difference in cut-off frequencies selected by cross-correlation and qualitative measures. However, when taking into consideration the variability inherent in biological systems, the margin of error between cut-off frequencies selected by the two methods should not be erroneously interpreted as biologically significant. Rather, variability of 0.5 Hz in the selection of cut-off frequencies was con-
232
L.C. Brereton, S.M. McGill / Journal of Electromyography and Kinesiology 8 (1998) 227–232
sidered acceptable. It should be noted that the window over which cross-correlations were performed began at approximately 0.5 s before the onset of force and terminated 2 s later, regardless of the duration of contraction which was invariably shorter. This method was selected to simplify data processing; however, it is recognized that any extraneous movement without concurrent force output, occurring during this 2 s window could have affected the cross-correlation value, and consequently the selected cut-off frequency. Frequently, the correlation values between two or three cut-off frequencies were within 1% of each other. Due to factors that may affect cross correlation values, it would be inappropriate to assume that the selected cut-off frequency for each trial was an absolute. We attempted to substantiate our results by comparing cross correlations measures to qualitative measures. Although this second methodology may not have been ideal, the qualitative results supported the cross correlation results within an acceptable margin of error. Our findings compare well with those presented Milner-Brown et al. [5], Buchthal and Schmalbruch [8], and Coggshall and Bekey [6] who reported cut-off frequencies for isometric contractions in the range of 1.7– 3 Hz. Our results also compared favourably with the work of Norman et al. [3] as muscle length was not a factor when correlating force to EMG. We encourage researchers who utilize RMS processing procedures for smoothing EMG to compare their results with a second order system to determine if RMS smoothing generates physiologically based results. In conclusion, it appears that the cut-off frequency response of the low back musculature during rapid isometric contractions is between 2 and 2.5 Hz indicating that the dynamic characteristics of the muscle can be captured by appropriately treated EMG.
References [1] Hof AL, VandenBerg JW. EMG to force processing I: an electrical analogue of the hill muscle model. J Biomech 1971;14(11):747–58. [2] McGill SM, Norman RW. Partitioning of the L4-L5 dynamic moment into disc, ligamentous and muscular components during lifting. Spine 1988;11(7):666–77. [3] Norman R, Gregor R, Dowling J. The prediction of cat tendon force form EMG in dynamic muscular contractions. In: Proceedings from the Fifth Biennial Conference Symposium, Canadian Society of Biomechanics, 1988:120–1. [4] Winter DA. Biomechanics and motor control of human movement, 2nd ed, ch. 2. New York: John Wiley and Sons, 1990. [5] Milner-Brown HS, Stein RB, Yemm R. The contractile properties of human motor units during voluntary isometric contractions. J Physiol 1973;228:285–306. [6] Coggshall JC, Bekey GA. EMG-force dynamics in human skeletal muscle. Med Biol Eng 1970;8:265–70. [7] Bawa P, Stein RB. Frequency response of human soleus muscle. J Neurophysiol 1976;39(4):788–93. [8] Buchthal F, Schmalbruch H. Contraction times and fibre types in intact human muscle. Acta Physiol Scand 1970;79:435–52. [9] Standards for reporting EMG data. Journal of Electromyography and Kinesiology 1996;6(1):III–IV. [10] Sutarno CG, McGill SM. Isovelocity investigation of the lengthening behaviour of the erector spinae muscle. Eur J Appl Physiol 1995;70:146–53. [11] Keegan JJ. Alterations of the lumbar curve related to posture and standing. J Bone Joint Surg 1953;35A:589. [12] McGill SM. Electromyographic activity of the abdominal and low back musculature during the generation of isometric and dynamic axial trunk torque: implications for lumbar mechanics. J Orthop Res 1991;9:91–103. Lisa C. Brereton received a B.Sc. and B. PHE degree (1996) from Queen’s University, Kingston, Ontario. She is presently a Masters student at the University of Waterloo, Waterloo, Ontario. Ms Brereton is the current student executive representative for the Canadian Society of Biomechanics. Her research focuses on how the spine becomes injured.
Acknowledgements Financial support for this research by the Natural Sciences and Engineering Research Council of Canada is gratefully acknowledged. Special thanks is extended to Vanessa Yingling for her contribution to the initial stages of this research.
Stuart M. McGill received a Ph.D. degree from the University of Waterloo in 1986. He is a professor of spine biomechanics in the Department of Kinesiology at the University of Waterloo where his research interests focus on how the spine works and becomes injured, and on developing better strategies for prevention and rehabilitation of low back pain. He is married, has two children and enjoys fishing.