Electromyography Pattern's During Swinging In Children With Upper Limb Loss: Wearing Their Prosthesis Versus Not Wearing It
Genn Carly Craig Katelynn Hill Wendy Chester Vicky Biden Edmund
This study focused on the question of how a child with unilateral limb loss performed activities of daily living while wearing a prosthesis compared with not wearing it.
The aim of this study was to determine if patterns of muscle activity in a child with unilateral limb loss were within normal limits during swinging and biking compared to normally limbed children doing the same task. This was done by looking at bilateral muscle symmetry, agonist/antagonist muscle pairs, activation timing and length of activation for each muscle.
The methods used included comparing a sample population of normally limbed children with a sample population of children with unilateral limb loss. The data collected included motion data captured using a motion analysis system called Vicon from Oxford Metrics and a wireless EMG system called Zero-wire from Aurion. Each subject had reflective markers placed on anatomical landmarks in order to observe their motion as well as 10 electrodes placed on 5 muscles on each side of the body.
During swinging there were statistically significant differences between the symmetry of the biceps, triceps, and deltoids of the control population compared to both test groups (prosthesis users wearing and not wearing their prosthesis). However, no statistically significant differences were observed between children with limb loss while wearing their prosthesis versus not wearing it.
Upper extremity limb loss is uncommon compared to lower extremity limb loss but is more prevalent in children because of the incidence of congenital cases. Upper extremity limb loss may occur traumatically, surgically, or congenitally.
Currently there are two very opposite schools of thought in terms of fitting unilateral, below elbow children with congenital deficiencies with electric prostheses. The first is to fit the child with a passive prosthesis when they can sit up and subsequently fit them with an electric device between 12 and 15 months1 2. The second is that since this group of children are able to perform most activities of daily living (ADL's) independently, the prosthesis provides little to no increase in functionality3 4. An interesting factor to consider in this argument could be the risk level of these children at acquiring repetitive strain injuries. There have been some studies that have already indicated that upper limb amputees have had overuse problems5 6.
The main focus of this thesis was to describe the muscle activity in children with unilateral limb loss and to decipher whether this activity lies within normal limits during activities of daily living. The three activities chosen were swinging, biking, and walking, however this paper will focus only on the swing task. The hypothesis was that throughout the activities chosen for study, children with unilateral limb loss would have muscle symmetry and activation patterns, which lie outside the normal range. A companion study is being done concurrently to look at the effects the prosthesis may have on joint loads when performing activities of daily living.
The tools that will be used to identify these potential differences are encompassed in the state-of-the-art motion laboratory located in the Institute of Biomedical Engineering at the University of New Brunswick. Included in this laboratory are an infrared eight-camera VICON motion capture system, a Zero Wire electromyography (EMG) system, and four Kistler forceplates embedded into the floor of the laboratory. The VICON motion capture system was utilized to observe how the patient moved while performing these gross motor tasks in terms of body segment angles and rotations. The surface electromyography (SEMG) system monitored the muscle activity of ten major upper body muscle groups while performing two activities. The force plates were used specifically for one of the activities (walking).
Individuals eligible to be included in the study were children between the ages of 5 and 14 years old. The control group had to have two asymptomatic arms. The prosthesis users were recruited from the clinical population of the Institute of Biomedical Engineering. They were in the same age range and had unilateral, below-elbow limb loss. The prosthesis users were tested once while wearing their myoelectric prosthesis and once without.
These experiments attempt to clarify muscle activity for the upper limb child amputee, and to determine whether this activity lies outside normal limits. Currently, there is little to no research that has reported the EMG activity of children with unilateral upper limb loss. There have been a few studies that have reported these individuals to be at a greater risk for repetitive strain injuries5 6 however there is not enough evidence currently to confirm this. This research attempts to initiate more interest in these problems in order to eventually provide a clearer prosthesis prescription plan for individuals who have unilateral limb loss (particularly those in the congenital populations).
Table 1 contains the demographic information for the subjects included in this paper.
|Table 1: Demographic information for control and test groups|
|Subject||Subject Type||Age||Gender||Dominant Side||Non-Dominant Side|
Range for the control group was between 4 and 14, and had an average of 9.3 years, with 50% males8, and 50% females8. The range for the test group was between 6 and 14, and had an average of 10.8 years, with 75% females3 and 25% males1.
The main focus of this paper is to describe the muscle activity in children with unilateral limb loss and to decipher whether this activity lies within normal limits during activities of daily living. In order to achieve this, the use of motion analysis, and electromyography were required.
The subjects who participated in this study were split into two groups between the ages of 5 and 14. The control group consisted of normally limbed children and were recruited by email throughout the faculty, staff and students at the University of New Brunswick as well as through the Scout Association in Fredericton. The group of individuals in the test group were clients of the upper-extremity fitting centre located in the Institute of Biomedical Engineering. The occupational therapist at the clinic asked children in the appropriate age bracket if they would consider being a part of the study. This led to four participants. Therefore, in total there were 16 children in the control group and 4 in the test group.
The two main components that were used throughout this study were the YICON motion analysis system as well as the zero-wire wireless EMG system. Both of these systems were connected to the Analog-to-Digital box provided by the VICON station, which allowed for 64 channels in total.
Motion Capture System
The Yicon Motion camera system uses eight infrared cameras to locate reflective marker balls in 3D space. Each camera emits infra-red rays that reflect off the markers, and it is these reflections that get captured by the cameras. For this study, the video system was calibrated at a sample rate of 60 Hz, which was more than double the frequency at which the subject's movements were occurring. Figure 4 in the data analysis section shows evidence of this. This satisfies the Nyquist Theorem requirements that any frequencies being captured must be half the sample rate of the system or lower.
Surface electromyography was used exclusively to determine the timing of the muscle activation during gross motor activities. This allowed for determination of muscle activation symmetry as well as timing and length of muscle activation. The analog system was calibrated separately, as it had an entirely different sample rate of 1080 Hz. All analog trials connected to the 64-channel Analog-to-Digital board were calibrated simultaneously. The analog calibration was done when the subject was lying face down on a massage table to ensure minimum muscle activity.
The Duotrodes used were silver-silver chloride, bi-polar, disk-shaped electrodes. Pre-gelled contact surfaces were 12.5 mm diameter and were mounted in pairs 19 mm from center to center.
The Zero Wire EMG system was the second main system used in this study. It has three main components, the duotrodes, transmitting electrodes, and the receiver which is integrated through Vicon's 64-channel A/D board. Figure 1 shows these components.
The protocol on the determination of muscles examined was as follows:
- Muscle had to be on the trunk or the upper limbs,
- Arm muscles had to be above the elbow (due to the individuals with limb loss wouldn't have certain muscles below the elbow and this would defeat the purpose),
- Muscles had to be easy to locate, and therefore were usually large muscles close to the surface of the skin,
- Muscles were thought to be active in at least 2 activities chosen.
According to these factors, 10 muscles, five on each side, were selected for observation. They were biceps, triceps, trapezuis, deltoidus medialus, and erector spinae. Out of the 10 muscles observed, the erector spinae was the most difficult muscle to accurately capture due to its deeper origins, and development in children. Since there is no documented observation of upper body muscles in children during gross motor activities, there was no literature to guide these muscle choices. The muscle activity patterns of these 10 selected muscles were described based on the swing cycle.
A European group called SENIUM7 has in place guidelines for electrode placement (shown in Figure 2 ) that is meant to increase the consistency and validity of EMG studies. In addition to the reflective markers, the 10 electrodes that monitored EMG activity were placed based on the SENIAM guidelines for surface electrodes.
The swing task (demonstrated in Figure 3 ) is something that children may do on a regular basis and provides an activity that is highly cyclic, making it relatively simple to observe. In addition, swinging requires upper body muscle activity, making it an ideal task for this experiment.
For the control group, each subject was required to perform the swing activity between 4 and 8 times, depending on their attention span and whether or not they were performing the task as they usually would. The test group had to perform two rounds of the swing activity, one while wearing their prosthesis and another while not wearing it.
The Electromyography data were used to determine what muscles were active during one cycle of each task. This analysis was simplified for the swing task because it was highly cyclic. The muscle activity was determined to be "on" or "off' based on a calculated threshold value; this is described in more detail in the Muscle Onset & Offset Calculation section. The MATLAB programming code was used for reading the EMG signal and performing these calculations. Once the activation timing for each muscle was determined it was described against a percentage of each cycle in much the same way gait data are usually reported.
Tools used during data analysis include, VI-CON workstation software, MATLAB, Minitab, and excel. The analog data were sampled at a rate of 1080 Hz which was at least two times the bandpass output of the Zero-wire electrodes (between 10 and 500 Hz). The video data were captured at a rate of 60 Hz, which was at least double the highest frequency occurring from the gross motor activities.
After the data were filtered and segmented, a cycle was defined for the swing task. The swing cycle was determined to start at the furthest point the toe (right or left) travelled in the backswing; the halfway point in the cycle was determined to be the furthest point the toe travelled in the forward swing; and the end of the cycle was the same as the next cycles starting point, the furthest point the toe travelled in the backswing. Figure 4 below describes the swing cycle.
Figure 5 describes the coordinate system relative to the swing task. Based on this, the maximum points the right toe travelled were actually the points where the toe is furthest point back, corresponding to the starting and ending of each swing cycle.
An example of the oscillations made by the right toe during swinging is shown in Figure 6 . This particular subject slowly increased their swinging amplitude with each swing. This was generally the case, however some subjects got up to their highest amplitude (or steady state amplitude) faster.
For each trial, the most typical swing cycle was chosen to be analyzed. This was done by comparing all cycles occurring within each trial, taking the average of these cycles and then choosing the cycle that most closely followed the same pattern as the average cycle. In Figure 7 , all of the cycles for Subject E were arranged on the same plot as well as the calculated average out of these cycles. For subject E, trial 2, cycle 4 was chosen to be analysed.
Muscle Onset & Offset Calculation
Muscle onset timing, muscle activation patterns, and muscle symmetry were the main factors being assessed and were described for each muscle. Muscle onset timing has been determined many various ways, including, visually, as a percent of the peak magnitude, and higher than a threshold value7-22. Since, the magnitude in this study was not normalised, no calculation of muscle onset could be relative to this value. A study done by Hodges & Bui, in 199623evaluated the accuracy of these various types of methods to determine muscle onset timing and additionally identified the best combination of these parameters for detection accuracy. Using signals with both high and low noise levels, the best combination of parameters to accurately detect muscle onset was shown in Equation 1: Threshold calculation to determine whether muscle is "on" or "off'. Since this equation was used in this analysis, the most important calculation necessary for the determination of muscle activation patterns was the threshold level that determined whether the muscle is on or off, depending on whether the muscle activity was above or below the threshold value respectively.
z = 3 * a(x(for 25ms)) *
Equation 1: Threshold calculation to determine whether muscle is "on" or "off'
* low pass filtered with 6th order elliptical @ 50Hz.
a: Standard Deviation
t: Threshold value
x: is the baseline noise for a segment of at least 50ms13 23 prior to muscle activation when the muscle is off.
Therefore, for a muscle to be considered "on", it had to exceed the standard deviation of the baseline mean noise by a factor of 3 for at least 25 ms. The baseline noise chosen for this experiment was a segment of muscle activity during the task trial where the muscle was inactive. The baseline noise had to be at least 50 ms long23 13. For consistency, all threshold calculations were calculated using Equation 1. This was done based on the results of the study done by Hodges and Bui, 1996, and also to keep repeatability as high as possible. The Matlab program written specifically for this data analysis and is described in the flow chart in Appendix A.
Results and Discussion
For each trial, five values that described muscle activity were calculated and are described in Table 1. Each of these variables assists in describing muscle symmetry and activation timing.
|Table 2: Explains each of the six values describing
muscle activity were calculated for each Subject
|Variable Name||Definition||Matrix Size|
|EMG Onset||Stores the onset and offset information for all 10 muscles. This is a matrix with only two values, either 1.0 for muscle "on", or 0.5 for muscle "off".||[rxlO]
|EMG Basic||Stores the muscle activity for all 10 muscles during the cycle that was chosen for analysis.||[lOxr]
|EMG Percentage||Stores the percentage of the task cycle that each muscle is "on".||1x10|
|EMG Symmetry||Stores the percentage of time the same muscles on the left and right side are both "on", or both "off".||1x5|
|EMG Agonist||Stores the percentage of time the agonist and antagonist muscles are co-contracting.||1x4|
A Mann-Whitney statistical test was performed for the percentage of time each muscle was active throughout the swing cycle, the results for all muscles are shown in the boxplot in Figure 8 . There were a few significant differences within this group and they are listed below,
- Non-Dominant Bicep Control versus Non-Dominant Bicep Test (WEARING)P<0.05 (0.021)
- Non-Dominant Tricep Control versus Non-Dominant Tricep Test (WEARING)p<0.05 (0.016)
- Non-Dominant Tricep Control versus Non-Dominant Tricep Test (WITHOUT)p<0.05 (0.016)
The data are grouped either within the dominant or non-dominant arm of the five muscle examined. In each group, there are three different subject conditions; control group, test group wearing their prosthesis, and test group without their prosthesis. The data are also sorted by the control group median in descending order. The median muscle activity in order is 'Non-Dominant Bicep', 'Dominant Bicep', 'Non-Dominant Tricep', 'Dominant Tricep', 'Non-Dominant Trapezuis', 'Dominant Deltoid', 'Dominant Trapezuis', 'Non-Dominant Deltoid', 'Non-Dominant Eresp', 'Dominant Eresp'.
The second variable that was statistically analysed was the occurrence of co-contraction between agonist/ antagonist muscles on the same side of the body. Figure 9 describes the boxplot for the occurrence of cocontraction throughout the swing cycle. Most of the muscle pairs had co-contraction throughout approximately 40% of the swing cycle. There were no significant differences between the control group, and either of the test groups, for all agonist/antagonist muscle pairs.
The last variable that was statistically analysed was thi symmetry between right and left sides for all 5 muscles. This may be the best indicator of compensations made by the test groups when performing the swing activity. After the Mann-Whitney statistical test was performed, it was found that there were some statistically significant differences between the test and control groups;
- Bicep of control group compared with test group wearing their prosthesis (p<0.01)
- Bicep of control group compared with test group not wearing their prosthesis (p<0.05)
- Triceps of control group compared with test group wearing their prosthesis (p<0.05)
- Triceps of control group compared with test group not wearing their prosthesis (p<0.01)
- Deltoids of control group compared with test group wearing their prosthesis (p<0.01)
- Deltoids of control group compared with test group not wearing their prosthesis (p<0.05)
Therefore, for muscle symmetry, 3 out of the 5 muscle groups examined showed significant differences between the control group and both the prosthesis wearers and non-wearers. The muscle symmetry for all muscles in the control group was at approximately 80% of the entire swing cycle. The muscles that displayed significant differences were always lower for the test groups than the control group, demonstrating less symmetry among children with limb loss (which was expected). Another interesting fact to observe from Figure 10 is that the control group, which had 16 individuals showed much less variance than the prosthesis user group, even though there were only 4 individuals. This was a general trend with the other two results studied as well. Therefore, these results should be done again as case studies, and compared.
Muscle Activation Patterns
To describe muscle activity in patterns, similar principals to a 95% confidence interval were used. Therefore, for any activity to be recorded as "normal", in the control group, at any one point during the swing cycle, the muscle examined had to be "on", in 15 out of 16 subjects (94%).
Since there were only four subjects in the test group, the muscle being examined had to be "on" in all four subjects at the same point in the swing cycle.
It is important to note that out of these four subjects, two had limb loss on the right side, and two had limb loss on the left. The first muscle group examined were the Biceps; the right and left sides are described below in Figure 11 .
For the control group, the dominant Bicep was active from 33% to 50% of the swing cycle, which is the portion of the cycle leading up to the backswing. The test group wearing their prosthesis had more activity between 5% and 50%. The test group not wearing their prosthesis was more active than the control group as well, between 5 and 45% and on again between 84 and 90%. The non-dominant bicep in the control group was active similar to the right side, but shifted slightly earlier in the cycle (on between 25 and 45%). In this case, the test group not wearing their prosthesis showed more normal activation patterns, on between 20 and 40 % with two additional periods active at the start and end of the cycle. The test group wearing their prosthesis had no consistent activation on the nondominant side.
The triceps have much less activity within the control group and both test groups. The control group showed the dominant Tricep active between 84 and 89%; the prosthesis wearers had more activity starting at the same time as the control's, "on" between 84 and 98%. The test group without their prosthesis had two periods of activity between 20 and 40% and 80 and 95%. This is described in Figure 12 . The non-dominant triceps control had similar activity to the dominant side however, with activation shifted earlier in the cycle (active between 79 and 84%).
The dominant trapezuis muscle for the control group was active between 32 and 57%. The prosthesis wearers and non-wearers had much more activity during the swing cycle. The test group wearing a prosthesis was on between 30 and 48%, again between 59 and 77%, and on briefly between 19 and 22% and another brief spurt between 87 and 89%. Similar to the dominant Bicep activity, the dominant trapezuis muscle in the control's appear to be on just prior to the back-swing. This is shown in Figure 13 . The non-dominant trapezuis control had similar activity to the dominant side however, with activation shifted earlier in the cycle (active between 28 and 53%). This is similar to the Biceps and Triceps in that the non-dominant side of the controls had similar activation as the dominant side, with activation occurring approximately 5% earlier in the swing cycle. Similar to the non-dominant Bicep in the test group wearing their prosthesis, there was no consistent activation. The test group wearing their prosthesis was actively briefly between 36 and 40%, and on almost constantly between 50 and 87%.
The dominant deltoids for the control group had similar activation to the dominant triceps for the controls, being active for a small portion of the swing cycle, between 81 and 85%. The dominant deltoid for the test group wearing a prosthesis had no consistent activation throughout the entire swing cycle. The test group not wearing a prosthesis had much longer activation, "on" between 32 and 41%, and then intermittently "on" and "off' between 57 and 91%. The non-dominant deltoids for the control group had even lower activity than the dominant side, on very briefly between 37 & 38%, and 80 & 81%. The test group wearing a prosthesis some activity on the non-dominant side between 67and 70%, which was more than the dominant side (no activity). The non-dominant side of the test group not wearing a prosthesis had no activity throughout the swing cycle. Figure 14 describes the activity of the right and left deltoids.
There was no observed activity consistent for 95% of all subjects for all groups for the dominant and non-dominant erector spinae muscles. This could partially be due to the difficulty in placing the erector spinae electrodes on children.
In general the agonist muscles (Biceps & Trapezuis) were activated approximately between 30 and 50 percent of the swing cycle and were much more active than the antagonist muscles (Tricep & Deltoid). This makes sense as an agonist by definition has the major responsibility of causing movement, while antagonist muscles oppose or reverse a movement24. In swinging, the arms themselves do not move a whole lot, but the power generated assists to propel the body forwards and backwards.
Another interesting trend within the control group, was that the dominant side for the biceps, triceps, and trapezuis muscles had very similar activation as the non-dominant side for the same muscles. However the dominant side muscles were activated approximately 5% later in the swing cycle in all cases. Additionally, for the biceps and trapezuis muscles, the nondominant side showed no consistent activity for the test group while wearing their prosthesis.
The test group not wearing a prosthesis generally showed more activation than the control group for the non-dominant sides of the biceps and trapezuis muscles.
In general, the prosthesis wearers and non-wearers had much more activity than the control group for the bicep, tricep and trapezuis muscles on the dominant side. The non-wearers and wearers had very different patterns compared with the controls for the non-dominant side; the non-wearers being much more active, and the wearers having no activity at all. Therefore, both the test groups (children with limb loss), had differing muscle activation patterns than the control group during swinging. Additionally, it should be noted that the erector spinae muscle was difficult to locate and often had inconsistent results; this is re-enforced by the fact that there seems to be approximately 10 total muscle activity (Figure 9 ) for the dominant and non-dominant erector spinae muscle in all groups, yet no consistent activity was found when looking at the muscle patterns throughout the swing cycle.
Overall, 3 out of the 5 muscle groups examined showed significant differences between the control group and both test groups for percent muscle symmetry throughout the swing cycle. Additionally, the differences in muscle patterns between the control group and the test groups give the hypothesis that children with limb loss would display muscle patterns outside the normal range was correct, however there were no significant differences between the group of individuals wearing a prosthesis and not wearing one.
Institute of Biomedical Engineering, University of New Brunswick
- Gaebler-Spira, D. and Uellendahl, J. Pediatric limb deficiencies, [book auth.] ed. Molnar. Pediatric Rehabilitation . Philadelphia : Hanley & Belfus, 1999, pp. 331-350.
- Early upper limb prosthesis fitting: When and what do we fit . Shaperman, J., Landsberger, S. and Setoguchi, Y. 2003, Prosthetics and Orthotics International, pp. 11-19.
- Prosthetic management of children with unilateral con-fenital-below-elbow deficiency . Davids, J., et al. 2006, Journal of Bone and Joint Surgery, pp. 1284-1300.
- Impact ofprostheses on function and quality of life for children with unilateral congenital below-the-elbow deficiency . James, M., et al. 2006, Journal of Bone and Joint Surgery, pp. 2356-2365.
- Save that arm: A study of problems in the remaining arm of unilateral upper limb amputees . Jones, L. and Davidson, J. 1999, Prosthetics and Orthotics International, pp. 55 -58.
- Functional outcome of patient with proximal upperlimb deficiency-acquired and congenital . Datta, D., Selvarajah, K. and Davey, N. 2004, Clinical Rehabilitation, pp. 172-177.
- Reflex activation of gluteal muscles in walking with balance shoes: an approach to restoration of function for chronic low back pain patients . Bullock-Saxton, J. 1993, Spine, pp. 704-708.
- Local sensation changes and altered hip muscle function following severe ankle sprain.Bullock-Saxton, J. 1994, Physical Therapy, pp. 17-31.
- Patterns of rapid motor responses during postural adjustments when standing in healthy subjects and hemiplegic patients . Badke, M. and Duncan, P.1983, Journal of Physical Therapy, pp. 13-20.
- Electromyographic study of muscular overglow during precision handgrip . Carey, J., Allison, J. and Mundale, M. 1983, Physical Therapy, pp. 505-511.
- Functionally complex muscles of the cat hindlimb, III: differencial activation within biceps femoris during postural perturbation . Chanaud, C. and Macpherson, J. 1991, Exp. Brain Res., pp. 271-280.
- The use of surface electromyography in biomechanics . De Luca, C. 1997, Journal of Applied Biomechanics, pp. 135-146.
- Reliability of computerized surface electromyography for determining the onset of muscle activity . DiFabio . 1987, Journal of Physical Therapy, pp. 43-48.
- Computerized interpretation of initiation and relaxation of EMG recorded dynamic muscle contraction . Greenisen, M. and Vroomen, L. 1979, Proceedings of the 4th Congress of the International Society of Electrophysiological Kinesiology, Boston, USA, pp. 172-173.
- Goal-directed arm movements, I: analysis of EMG records in shoulder and elbow muscles . Happee, R. 1992, Journal of Electromyography in Kinesiology, pp. 165-178.
- Onset timing of electromyographic activity in the vastus medialis oblique and vastus lateralis muscles in subjects with and without patellofemoral pain syndrome . Karst, G. and Willet, G. 1995, Physical Therapy, pp.813-823.
- The ABC of EMG. A practical introduction to kinesi-ological electromyography . Konrad, P. 2005, Electromyography.
- Stance posture control in select groups of children with cerebral palsy: deficits in sensory organisation and muscular coordination . Nashner, L., Shumway-Cook, A. and Marin, O. 1983, Exp Brain Res, pp. 393-409.
- Preparation for movement in the cat, II: unit activity in the basal ganglia and thalamus . Neafsey, E., Hull, C and Buchwald, N. 1978, Clinical Neurophisiology, pp. 714-723.
- Postural responses and effector factors in persons with unexplainedfalls: results and methodological issues . Studenski, S. Duncan, P. and Chandler, J.1991, Journal of American Geriatric Society, pp. 229-234.
- Landing from a jump: the role of vision when landing from known and unknown heights . McKinley and Thompson . 1995.
- Relation between muscle response onset and body segmental movements during postural perturbations in humans . Woollacott, M., von Hosten, C. and Ros-blad, B. 1988, Exp. Brain Res., pp. 593-604.
- A comparison of computer-based methods for the determination of onset of muscle contraction using electromyography . Hodges, P. and Bui, B. 1996, Electromyography and Clinical, pp. 511-519.
- Marieb, E.Essentials of Human Anatomy and Physiology . San Fransisco, CA : Pearson Education Inc., 2006.