- Email: [email protected]
Power spectral analysis of surface electromyography (EMG) at matched contraction levels of the first dorsal interosseous muscle in stroke survivors
Clinical Neurophysiology 125 (2014) 988–994
Contents lists available at ScienceDirect
Clinical Neurophysiology journal homepage: www.elsevier.com/locate/clinph
Power spectral analysis of surface electromyography (EMG) at matched contraction levels of the first dorsal interosseous muscle in stroke survivors Xiaoyan Li a,⇑, Henry Shin b, Ping Zhou a,c,d, Xun Niu a, Jie Liu a, William Zev Rymer a,b,c a
Sensory Motor Performance Program, Rehabilitation Institute of Chicago, Chicago, IL, USA Department of Biomedical Engineering, Northwestern University, Chicago, IL, USA Department of Physical Medicine and Rehabilitation, Northwestern University, Chicago, IL, USA d Institute of Biomedical Engineering, University of Science and Technology of China, Hefei, China b c
a r t i c l e
i n f o
Article history: Accepted 30 September 2013 Available online 21 November 2013 Keywords: Surface electromyography (EMG) Stroke Spectral analysis First dorsal interosseous (FDI) muscle
h i g h l i g h t s Spectral analysis of surface EMG post stroke was performed in FDI muscles. Subjects showed reduced mean frequency in paretic side at matched forces. Spectral difference between two sides was not correlated to clinical scales.
a b s t r a c t Objective: The objective of this study was to help assess complex neural and muscular changes induced by stroke using power spectral analysis of surface electromyogram (EMG) signals. Methods: Fourteen stroke subjects participated in the study. They were instructed to perform isometric voluntary contractions by abducting the index finger. Surface EMG signals were collected from the paretic and contralateral first dorsal interosseous (FDI) muscles with forces ranging from 30% to 70% maximum voluntary contraction (MVC) of the paretic muscle. Power spectral analysis was performed to characterize features of the surface EMG in paretic and contralateral muscles at matched forces. A Linear Mixed Model was applied to identify the spectral changes in the hemiparetic muscle and to examine the relation between spectral parameters and contraction levels. Regression analysis was performed to examine the correlations between spectral characteristics and clinical features. Results: Differences in power spectrum distribution patterns were observed in paretic muscles when compared with their contralateral pairs. Nine subjects showed increased mean power frequency (MPF) in the contralateral side (>15 Hz). No evident spectrum difference was observed in 3 subjects. Only 2 subjects had higher MPF in the paretic muscle than the contralateral muscle. Pooling all subjects’ data, there was a significant reduction of MPF in the paretic muscle compared with the contralateral muscle (paretic: 168.7 ± 7.6 Hz, contralateral: 186.1 ± 8.7 Hz, mean ± standard error, F = 36.56, p < 0.001). Examination of force factor on the surface EMG power spectrum did not confirm a significant correlation between the MPF and contraction force in either hand (F = 0.7, p > 0.5). There was no correlation between spectrum difference and Fugl–Meyer or Chedoke scores, or ratio of paretic and contralateral MVC (p > 0.2). Conclusions: There appears to be complex muscular and neural processes at work post stroke that may impact the surface EMG power spectrum. The majority of the tested stroke subjects had lower MPF in the paretic muscle than in the contralateral muscle at matched isometric contraction force. The reduced MPF of paretic muscles can be attributed to different factors such as increased motor unit synchronization, impairments in motor unit control properties, loss of large motor units, and atrophy of muscle fibers. Significance: Surface EMG power spectral analysis can serve as a useful tool to indicate complex neural and muscular changes after stroke. Ó 2013 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved.
⇑ Corresponding author. Address: Sensory Motor Performance Program, Rehabilitation Institute of Chicago, 345 E Superior St, Suite 1406, Chicago, IL, USA. Tel.: +1 312 238 1174. E-mail address: [email protected] (X. Li). 1388-2457/$36.00 Ó 2013 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.clinph.2013.09.044
X. Li et al. / Clinical Neurophysiology 125 (2014) 988–994
1. Introduction Spectral analysis of electromyogram (EMG) signals has been used for decades to reveal pathophysiology of neuromuscular impairment (Dimitrov et al., 2008; Farina et al., 2004; Kaplanis et al., 2009; Lindstrom and Magnusson, 1977; Muro et al., 1982). Changes in spectral frequency of surface EMG are associated with peripheral factors, such as muscle fiber conduction velocity, muscle fiber length and orientation, motor unit location, intracellular action potential shape and its negative after-potentials (Dimitrov et al., 2008; Farina et al., 2004; Lindstrom et al., 1970; Lowery et al., 2000; Zaman et al., 2011). Modifications of the spectral characteristics are also related to synchronization of multiple motor units (mainly associated with central nervous system) or motor unit recruitment and firing rate changes (associated with both central nerve system and motor unit properties) (Gabriel and Kamen, 2009; Lago and Jones, 1977; Solomonow et al., 1990; Wakeling, 2009; Yao et al., 2000). Other non-physiological factors such as temperature, thickness of the subcutaneous layers, electrode position and configuration, etc. may also influence the EMG spectral distribution (Moritani and Muro, 1987; Petrofsky and Lind, 1980; Zipp, 1978). Subsequent to a cerebral lesion, progressive peripheral changes in hemiparetic muscles have been reported during the course of the disease involving loss of muscle fibers, changes of motor unit type composition, loss of functioning motor units, or structural reorganization of survival motor units (Brown and Snow, 1990; Charcot, 1893; Dattola et al., 1993; Hara et al., 2004; Spaans and Wilts, 1982). Meanwhile, disorganization of motor unit control properties have been observed in the hemiparetic muscles described as compressed motor unit recruitment range, abnormally low motor unit discharge rate (Gemperline et al., 1995; Rosenfalck and Andreassen, 1980), as well as increased intensity of motor unit synchronization (Farmer et al., 1993). Presently, there are few studies documenting the impact of post stroke neuromuscular modifications on the spectral characteristics of the surface EMG. In one study an increase of power density in the lower spectral frequency section was observed in one subject’s paretic biceps brachii muscle whereas no significant spectrum difference was found between paretic and contralateral muscles in the remaining five subjects (Gemperline et al., 1995). The spectral analysis was also used to characterize individual motor unit action potentials (MUAPs) which demonstrated lower mean power frequency in the paretic side compared with the contralateral side (Kallenberg and Hermens, 2009). However, there was no significant difference of the mean power frequency in the global surface EMG analysis between two sides in the same group of subjects. Another question that has not been intensively explored in stroke is the relation between the EMG power spectrum and the contraction force. Previous investigations of such relation in healthy subjects have provided contradictory observations (Farina et al., 2002; Kaplanis et al., 2009; Rainoldi et al., 1999; Seki et al., 1991). For example, it was reported that the mean power frequency (MPF) or median frequency (MF) values of surface EMG continuously increase with isometric force up to 80% maximal voluntary contraction (MVC) (Kaplanis et al., 2009; Moritani and Muro, 1987). Conversely, decreased spectrum frequency was also observed with the increment of contraction levels (Gabriel and Kamen, 2009; Rainoldi et al., 1999). In stoke there was only one study investigating the relation of surface EMG spectrum and force, which reported a reduction of median frequency in the hemiparetic muscles and a slight increase of median frequency in the contralateral side along force (Kallenberg and Hermens, 2009).
989
In light of the few and inconsistent observations on post stroke power spectral changes of surface EMG, the objective of the current study was to evaluate the spectral characteristics of the surface EMG in stroke survivors. Particularly, we compared power spectrum patterns of surface EMG between paretic and contralateral muscles during sustained voluntary contractions. We further examined the association between surface EMG spectral characteristics and isometric muscle contraction levels. 2. Methods 2.1. Subjects Fourteen chronic stroke survivors (6 females and 8 males, aged 45–70 years old) with mild to severe weakness in the contralesional side participated in the study. All subjects were screened by a physician based on clinical history and physical examination. Subjects with concurrent neurological disorders or other symptoms (such as neuropathy, radiculopathy, cervical spondylosis, and hyperglycemia, etc.) were excluded. No subjects reported any arm pain, numbness or paresthesia. The disease duration since the onset of the stroke ranged from 1 year 6 months to 24 years. All subjects submitted written consent approved by the Institutional Review Board of Northwestern University (Chicago, IL, USA) before experiments. Their motor function was evaluated based on the Fugl-Meyer test and the Chedoke-McMaster test. Maximal voluntary contraction (MVC) force was additionally measured from the first dorsal interosseous muscles of the paretic and contralateral hands, respectively. The MVC ratio between paretic MVC and contralateral MVC was used as an index of relative weakness (Gemperline et al., 1995). A summary of subjects’ information is presented in Table 1. 2.2. Experiments Subjects were seated comfortably in a mobile Biodex chair. Their forearm and wrist were casted and positioned on a plastic platform, where the wrist was fastened inside a ring-mount interface. This setup kept the wrist in a pronated position and minimized its movement (Fig. 1a). The index finger was immobilized in a fiberglass cast and placed inside a small ring interface that was attached to a six degree-of-freedom load cell (ATI, Apex, NC). The position of the index finger was aligned with the center of the load cell. Surface EMG signals were recorded from a four-channel sensor array (Delsys, Boston, MA) dwelling on top of the first dorsal interosseous (FDI) muscle. The configuration of the sensor is illustrated in Fig. 1b, as five pin electrodes in a square, one in each corner and the fifth in the center. The inter-electrode distance between two neighboring corner electrodes is 5 mm. Each pin electrode is 0.5 mm in diameter. Details of the sensor can be found in (De Luca and Hostage, 2010). Since it was difficult to align all the four bipolar recording pairs of pins (i.e. the center pin with respect to each of the four corner pins) with the muscle fiber orientation, placement of the sensory array was arranged in a way that three electrode-pins (two diagonal pins and the center pin) were in parallel with the orientation of the muscle fibers. This assures the same alignment of the electrode with respect to the FDI muscle. The output differential signals were filtered with a bandpass filter of 20–2000 Hz. Voluntary contraction force exerted by the index finger was measured in abduction (Fx), flexion (Fy), and the anterior–posterior direction (Fz) in the sagittal plane. All forces and surface EMG signals were sampled at 20,000 Hz using EMGWorksÒ (Delsys, Boston, MA).
990
X. Li et al. / Clinical Neurophysiology 125 (2014) 988–994 Table 1 Stroke subjects’ information. ID
Sex
Age (years)
Duration (years)
Paretic side
Chedoke score
FM
MVC ratio
1 2 3 4 5 6 7 8 9 10 11 12 13 14
M M F M M M M F F F F M M F
62.2 45.2 52.3 61.8 56.8 66.8 70.0 68.2 44.9 59.5 59.7 48.7 61.4 58.6
1.7 5.7 6.8 7.8 8.0 8.7 1.5 5.0 13.7 19.7 24.1 1.7 5.0 5.5
Left Right Right Right Right Left Left Right Left Left Right Right Right Right
7 5 5 4 5 4 7 5 4 3 2 6 6 4
66 58 58 56 51 24 56 53 16 29 22 48 63 40
0.75 0.60 0.66 0.19 0.82 0.27 0.77 0.28 0.17 0.36 0.52 0.61 0.85 0.87
FM: Fugl-Meyer test; MVC ratio: MVC paretic/MVC contralateral.
Fig. 1. (a) Experimental setup; (b) configuration of the array sensor (Delsys, Boston, MA) and (c) an example of a typical trial includes a rest period (baseline), a rising ramp, and a constant force period.
In the beginning of the experiment, the subjects were required to perform isometric maximum voluntary contractions (MVC) three times using the paretic index finger in the direction of abduction. The MVC of the paretic FDI muscle was determined as the highest value of the three efforts, and then this value was used to determine the desired force levels for both paretic and contralateral muscles of the subjects. The magnitude of desired force levels was set as 30–70% of the paretic muscle MVC in 10% MVC increments. For each desired force level, the duration of a typical trial included a rest period of 3 s, two ramp periods when the abduction force inclined or declined at a speed of 10% paretic muscle MVC per second, and a sustained force period for at least 10 s (Fig. 1c). Depending on the magnitude of the desired force levels, the duration of a trial varied from 20 to 30 s. Following a brief rest, the subjects were instructed to generate isometric contractions to match the magnitude of desired force using the index finger. The magnitude of a desired force and the contraction force generated by the subjects were displayed as real time trajectories in different colors. Such force feedback helped subjects adjust their contraction force to match the desired force level. Throughout the experiment, the subjects were encouraged to maintain the abduction force as stable as possible and minimize the flexion force simultaneously. At least
two successful trials at each contraction level were obtained before advancing to the next contraction level. The order of trials at different force levels was randomized. Practice was given to help them become familiarized with the task at each force level. A brief rest period was provided between trials to prevent potential fatigue. Subjects were scheduled for the experiments on separate days for the paretic and contralateral muscles. The experiment on the paretic muscle was always performed first so that the desired force levels could be determined prior to the contralateral muscle experiment. In addition to desired force levels, the MVC of the contralateral muscle was also recorded. 2.3. Data analysis Offline analysis was performed on the surface EMG signals as well as the abduction (Fx) and flexion (Fy) forces. We averaged the abduction and flexion forces within the sustained force period to calculate the flexion deviation, defined as the inverse tangent of the ratio between flexion and abduction forces (Fy/Fx). Trials with flexion deviation larger than 30° were excluded from analysis. Next, the signal to noise ratio (SNR) of surface EMG was examined for each channel. Spectral analysis was conducted on the channel
X. Li et al. / Clinical Neurophysiology 125 (2014) 988–994
that had the largest SNR values. The pair of pins that demonstrated the largest SNR of the differential signal was the one arranged along the muscle fibers. This was consistent for all the recordings. The Hamming window was used to smooth the signal before the periodogram method was applied to estimate the power spectrum of the surface EMG signals. The frequency resolution was 2 Hz. The mean power frequency (MPF) was calculated for each trial and then averaged over trials at the same level. Comparisons were made between the paretic and contralateral sides at the same force levels (i.e. matched contractions). The spectrum difference between the paretic and contralateral muscles was further analyzed by employing the spectral distribution function to detect small variations between two waveforms of the power spectra distribution from paretic and contralateral muscles. The spectral distribution function was defined as the normalized integral of the power spectrum (Lowery et al., 2000; Merletti and Lo Conte, 1997; Rix and Malenge, 1980). At a given distribution percentile, the frequency value in the hemiparetic muscle (fp in Fig. 2) could be different from that in the contralateral muscle (fcon in Fig. 2). Thus, a scaling factor was computed as the mean ratio of fp/fcon to indicate frequency compression. This technique calculated the shifts of spectral distribution across all percentile frequencies of the entire waveform and provided a more accurate index for the estimation of muscle fiber conduction velocity compared with the use of the median frequency of the power spectrum (Lowery et al., 2000). The following two main effects on the surface EMG spectral characteristics were investigated: (1) side effect (paretic muscle vs. contralateral side), and (2) force effect (from 30% to 70% paretic MVC). The analysis was implemented in a Linear Mixed Model (SPSS Inc., Chicago, IL) combined with the Bayesian Information Criteria (BIC) to achieve the best covariance structure. Covariates involved Fugl-Meyer score, Chedoke score, and the MVC ratio between the paretic and contralateral muscles. Post hoc test included pairwise comparisons and the Bonferroni correction. A family confidence coefficient of 0.95 was used to determine the significance.
3. Results Different patterns of power spectrum distribution were observed in paretic and contralateral muscles (upper part of Fig. 2). The spectral distribution functions corresponding to individual EMG power density distribution of the upper plots are displayed
991
on the lower part of the figure. All three subjects maintained 60% paretic MVC using the paretic and contralateral FDI muscles, respectively. In Fig. 2a, there is an increase of power density in the lower frequency portion of spectral distribution in the Subject 1’s hemiparetic muscle compared with his contralateral side. As a result, the mean power density frequency (MPF) of the paretic muscle (144.8 Hz) is smaller than that of the contralateral side (187.8 Hz). Calculation of the spectral distribution function indicated spectral compression in the paretic side with scaling factor around 0.7. The second subject’s surface EMG spectra were presented in the middle plot (Fig. 2b), where similar power density distributions were observed in the paretic muscle and the contralateral side. Computation of the MPF showed proximate values in the two muscles (paretic: 198 Hz; contralateral: 196.3 Hz). Likewise, the two spectral distribution curves were largely overlapped and no observance of spectral compression was found in this subject (scaling ratio: 1.01). In the third subject the paretic side (176.4 Hz) showed higher MFP value compared with the contralateral side (154.8 Hz) (Fig. 2c). Correspondingly, a spectral compression was found in the contralateral muscle (scaling factor: 1.2). In total, we found that 9 subjects showed increased MPF in the contralateral side (>15 Hz). No evident spectrum difference was observed in 3 subjects. Only 2 subjects had higher MPF in the paretic muscle than the contralateral muscle. Pooling all subjects’ data, we performed statistical analysis to examine the side and force effects on the spectrum characteristics of the surface EMG. The results indicated a significant reduction of MPF in the paretic muscle compared with the contralateral muscle (Fig. 3a, paretic: 168.7 ± 7.6 Hz, contralateral: 186.1 ± 8.7 Hz, mean ± standard error, F = 36.56, p < 0.001). A continuously lower average MPF was observed in the paretic muscles across all isometric voluntary force levels from 30% to 70% paretic MVC (Fig. 3b). However, the examination of the effects of contraction level on the surface EMG power spectrum did not disclose any significant correlation between the MPF and contraction force in either hand (F = 0.7, p > 0.5). Specifically, the averaged MPFs of the hemiparetic muscles at isometric contractions were: 167.6 ± 6.7 Hz (30% paretic MVC), 168.9 ± 7.3 Hz (40%), 165.2 ± 8.5 Hz (50%), 171.4 ± 8.5 Hz (60%), and 170.3 ± 7.7 Hz (70%). Additionally, the averaged MPFs of the contralateral side were: 190.8 ± 9.2 Hz (30% paretic MVC), 191.2 ± 9.4 Hz (40%), 182.2 ± 8.3 Hz (50%), 185.8 ± 9.1 Hz (60%), and 180.4 ± 8.5 Hz (70%) (Fig. 3b). Calculation of the averaged scaling factor across all subjects and force levels indicated spectral compression in the hemiparetic muscles (0.89 ± 0.04, mean ± standard error). With the force varying from
Fig. 2. Power spectral distribution of surface EMG at matched force from three subjects. (a) Subject 1, MPF: paretic side: 144.8 Hz; contralateral side: 187.8 Hz, scaling factor: 0.7. (b) Subject 12, MPF: paretic side: 198 Hz; contralateral side: 196.3 Hz, scaling factor: 1.01. (c) Subject 6, MPF: paretic side: 176.4 Hz; contralateral side: 154.8 Hz, scaling factor: 1.2.
992
X. Li et al. / Clinical Neurophysiology 125 (2014) 988–994
Fig. 3. (a) Averaged MPF of surface EMG in the paretic (black: 168.7 ± 7.6 Hz, mean ± standard error) and contralateral side (white: 186.1 ± 8.7 Hz) for all subjects. ⁄p < 0.05. (b) Averaged MPF of surface EMG at different contractions (from 30% to 70% paretic MVC). Paretic side: circle with solid line, 167.6 ± 6.7, 168.9 ± 7.3, 165.2 ± 8.5, 171.4 ± 8.5, and 170.3 ± 7.7 Hz. Contralateral side: square with dash line, 190.8 ± 9.2, 191.2 ± 9.4, 182.2 ± 8.3, 185.8 ± 9.1, and 180.4 ± 8.5 Hz.
30% to 70% paretic muscle MVC, the scaling factors increased slightly: 0.86 ± 0.04 (30%), 0.87 ± 0.06 (40%), 0.91 ± 0.05 (50%), 0.91 ± 0.05 (60%), and 0.94 ± 0.04 (70%). However, no significant correlation between the scaling factor and force increment was found (F = 2.08, p > 0.1). We also examined whether subjects’ clinical data, including Fugl-Meyer and Chedoke scores and the ratio of paretic and contralateral MVC, influenced the significant spectral alterations between the paretic and contralateral muscles. It was found that there was no correlation between spectrum difference and Fugl-Meyer or Chedoke scores, or ratio of paretic and contralateral MVC (p > 0.2). 4. Discussion The main finding of the current study is that different surface EMG power spectrum distributions were observed in paretic and contralateral muscles of stroke subjects. No correlation was observed between the spectral difference and the Fugl-Meyer or Chedoke scores, or ratio of paretic and contralateral muscle strength (p > 0.2). In addition, neither the paretic nor the contralateral FDI muscles showed a significant relationship between surface EMG spectrum and contraction levels. The findings of this study suggest that there appear to be complex muscular and neural processes at work post stroke that may impact the surface EMG power spectrum. Pooled analysis of all the tested stroke subjects showed reduced MPF of surface EMG power spectrum in the paretic muscles compared with the contralateral muscles at matched isometric contraction forces. Such pattern of spectrum alteration was also reported in previous studies (Gemperline et al., 1995; Muro et al., 1982). The reduced MPF of paretic muscles can be attributed to a variety of factors. For example, the increased synchronization strength of MUAPs was reported to occur in paretic muscles (Datta et al., 1991; Hausmanowa-Petrusewicz and Kopec, 1983), which may contribute to the MPF reduction of the power spectrum (Latash, 1988a,b; Muro et al., 1982). The increased lower frequency components in the paretic muscles could also be due to impairments in motor unit control properties. It was suggested that the EMG spectrum depends not only on the individual spike spectrum but also on the firing rate characteristics (De Luca, 1984; Lago and Jones, 1977; Weytjens and van Steenberghe, 1984). The firing rate of motor units was reported to be lower in paretic muscle compared with contralateral muscle (Gemperline et al., 1995; Hu et al., 2006; Rosenfalck and Andreassen, 1980), which may affect the spectrum at low frequency components. Both increased degree of motor unit
synchronization and reduction of motor unit firing rates may lead to enlarged EMG amplitude (Weytjens and van Steenberghe, 1984; Yao et al., 2000). At matched contraction levels, larger peak amplitudes on the paretic muscles was confirmed in our previous study from 9 of the 14 tested stroke subjects (Li et al., 2013). Since spectral parameters are closely correlated with motor unit depth and changes of motor unit conduction velocity (Broman et al., 1985; Lindstrom and Magnusson, 1977; Linnamo et al., 2001), the reduction of the MPF can be an indication of loss of large motor units in the paretic muscles. Previous studies suggested a selective degeneration of the large motor units following a stroke (Knight and Kamen, 2005; Lukacs et al., 2008). Those large motor units were reported to have higher conduction velocity and be located in more superficial regions of the muscles (Dengler et al., 1988; Knight and Kamen, 2005). Presumably, for the same force generation, the surface EMG signals derived from smaller motor units in relatively deep regions of the muscle will have lower amplitudes compared with those derived from larger motor units in the more superficial regions. Smaller peak amplitudes of surface EMG at matched forces were observed in the paretic muscles from 5 of the 14 tested stroke subjects (Li et al., 2013). In Fig. 3c, we also found that two of the tested stroke subjects had higher MPF in the paretic muscle than in the contralateral muscle at matched isometric force. This may be associated with muscle fiber reinnervation following spinal motoneuron degeneration post stroke. Increased muscle fiber density and neuromuscular jitter have been reported in stroke subjects as evidence of muscle fiber reinnervation (Chang, 1998; Lukacs et al., 2009). The reduction of the MPF for paretic muscles may also be due to atrophy of muscle fibers, especially the type II muscle fibers. In a study that applied supramaximal tetanic electrical stimulation to explore the alterations of muscle fiber composition in the hemiparetic limbs, stroke survivors, when compared with the healthy subjects, exhibited decreased spectral frequency in the initial stage but less reduction of median frequencies after 10 s of electrical stimulation (Toffola et al., 2001). The less spectrum reduction in the hemiparetic muscles in the fatigue tests indicated rearrangement of muscle composition after a stroke. There are few studies that examined the effects of increased isometric contractions on the spectral characteristics in stroke. In the current study we could not find a clear correlation between spectral features and contraction levels in either paretic or contralateral muscles. The finding is different from a previous report that observed a decreasing MPF in the paretic muscles over a 5–50% MVC force range and an increasing MPF in the contralateral
X. Li et al. / Clinical Neurophysiology 125 (2014) 988–994
muscles across the same force range (Kallenberg and Hermens, 2009). There are many factors that may contribute to the preceding differences. For example, different muscles were involved in the two studies, i.e. the previous report was based on the biceps brachii muscle whereas our results were obtained from the FDI muscles. It is known that biceps brachii muscles have a wide recruitment range up to 80% MVC, so force increment is managed mainly by the recruitment of more motor units. Conversely, the FDI muscles have narrower recruitment range. Therefore, force increase in the FDI muscle is achieved by adjusting the firing rate of active motor units. The differences in force enhancement strategy may influence the spectral characteristics of surface EMG. Additionally, the proximal and distal muscles demonstrate different pathological changes and recovery time post-stroke (Twitchell, 1951), which may also have an effect on the production of surface EMG. The different observations may be due to many other factors such as the recording property of the electrode, the contraction levels or types of muscle activation (ramp increment or step increment), limited number of subjects participated, variations of the experimental conditions (Kaplanis et al., 2009; Moritani and Muro, 1987). Indeed, even for healthy subjects different observations between the power spectrum and muscle contraction force have been reported. The increase of power spectrum with force was observed in healthy subjects performing continuous ramp contractions or constant contractions (Moritani and Muro, 1987; Muro et al., 1983). However, other studies reported a significant decrease of the spectra with force or no changes of mean or median frequencies with force using the same muscle group (Beck et al., 2006; Farina et al., 2002; Gabriel and Kamen, 2009; Kaplanis et al., 2009). Finally, we acknowledge that the interpretation of spectrum characteristics of surface EMG must additionally take into account any modifying factors other than physiological properties of the tested muscle (Broman et al., 1985; Dimitrov et al., 2003; Farina et al., 2002; Lowery et al., 2003). For example, the inter-electrode distance was reported to have a strong influence on the power spectrum distribution of surface EMG especially in the high frequency power content (Moritani and Muro, 1987; Zipp, 1978). It was suggested that larger inter-electrode distance leads to a lower spectrum bandwidth of surface EMG in bipolar recording. In the current study, the inter-electrode distance of the array sensor is 3.6 mm which is much smaller than conventional surface electrodes (10–20 mm). This may explain the relatively higher MPF of the surface EMG we collected from the paretic and contralateral muscles as compared with the frequencies of 70–150 Hz reported in the literature (Gabriel and Kamen, 2009; Kaplanis et al., 2009).
Acknowledgements The authors thank Nina Suresh, PhD for helping with data collection. This work was supported by Brinson Stroke Foundation, the National Institute on Disability and Rehabilitation Research of the U.S. Department of Education (Grant #: H133F110033) and the National Institutes of Health of the U.S. Department of Health and Human Services (Grant #: 2R24 HD050821).
References Beck TW, Housh TJ, Johnson GO, Cramer JT, Weir JP, Coburn JW, et al. Electromyographic instantaneous amplitude and instantaneous mean power frequency patterns across a range of motion during a concentric isokinetic muscle action of the biceps brachii. J Electromyogr Kinesiol 2006;16:531–9. Broman H, Bilotto G, De Luca CJ. Myoelectric signal conduction velocity and spectral parameters: influence of force and time. J Appl Physiol 1985;58:1428–37. Brown WF, Snow R. Denervation in hemiplegic muscles. Stroke 1990;21:1700–4.
993
Chang CW. Evident trans-synaptic degeneration of motor neurons after stroke: a study of neuromuscular jitter by axonal microstimulation. Electroencephalogr Clin Neurophysiol 1998;109:199–202. Charcot JM. Dégénerations secondaries de cause cérébrale. Paris: Bureaux du Progrès Medical; 1893. Datta AK, Farmer SF, Stephens JA. Central nervous pathways underlying synchronization of human motor unit firing studied during voluntary contractions. J Physiol 1991;432:401–25. Dattola R, Girlanda P, Vita G, Santoro M, Roberto ML, Toscano A, et al. Muscle rearrangement in patients with hemiparesis after stroke: an electrophysiological and morphological study. Eur Neurol 1993;33:109–14. De Luca CJ. Myoelectrical manifestations of localized muscular fatigue in humans. Crit Rev Biomed Eng 1984;11:251–79. De Luca CJ, Hostage EC. Relationship between firing rate and recruitment threshold of motoneurons in voluntary isometric contractions. J Neurophysiol 2010;104:1034–46. Dengler R, Stein RB, Thomas CK. Axonal conduction velocity and force of single human motor units. Muscle Nerve 1988;11:136–45. Dimitrov GV, Arabadzhiev TI, Hogrel JY, Dimitrova NA. Simulation analysis of interference EMG during fatiguing voluntary contractions. Part II – changes in amplitude and spectral characteristics. J Electromyogr Kinesiol 2008;18:35–43. Dimitrov GV, Disselhorst-Klug C, Dimitrova NA, Schulte E, Rau G. Simulation analysis of the ability of different types of multi-electrodes to increase selectivity of detection and to reduce cross-talk. J Electromyogr Kinesiol 2003;13:125–38. Farina D, Fosci M, Merletti R. Motor unit recruitment strategies investigated by surface EMG variables. J Appl Physiol 2002;92:235–47. Farina D, Merletti R, Enoka RM. The extraction of neural strategies from the surface EMG. J Appl Physiol 2004;96:1486–95. Farmer SF, Swash M, Ingram DA, Stephens JA. Changes in motor unit synchronization following central nervous lesions in man. J Physiol 1993;463:83–105. Gabriel DA, Kamen G. Experimental and modeling investigation of spectral compression of biceps brachii SEMG activity with increasing force levels. J Electromyogr Kinesiol 2009;19:437–48. Gemperline JJ, Allen S, Walk D, Rymer WZ. Characteristics of motor unit discharge in subjects with hemiparesis. Muscle Nerve 1995;18:1101–14. Hara Y, Masakado Y, Chino N. The physiological functional loss of single thenar motor units in the stroke patients: when does it occur? Does it progress? Clin Neurophysiol 2004;115:97–103. Hausmanowa-Petrusewicz I, Kopec J. EMG parameters changes in the effort pattern at various loads in diseased muscle. Electromyogr Clin Neurophysiol 1983;23:213–28. Hu XL, Tong KY, Hung LK. Firing properties of motor units during fatigue in subjects after stroke. J Electromyogr Kinesiol 2006;16:469–76. Kallenberg LA, Hermens HJ. Motor unit properties of biceps brachii in chronic stroke patients assessed with high-density surface EMG. Muscle Nerve 2009;39:177–85. Kaplanis PA, Pattichis CS, Hadjileontiadis LJ, Roberts VC. Surface EMG analysis on normal subjects based on isometric voluntary contraction. J Electromyogr Kinesiol 2009;19:157–71. Knight CA, Kamen G. Superficial motor units are larger than deeper motor units in human vastus lateralis muscle. Muscle Nerve 2005;31:475–80. Lago P, Jones NB. Effect of motor-unit firing time statistics on e.m.g. spectra. Med Biol Eng Comput 1977;15:648–55. Latash ML. Spectral analysis of the electromyogram (EMG) in spinal cord trauma patients: I. Different types of the EMG and corresponding spectra. Electromyogr Clin Neurophysiol 1988a;28:319–27. Latash ML. Spectral analysis of the electromyogram (EMG) in spinal cord trauma patients: II. Motor unit and interference EMG power spectra. Electromyogr Clin Neurophysiol 1988b;28:329–34. Li X, Suresh A, Zhou P, Rymer W. Alterations in the spike amplitude distribution of the surface electromyogram post-stroke. IEEE Trans Biomed Eng 2013;60:845–52. Lindstrom L, Magnusson R. Interpretation of myoelectric power spectra: a model and its applications. Proc IEEE 1977;65:653–62. Lindstrom L, Magnusson R, Petersen I. Muscular fatigue and action potential conduction velocity changes studied with frequency analysis of EMG signals. Electromyography 1970;10:341–56. cLinnamo V, Strojnik V, Komi PV. Electromyogram power spectrum and features of the superimposed maximal M-wave during voluntary isometric actions in humans at different activation levels. Eur J Appl Physiol 2001;86:28–33. Lowery MM, Stoykov NS, Kuiken TA. A simulation study to examine the use of crosscorrelation as an estimate of surface EMG cross talk. J Appl Physiol 2003;94:1324–34. Lowery MM, Vaughan CL, Nolan PJ, O’Malley MJ. Spectral compression of the electromyographic signal due to decreasing muscle fiber conduction velocity. IEEE Trans Rehabil Eng 2000;8:353–61. Lukacs M, Vecsei L, Beniczky S. Large motor units are selectively affected following a stroke. Clin Neurophysiol 2008;119:2555–8. Lukacs M, Vecsei L, Beniczky S. Changes in muscle fiber density following a stroke. Clin Neurophysiol 2009;120:1539–42. Merletti R, Lo Conte LR. Surface EMG signal processing during isometric contractions. J Electromyogr Kinesiol 1997;7:241–50.
994
X. Li et al. / Clinical Neurophysiology 125 (2014) 988–994
Moritani T, Muro M. Motor unit activity and surface electromyogram power spectrum during increasing force of contraction. Eur J Appl Physiol Occup Physiol 1987;56:260–5. Muro M, Nagata A, Moritani T. Analysis of myoelectric signals during dynamic and isometric contractions. In: Matsui H, Kobayashi K, editors. Biomechanics VIIIA. Human Kinetics Publishers; 1983. p. 432–9. Muro M, Nagata A, Murakami K, Moritani T. Surface EMG power spectral analysis of neuromuscular disorders during isometric and isotonic contractions. Am J Phys Med 1982;61:244–54. Petrofsky JS, Lind AR. The influence of temperature on the amplitude and frequency components of the EMG during brief and sustained isometric contractions. Eur J Appl Physiol Occup Physiol 1980;44:189–200. Rainoldi A, Galardi G, Maderna L, Comi G, Lo Conte L, Merletti R. Repeatability of surface EMG variables during voluntary isometric contractions of the biceps brachii muscle. J Electromyogr Kinesiol 1999;9:105–19. Rix H, Malenge JP. Detecting small variations in shape. IEEE Trans Syst Man Cybern 1980;10:90–6. Rosenfalck A, Andreassen S. Impaired regulation of force and firing pattern of single motor units in patients with spasticity. J Neurol Neurosurg Psychiatry 1980;43:907–16. Seki K, Miyazaki Y, Watanabe M, Nagata A, Narusawa M. Surface electromyogram spectral characterization and motor unit activity during voluntary ramp contraction in men. Eur J Appl Physiol Occup Physiol 1991;63:165–72.
Solomonow M, Baten C, Smit J, Baratta R, Hermens H, D’Ambrosia R, et al. Electromyogram power spectra frequencies associated with motor unit recruitment strategies. J Appl Physiol 1990;68:1177–85. Spaans F, Wilts G. Denervation due to lesions of the central nervous system. An EMG study in cases of cerebral contusion and cerebrovascular accidents. J Neurol Sci 1982;57:291–305. Toffola ED, Sparpaglione D, Pistorio A, Buonocore M. Myoelectric manifestations of muscle changes in stroke patients. Arch Phys Med Rehabil 2001;82:661–5. Twitchell TE. The restoration of motor function following hemiplegia in man. Brain 1951;74:443–80. Wakeling JM. Patterns of motor recruitment can be determined using surface EMG. J Electromyogr Kinesiol 2009;19:199–207. Weytjens JL, van Steenberghe D. The effects of motor unit synchronization on the power spectrum of the electromyogram. Biol Cybern 1984;51:71–7. Yao W, Fuglevand RJ, Enoka RM. Motor-unit synchronization increases EMG amplitude and decreases force steadiness of simulated contractions. J Neurophysiol 2000;83:441–52. Zaman SA, Macisaac DT, Parker PA. Repeatability of surface EMG-based single parameter muscle fatigue assessment strategies in static and cyclic contractions. Conf Proc IEEE Eng Med Biol Soc 2011;2011:3857–60. Zipp P. Effect of electrode parameters on the bandwidth of the surface e.m.g. powerdensity spectrum. Med Biol Eng Comput 1978;16:537–41.
Contents lists available at ScienceDirect
Clinical Neurophysiology journal homepage: www.elsevier.com/locate/clinph
Power spectral analysis of surface electromyography (EMG) at matched contraction levels of the first dorsal interosseous muscle in stroke survivors Xiaoyan Li a,⇑, Henry Shin b, Ping Zhou a,c,d, Xun Niu a, Jie Liu a, William Zev Rymer a,b,c a
Sensory Motor Performance Program, Rehabilitation Institute of Chicago, Chicago, IL, USA Department of Biomedical Engineering, Northwestern University, Chicago, IL, USA Department of Physical Medicine and Rehabilitation, Northwestern University, Chicago, IL, USA d Institute of Biomedical Engineering, University of Science and Technology of China, Hefei, China b c
a r t i c l e
i n f o
Article history: Accepted 30 September 2013 Available online 21 November 2013 Keywords: Surface electromyography (EMG) Stroke Spectral analysis First dorsal interosseous (FDI) muscle
h i g h l i g h t s Spectral analysis of surface EMG post stroke was performed in FDI muscles. Subjects showed reduced mean frequency in paretic side at matched forces. Spectral difference between two sides was not correlated to clinical scales.
a b s t r a c t Objective: The objective of this study was to help assess complex neural and muscular changes induced by stroke using power spectral analysis of surface electromyogram (EMG) signals. Methods: Fourteen stroke subjects participated in the study. They were instructed to perform isometric voluntary contractions by abducting the index finger. Surface EMG signals were collected from the paretic and contralateral first dorsal interosseous (FDI) muscles with forces ranging from 30% to 70% maximum voluntary contraction (MVC) of the paretic muscle. Power spectral analysis was performed to characterize features of the surface EMG in paretic and contralateral muscles at matched forces. A Linear Mixed Model was applied to identify the spectral changes in the hemiparetic muscle and to examine the relation between spectral parameters and contraction levels. Regression analysis was performed to examine the correlations between spectral characteristics and clinical features. Results: Differences in power spectrum distribution patterns were observed in paretic muscles when compared with their contralateral pairs. Nine subjects showed increased mean power frequency (MPF) in the contralateral side (>15 Hz). No evident spectrum difference was observed in 3 subjects. Only 2 subjects had higher MPF in the paretic muscle than the contralateral muscle. Pooling all subjects’ data, there was a significant reduction of MPF in the paretic muscle compared with the contralateral muscle (paretic: 168.7 ± 7.6 Hz, contralateral: 186.1 ± 8.7 Hz, mean ± standard error, F = 36.56, p < 0.001). Examination of force factor on the surface EMG power spectrum did not confirm a significant correlation between the MPF and contraction force in either hand (F = 0.7, p > 0.5). There was no correlation between spectrum difference and Fugl–Meyer or Chedoke scores, or ratio of paretic and contralateral MVC (p > 0.2). Conclusions: There appears to be complex muscular and neural processes at work post stroke that may impact the surface EMG power spectrum. The majority of the tested stroke subjects had lower MPF in the paretic muscle than in the contralateral muscle at matched isometric contraction force. The reduced MPF of paretic muscles can be attributed to different factors such as increased motor unit synchronization, impairments in motor unit control properties, loss of large motor units, and atrophy of muscle fibers. Significance: Surface EMG power spectral analysis can serve as a useful tool to indicate complex neural and muscular changes after stroke. Ó 2013 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved.
⇑ Corresponding author. Address: Sensory Motor Performance Program, Rehabilitation Institute of Chicago, 345 E Superior St, Suite 1406, Chicago, IL, USA. Tel.: +1 312 238 1174. E-mail address: [email protected] (X. Li). 1388-2457/$36.00 Ó 2013 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.clinph.2013.09.044
X. Li et al. / Clinical Neurophysiology 125 (2014) 988–994
1. Introduction Spectral analysis of electromyogram (EMG) signals has been used for decades to reveal pathophysiology of neuromuscular impairment (Dimitrov et al., 2008; Farina et al., 2004; Kaplanis et al., 2009; Lindstrom and Magnusson, 1977; Muro et al., 1982). Changes in spectral frequency of surface EMG are associated with peripheral factors, such as muscle fiber conduction velocity, muscle fiber length and orientation, motor unit location, intracellular action potential shape and its negative after-potentials (Dimitrov et al., 2008; Farina et al., 2004; Lindstrom et al., 1970; Lowery et al., 2000; Zaman et al., 2011). Modifications of the spectral characteristics are also related to synchronization of multiple motor units (mainly associated with central nervous system) or motor unit recruitment and firing rate changes (associated with both central nerve system and motor unit properties) (Gabriel and Kamen, 2009; Lago and Jones, 1977; Solomonow et al., 1990; Wakeling, 2009; Yao et al., 2000). Other non-physiological factors such as temperature, thickness of the subcutaneous layers, electrode position and configuration, etc. may also influence the EMG spectral distribution (Moritani and Muro, 1987; Petrofsky and Lind, 1980; Zipp, 1978). Subsequent to a cerebral lesion, progressive peripheral changes in hemiparetic muscles have been reported during the course of the disease involving loss of muscle fibers, changes of motor unit type composition, loss of functioning motor units, or structural reorganization of survival motor units (Brown and Snow, 1990; Charcot, 1893; Dattola et al., 1993; Hara et al., 2004; Spaans and Wilts, 1982). Meanwhile, disorganization of motor unit control properties have been observed in the hemiparetic muscles described as compressed motor unit recruitment range, abnormally low motor unit discharge rate (Gemperline et al., 1995; Rosenfalck and Andreassen, 1980), as well as increased intensity of motor unit synchronization (Farmer et al., 1993). Presently, there are few studies documenting the impact of post stroke neuromuscular modifications on the spectral characteristics of the surface EMG. In one study an increase of power density in the lower spectral frequency section was observed in one subject’s paretic biceps brachii muscle whereas no significant spectrum difference was found between paretic and contralateral muscles in the remaining five subjects (Gemperline et al., 1995). The spectral analysis was also used to characterize individual motor unit action potentials (MUAPs) which demonstrated lower mean power frequency in the paretic side compared with the contralateral side (Kallenberg and Hermens, 2009). However, there was no significant difference of the mean power frequency in the global surface EMG analysis between two sides in the same group of subjects. Another question that has not been intensively explored in stroke is the relation between the EMG power spectrum and the contraction force. Previous investigations of such relation in healthy subjects have provided contradictory observations (Farina et al., 2002; Kaplanis et al., 2009; Rainoldi et al., 1999; Seki et al., 1991). For example, it was reported that the mean power frequency (MPF) or median frequency (MF) values of surface EMG continuously increase with isometric force up to 80% maximal voluntary contraction (MVC) (Kaplanis et al., 2009; Moritani and Muro, 1987). Conversely, decreased spectrum frequency was also observed with the increment of contraction levels (Gabriel and Kamen, 2009; Rainoldi et al., 1999). In stoke there was only one study investigating the relation of surface EMG spectrum and force, which reported a reduction of median frequency in the hemiparetic muscles and a slight increase of median frequency in the contralateral side along force (Kallenberg and Hermens, 2009).
989
In light of the few and inconsistent observations on post stroke power spectral changes of surface EMG, the objective of the current study was to evaluate the spectral characteristics of the surface EMG in stroke survivors. Particularly, we compared power spectrum patterns of surface EMG between paretic and contralateral muscles during sustained voluntary contractions. We further examined the association between surface EMG spectral characteristics and isometric muscle contraction levels. 2. Methods 2.1. Subjects Fourteen chronic stroke survivors (6 females and 8 males, aged 45–70 years old) with mild to severe weakness in the contralesional side participated in the study. All subjects were screened by a physician based on clinical history and physical examination. Subjects with concurrent neurological disorders or other symptoms (such as neuropathy, radiculopathy, cervical spondylosis, and hyperglycemia, etc.) were excluded. No subjects reported any arm pain, numbness or paresthesia. The disease duration since the onset of the stroke ranged from 1 year 6 months to 24 years. All subjects submitted written consent approved by the Institutional Review Board of Northwestern University (Chicago, IL, USA) before experiments. Their motor function was evaluated based on the Fugl-Meyer test and the Chedoke-McMaster test. Maximal voluntary contraction (MVC) force was additionally measured from the first dorsal interosseous muscles of the paretic and contralateral hands, respectively. The MVC ratio between paretic MVC and contralateral MVC was used as an index of relative weakness (Gemperline et al., 1995). A summary of subjects’ information is presented in Table 1. 2.2. Experiments Subjects were seated comfortably in a mobile Biodex chair. Their forearm and wrist were casted and positioned on a plastic platform, where the wrist was fastened inside a ring-mount interface. This setup kept the wrist in a pronated position and minimized its movement (Fig. 1a). The index finger was immobilized in a fiberglass cast and placed inside a small ring interface that was attached to a six degree-of-freedom load cell (ATI, Apex, NC). The position of the index finger was aligned with the center of the load cell. Surface EMG signals were recorded from a four-channel sensor array (Delsys, Boston, MA) dwelling on top of the first dorsal interosseous (FDI) muscle. The configuration of the sensor is illustrated in Fig. 1b, as five pin electrodes in a square, one in each corner and the fifth in the center. The inter-electrode distance between two neighboring corner electrodes is 5 mm. Each pin electrode is 0.5 mm in diameter. Details of the sensor can be found in (De Luca and Hostage, 2010). Since it was difficult to align all the four bipolar recording pairs of pins (i.e. the center pin with respect to each of the four corner pins) with the muscle fiber orientation, placement of the sensory array was arranged in a way that three electrode-pins (two diagonal pins and the center pin) were in parallel with the orientation of the muscle fibers. This assures the same alignment of the electrode with respect to the FDI muscle. The output differential signals were filtered with a bandpass filter of 20–2000 Hz. Voluntary contraction force exerted by the index finger was measured in abduction (Fx), flexion (Fy), and the anterior–posterior direction (Fz) in the sagittal plane. All forces and surface EMG signals were sampled at 20,000 Hz using EMGWorksÒ (Delsys, Boston, MA).
990
X. Li et al. / Clinical Neurophysiology 125 (2014) 988–994 Table 1 Stroke subjects’ information. ID
Sex
Age (years)
Duration (years)
Paretic side
Chedoke score
FM
MVC ratio
1 2 3 4 5 6 7 8 9 10 11 12 13 14
M M F M M M M F F F F M M F
62.2 45.2 52.3 61.8 56.8 66.8 70.0 68.2 44.9 59.5 59.7 48.7 61.4 58.6
1.7 5.7 6.8 7.8 8.0 8.7 1.5 5.0 13.7 19.7 24.1 1.7 5.0 5.5
Left Right Right Right Right Left Left Right Left Left Right Right Right Right
7 5 5 4 5 4 7 5 4 3 2 6 6 4
66 58 58 56 51 24 56 53 16 29 22 48 63 40
0.75 0.60 0.66 0.19 0.82 0.27 0.77 0.28 0.17 0.36 0.52 0.61 0.85 0.87
FM: Fugl-Meyer test; MVC ratio: MVC paretic/MVC contralateral.
Fig. 1. (a) Experimental setup; (b) configuration of the array sensor (Delsys, Boston, MA) and (c) an example of a typical trial includes a rest period (baseline), a rising ramp, and a constant force period.
In the beginning of the experiment, the subjects were required to perform isometric maximum voluntary contractions (MVC) three times using the paretic index finger in the direction of abduction. The MVC of the paretic FDI muscle was determined as the highest value of the three efforts, and then this value was used to determine the desired force levels for both paretic and contralateral muscles of the subjects. The magnitude of desired force levels was set as 30–70% of the paretic muscle MVC in 10% MVC increments. For each desired force level, the duration of a typical trial included a rest period of 3 s, two ramp periods when the abduction force inclined or declined at a speed of 10% paretic muscle MVC per second, and a sustained force period for at least 10 s (Fig. 1c). Depending on the magnitude of the desired force levels, the duration of a trial varied from 20 to 30 s. Following a brief rest, the subjects were instructed to generate isometric contractions to match the magnitude of desired force using the index finger. The magnitude of a desired force and the contraction force generated by the subjects were displayed as real time trajectories in different colors. Such force feedback helped subjects adjust their contraction force to match the desired force level. Throughout the experiment, the subjects were encouraged to maintain the abduction force as stable as possible and minimize the flexion force simultaneously. At least
two successful trials at each contraction level were obtained before advancing to the next contraction level. The order of trials at different force levels was randomized. Practice was given to help them become familiarized with the task at each force level. A brief rest period was provided between trials to prevent potential fatigue. Subjects were scheduled for the experiments on separate days for the paretic and contralateral muscles. The experiment on the paretic muscle was always performed first so that the desired force levels could be determined prior to the contralateral muscle experiment. In addition to desired force levels, the MVC of the contralateral muscle was also recorded. 2.3. Data analysis Offline analysis was performed on the surface EMG signals as well as the abduction (Fx) and flexion (Fy) forces. We averaged the abduction and flexion forces within the sustained force period to calculate the flexion deviation, defined as the inverse tangent of the ratio between flexion and abduction forces (Fy/Fx). Trials with flexion deviation larger than 30° were excluded from analysis. Next, the signal to noise ratio (SNR) of surface EMG was examined for each channel. Spectral analysis was conducted on the channel
X. Li et al. / Clinical Neurophysiology 125 (2014) 988–994
that had the largest SNR values. The pair of pins that demonstrated the largest SNR of the differential signal was the one arranged along the muscle fibers. This was consistent for all the recordings. The Hamming window was used to smooth the signal before the periodogram method was applied to estimate the power spectrum of the surface EMG signals. The frequency resolution was 2 Hz. The mean power frequency (MPF) was calculated for each trial and then averaged over trials at the same level. Comparisons were made between the paretic and contralateral sides at the same force levels (i.e. matched contractions). The spectrum difference between the paretic and contralateral muscles was further analyzed by employing the spectral distribution function to detect small variations between two waveforms of the power spectra distribution from paretic and contralateral muscles. The spectral distribution function was defined as the normalized integral of the power spectrum (Lowery et al., 2000; Merletti and Lo Conte, 1997; Rix and Malenge, 1980). At a given distribution percentile, the frequency value in the hemiparetic muscle (fp in Fig. 2) could be different from that in the contralateral muscle (fcon in Fig. 2). Thus, a scaling factor was computed as the mean ratio of fp/fcon to indicate frequency compression. This technique calculated the shifts of spectral distribution across all percentile frequencies of the entire waveform and provided a more accurate index for the estimation of muscle fiber conduction velocity compared with the use of the median frequency of the power spectrum (Lowery et al., 2000). The following two main effects on the surface EMG spectral characteristics were investigated: (1) side effect (paretic muscle vs. contralateral side), and (2) force effect (from 30% to 70% paretic MVC). The analysis was implemented in a Linear Mixed Model (SPSS Inc., Chicago, IL) combined with the Bayesian Information Criteria (BIC) to achieve the best covariance structure. Covariates involved Fugl-Meyer score, Chedoke score, and the MVC ratio between the paretic and contralateral muscles. Post hoc test included pairwise comparisons and the Bonferroni correction. A family confidence coefficient of 0.95 was used to determine the significance.
3. Results Different patterns of power spectrum distribution were observed in paretic and contralateral muscles (upper part of Fig. 2). The spectral distribution functions corresponding to individual EMG power density distribution of the upper plots are displayed
991
on the lower part of the figure. All three subjects maintained 60% paretic MVC using the paretic and contralateral FDI muscles, respectively. In Fig. 2a, there is an increase of power density in the lower frequency portion of spectral distribution in the Subject 1’s hemiparetic muscle compared with his contralateral side. As a result, the mean power density frequency (MPF) of the paretic muscle (144.8 Hz) is smaller than that of the contralateral side (187.8 Hz). Calculation of the spectral distribution function indicated spectral compression in the paretic side with scaling factor around 0.7. The second subject’s surface EMG spectra were presented in the middle plot (Fig. 2b), where similar power density distributions were observed in the paretic muscle and the contralateral side. Computation of the MPF showed proximate values in the two muscles (paretic: 198 Hz; contralateral: 196.3 Hz). Likewise, the two spectral distribution curves were largely overlapped and no observance of spectral compression was found in this subject (scaling ratio: 1.01). In the third subject the paretic side (176.4 Hz) showed higher MFP value compared with the contralateral side (154.8 Hz) (Fig. 2c). Correspondingly, a spectral compression was found in the contralateral muscle (scaling factor: 1.2). In total, we found that 9 subjects showed increased MPF in the contralateral side (>15 Hz). No evident spectrum difference was observed in 3 subjects. Only 2 subjects had higher MPF in the paretic muscle than the contralateral muscle. Pooling all subjects’ data, we performed statistical analysis to examine the side and force effects on the spectrum characteristics of the surface EMG. The results indicated a significant reduction of MPF in the paretic muscle compared with the contralateral muscle (Fig. 3a, paretic: 168.7 ± 7.6 Hz, contralateral: 186.1 ± 8.7 Hz, mean ± standard error, F = 36.56, p < 0.001). A continuously lower average MPF was observed in the paretic muscles across all isometric voluntary force levels from 30% to 70% paretic MVC (Fig. 3b). However, the examination of the effects of contraction level on the surface EMG power spectrum did not disclose any significant correlation between the MPF and contraction force in either hand (F = 0.7, p > 0.5). Specifically, the averaged MPFs of the hemiparetic muscles at isometric contractions were: 167.6 ± 6.7 Hz (30% paretic MVC), 168.9 ± 7.3 Hz (40%), 165.2 ± 8.5 Hz (50%), 171.4 ± 8.5 Hz (60%), and 170.3 ± 7.7 Hz (70%). Additionally, the averaged MPFs of the contralateral side were: 190.8 ± 9.2 Hz (30% paretic MVC), 191.2 ± 9.4 Hz (40%), 182.2 ± 8.3 Hz (50%), 185.8 ± 9.1 Hz (60%), and 180.4 ± 8.5 Hz (70%) (Fig. 3b). Calculation of the averaged scaling factor across all subjects and force levels indicated spectral compression in the hemiparetic muscles (0.89 ± 0.04, mean ± standard error). With the force varying from
Fig. 2. Power spectral distribution of surface EMG at matched force from three subjects. (a) Subject 1, MPF: paretic side: 144.8 Hz; contralateral side: 187.8 Hz, scaling factor: 0.7. (b) Subject 12, MPF: paretic side: 198 Hz; contralateral side: 196.3 Hz, scaling factor: 1.01. (c) Subject 6, MPF: paretic side: 176.4 Hz; contralateral side: 154.8 Hz, scaling factor: 1.2.
992
X. Li et al. / Clinical Neurophysiology 125 (2014) 988–994
Fig. 3. (a) Averaged MPF of surface EMG in the paretic (black: 168.7 ± 7.6 Hz, mean ± standard error) and contralateral side (white: 186.1 ± 8.7 Hz) for all subjects. ⁄p < 0.05. (b) Averaged MPF of surface EMG at different contractions (from 30% to 70% paretic MVC). Paretic side: circle with solid line, 167.6 ± 6.7, 168.9 ± 7.3, 165.2 ± 8.5, 171.4 ± 8.5, and 170.3 ± 7.7 Hz. Contralateral side: square with dash line, 190.8 ± 9.2, 191.2 ± 9.4, 182.2 ± 8.3, 185.8 ± 9.1, and 180.4 ± 8.5 Hz.
30% to 70% paretic muscle MVC, the scaling factors increased slightly: 0.86 ± 0.04 (30%), 0.87 ± 0.06 (40%), 0.91 ± 0.05 (50%), 0.91 ± 0.05 (60%), and 0.94 ± 0.04 (70%). However, no significant correlation between the scaling factor and force increment was found (F = 2.08, p > 0.1). We also examined whether subjects’ clinical data, including Fugl-Meyer and Chedoke scores and the ratio of paretic and contralateral MVC, influenced the significant spectral alterations between the paretic and contralateral muscles. It was found that there was no correlation between spectrum difference and Fugl-Meyer or Chedoke scores, or ratio of paretic and contralateral MVC (p > 0.2). 4. Discussion The main finding of the current study is that different surface EMG power spectrum distributions were observed in paretic and contralateral muscles of stroke subjects. No correlation was observed between the spectral difference and the Fugl-Meyer or Chedoke scores, or ratio of paretic and contralateral muscle strength (p > 0.2). In addition, neither the paretic nor the contralateral FDI muscles showed a significant relationship between surface EMG spectrum and contraction levels. The findings of this study suggest that there appear to be complex muscular and neural processes at work post stroke that may impact the surface EMG power spectrum. Pooled analysis of all the tested stroke subjects showed reduced MPF of surface EMG power spectrum in the paretic muscles compared with the contralateral muscles at matched isometric contraction forces. Such pattern of spectrum alteration was also reported in previous studies (Gemperline et al., 1995; Muro et al., 1982). The reduced MPF of paretic muscles can be attributed to a variety of factors. For example, the increased synchronization strength of MUAPs was reported to occur in paretic muscles (Datta et al., 1991; Hausmanowa-Petrusewicz and Kopec, 1983), which may contribute to the MPF reduction of the power spectrum (Latash, 1988a,b; Muro et al., 1982). The increased lower frequency components in the paretic muscles could also be due to impairments in motor unit control properties. It was suggested that the EMG spectrum depends not only on the individual spike spectrum but also on the firing rate characteristics (De Luca, 1984; Lago and Jones, 1977; Weytjens and van Steenberghe, 1984). The firing rate of motor units was reported to be lower in paretic muscle compared with contralateral muscle (Gemperline et al., 1995; Hu et al., 2006; Rosenfalck and Andreassen, 1980), which may affect the spectrum at low frequency components. Both increased degree of motor unit
synchronization and reduction of motor unit firing rates may lead to enlarged EMG amplitude (Weytjens and van Steenberghe, 1984; Yao et al., 2000). At matched contraction levels, larger peak amplitudes on the paretic muscles was confirmed in our previous study from 9 of the 14 tested stroke subjects (Li et al., 2013). Since spectral parameters are closely correlated with motor unit depth and changes of motor unit conduction velocity (Broman et al., 1985; Lindstrom and Magnusson, 1977; Linnamo et al., 2001), the reduction of the MPF can be an indication of loss of large motor units in the paretic muscles. Previous studies suggested a selective degeneration of the large motor units following a stroke (Knight and Kamen, 2005; Lukacs et al., 2008). Those large motor units were reported to have higher conduction velocity and be located in more superficial regions of the muscles (Dengler et al., 1988; Knight and Kamen, 2005). Presumably, for the same force generation, the surface EMG signals derived from smaller motor units in relatively deep regions of the muscle will have lower amplitudes compared with those derived from larger motor units in the more superficial regions. Smaller peak amplitudes of surface EMG at matched forces were observed in the paretic muscles from 5 of the 14 tested stroke subjects (Li et al., 2013). In Fig. 3c, we also found that two of the tested stroke subjects had higher MPF in the paretic muscle than in the contralateral muscle at matched isometric force. This may be associated with muscle fiber reinnervation following spinal motoneuron degeneration post stroke. Increased muscle fiber density and neuromuscular jitter have been reported in stroke subjects as evidence of muscle fiber reinnervation (Chang, 1998; Lukacs et al., 2009). The reduction of the MPF for paretic muscles may also be due to atrophy of muscle fibers, especially the type II muscle fibers. In a study that applied supramaximal tetanic electrical stimulation to explore the alterations of muscle fiber composition in the hemiparetic limbs, stroke survivors, when compared with the healthy subjects, exhibited decreased spectral frequency in the initial stage but less reduction of median frequencies after 10 s of electrical stimulation (Toffola et al., 2001). The less spectrum reduction in the hemiparetic muscles in the fatigue tests indicated rearrangement of muscle composition after a stroke. There are few studies that examined the effects of increased isometric contractions on the spectral characteristics in stroke. In the current study we could not find a clear correlation between spectral features and contraction levels in either paretic or contralateral muscles. The finding is different from a previous report that observed a decreasing MPF in the paretic muscles over a 5–50% MVC force range and an increasing MPF in the contralateral
X. Li et al. / Clinical Neurophysiology 125 (2014) 988–994
muscles across the same force range (Kallenberg and Hermens, 2009). There are many factors that may contribute to the preceding differences. For example, different muscles were involved in the two studies, i.e. the previous report was based on the biceps brachii muscle whereas our results were obtained from the FDI muscles. It is known that biceps brachii muscles have a wide recruitment range up to 80% MVC, so force increment is managed mainly by the recruitment of more motor units. Conversely, the FDI muscles have narrower recruitment range. Therefore, force increase in the FDI muscle is achieved by adjusting the firing rate of active motor units. The differences in force enhancement strategy may influence the spectral characteristics of surface EMG. Additionally, the proximal and distal muscles demonstrate different pathological changes and recovery time post-stroke (Twitchell, 1951), which may also have an effect on the production of surface EMG. The different observations may be due to many other factors such as the recording property of the electrode, the contraction levels or types of muscle activation (ramp increment or step increment), limited number of subjects participated, variations of the experimental conditions (Kaplanis et al., 2009; Moritani and Muro, 1987). Indeed, even for healthy subjects different observations between the power spectrum and muscle contraction force have been reported. The increase of power spectrum with force was observed in healthy subjects performing continuous ramp contractions or constant contractions (Moritani and Muro, 1987; Muro et al., 1983). However, other studies reported a significant decrease of the spectra with force or no changes of mean or median frequencies with force using the same muscle group (Beck et al., 2006; Farina et al., 2002; Gabriel and Kamen, 2009; Kaplanis et al., 2009). Finally, we acknowledge that the interpretation of spectrum characteristics of surface EMG must additionally take into account any modifying factors other than physiological properties of the tested muscle (Broman et al., 1985; Dimitrov et al., 2003; Farina et al., 2002; Lowery et al., 2003). For example, the inter-electrode distance was reported to have a strong influence on the power spectrum distribution of surface EMG especially in the high frequency power content (Moritani and Muro, 1987; Zipp, 1978). It was suggested that larger inter-electrode distance leads to a lower spectrum bandwidth of surface EMG in bipolar recording. In the current study, the inter-electrode distance of the array sensor is 3.6 mm which is much smaller than conventional surface electrodes (10–20 mm). This may explain the relatively higher MPF of the surface EMG we collected from the paretic and contralateral muscles as compared with the frequencies of 70–150 Hz reported in the literature (Gabriel and Kamen, 2009; Kaplanis et al., 2009).
Acknowledgements The authors thank Nina Suresh, PhD for helping with data collection. This work was supported by Brinson Stroke Foundation, the National Institute on Disability and Rehabilitation Research of the U.S. Department of Education (Grant #: H133F110033) and the National Institutes of Health of the U.S. Department of Health and Human Services (Grant #: 2R24 HD050821).
References Beck TW, Housh TJ, Johnson GO, Cramer JT, Weir JP, Coburn JW, et al. Electromyographic instantaneous amplitude and instantaneous mean power frequency patterns across a range of motion during a concentric isokinetic muscle action of the biceps brachii. J Electromyogr Kinesiol 2006;16:531–9. Broman H, Bilotto G, De Luca CJ. Myoelectric signal conduction velocity and spectral parameters: influence of force and time. J Appl Physiol 1985;58:1428–37. Brown WF, Snow R. Denervation in hemiplegic muscles. Stroke 1990;21:1700–4.
993
Chang CW. Evident trans-synaptic degeneration of motor neurons after stroke: a study of neuromuscular jitter by axonal microstimulation. Electroencephalogr Clin Neurophysiol 1998;109:199–202. Charcot JM. Dégénerations secondaries de cause cérébrale. Paris: Bureaux du Progrès Medical; 1893. Datta AK, Farmer SF, Stephens JA. Central nervous pathways underlying synchronization of human motor unit firing studied during voluntary contractions. J Physiol 1991;432:401–25. Dattola R, Girlanda P, Vita G, Santoro M, Roberto ML, Toscano A, et al. Muscle rearrangement in patients with hemiparesis after stroke: an electrophysiological and morphological study. Eur Neurol 1993;33:109–14. De Luca CJ. Myoelectrical manifestations of localized muscular fatigue in humans. Crit Rev Biomed Eng 1984;11:251–79. De Luca CJ, Hostage EC. Relationship between firing rate and recruitment threshold of motoneurons in voluntary isometric contractions. J Neurophysiol 2010;104:1034–46. Dengler R, Stein RB, Thomas CK. Axonal conduction velocity and force of single human motor units. Muscle Nerve 1988;11:136–45. Dimitrov GV, Arabadzhiev TI, Hogrel JY, Dimitrova NA. Simulation analysis of interference EMG during fatiguing voluntary contractions. Part II – changes in amplitude and spectral characteristics. J Electromyogr Kinesiol 2008;18:35–43. Dimitrov GV, Disselhorst-Klug C, Dimitrova NA, Schulte E, Rau G. Simulation analysis of the ability of different types of multi-electrodes to increase selectivity of detection and to reduce cross-talk. J Electromyogr Kinesiol 2003;13:125–38. Farina D, Fosci M, Merletti R. Motor unit recruitment strategies investigated by surface EMG variables. J Appl Physiol 2002;92:235–47. Farina D, Merletti R, Enoka RM. The extraction of neural strategies from the surface EMG. J Appl Physiol 2004;96:1486–95. Farmer SF, Swash M, Ingram DA, Stephens JA. Changes in motor unit synchronization following central nervous lesions in man. J Physiol 1993;463:83–105. Gabriel DA, Kamen G. Experimental and modeling investigation of spectral compression of biceps brachii SEMG activity with increasing force levels. J Electromyogr Kinesiol 2009;19:437–48. Gemperline JJ, Allen S, Walk D, Rymer WZ. Characteristics of motor unit discharge in subjects with hemiparesis. Muscle Nerve 1995;18:1101–14. Hara Y, Masakado Y, Chino N. The physiological functional loss of single thenar motor units in the stroke patients: when does it occur? Does it progress? Clin Neurophysiol 2004;115:97–103. Hausmanowa-Petrusewicz I, Kopec J. EMG parameters changes in the effort pattern at various loads in diseased muscle. Electromyogr Clin Neurophysiol 1983;23:213–28. Hu XL, Tong KY, Hung LK. Firing properties of motor units during fatigue in subjects after stroke. J Electromyogr Kinesiol 2006;16:469–76. Kallenberg LA, Hermens HJ. Motor unit properties of biceps brachii in chronic stroke patients assessed with high-density surface EMG. Muscle Nerve 2009;39:177–85. Kaplanis PA, Pattichis CS, Hadjileontiadis LJ, Roberts VC. Surface EMG analysis on normal subjects based on isometric voluntary contraction. J Electromyogr Kinesiol 2009;19:157–71. Knight CA, Kamen G. Superficial motor units are larger than deeper motor units in human vastus lateralis muscle. Muscle Nerve 2005;31:475–80. Lago P, Jones NB. Effect of motor-unit firing time statistics on e.m.g. spectra. Med Biol Eng Comput 1977;15:648–55. Latash ML. Spectral analysis of the electromyogram (EMG) in spinal cord trauma patients: I. Different types of the EMG and corresponding spectra. Electromyogr Clin Neurophysiol 1988a;28:319–27. Latash ML. Spectral analysis of the electromyogram (EMG) in spinal cord trauma patients: II. Motor unit and interference EMG power spectra. Electromyogr Clin Neurophysiol 1988b;28:329–34. Li X, Suresh A, Zhou P, Rymer W. Alterations in the spike amplitude distribution of the surface electromyogram post-stroke. IEEE Trans Biomed Eng 2013;60:845–52. Lindstrom L, Magnusson R. Interpretation of myoelectric power spectra: a model and its applications. Proc IEEE 1977;65:653–62. Lindstrom L, Magnusson R, Petersen I. Muscular fatigue and action potential conduction velocity changes studied with frequency analysis of EMG signals. Electromyography 1970;10:341–56. cLinnamo V, Strojnik V, Komi PV. Electromyogram power spectrum and features of the superimposed maximal M-wave during voluntary isometric actions in humans at different activation levels. Eur J Appl Physiol 2001;86:28–33. Lowery MM, Stoykov NS, Kuiken TA. A simulation study to examine the use of crosscorrelation as an estimate of surface EMG cross talk. J Appl Physiol 2003;94:1324–34. Lowery MM, Vaughan CL, Nolan PJ, O’Malley MJ. Spectral compression of the electromyographic signal due to decreasing muscle fiber conduction velocity. IEEE Trans Rehabil Eng 2000;8:353–61. Lukacs M, Vecsei L, Beniczky S. Large motor units are selectively affected following a stroke. Clin Neurophysiol 2008;119:2555–8. Lukacs M, Vecsei L, Beniczky S. Changes in muscle fiber density following a stroke. Clin Neurophysiol 2009;120:1539–42. Merletti R, Lo Conte LR. Surface EMG signal processing during isometric contractions. J Electromyogr Kinesiol 1997;7:241–50.
994
X. Li et al. / Clinical Neurophysiology 125 (2014) 988–994
Moritani T, Muro M. Motor unit activity and surface electromyogram power spectrum during increasing force of contraction. Eur J Appl Physiol Occup Physiol 1987;56:260–5. Muro M, Nagata A, Moritani T. Analysis of myoelectric signals during dynamic and isometric contractions. In: Matsui H, Kobayashi K, editors. Biomechanics VIIIA. Human Kinetics Publishers; 1983. p. 432–9. Muro M, Nagata A, Murakami K, Moritani T. Surface EMG power spectral analysis of neuromuscular disorders during isometric and isotonic contractions. Am J Phys Med 1982;61:244–54. Petrofsky JS, Lind AR. The influence of temperature on the amplitude and frequency components of the EMG during brief and sustained isometric contractions. Eur J Appl Physiol Occup Physiol 1980;44:189–200. Rainoldi A, Galardi G, Maderna L, Comi G, Lo Conte L, Merletti R. Repeatability of surface EMG variables during voluntary isometric contractions of the biceps brachii muscle. J Electromyogr Kinesiol 1999;9:105–19. Rix H, Malenge JP. Detecting small variations in shape. IEEE Trans Syst Man Cybern 1980;10:90–6. Rosenfalck A, Andreassen S. Impaired regulation of force and firing pattern of single motor units in patients with spasticity. J Neurol Neurosurg Psychiatry 1980;43:907–16. Seki K, Miyazaki Y, Watanabe M, Nagata A, Narusawa M. Surface electromyogram spectral characterization and motor unit activity during voluntary ramp contraction in men. Eur J Appl Physiol Occup Physiol 1991;63:165–72.
Solomonow M, Baten C, Smit J, Baratta R, Hermens H, D’Ambrosia R, et al. Electromyogram power spectra frequencies associated with motor unit recruitment strategies. J Appl Physiol 1990;68:1177–85. Spaans F, Wilts G. Denervation due to lesions of the central nervous system. An EMG study in cases of cerebral contusion and cerebrovascular accidents. J Neurol Sci 1982;57:291–305. Toffola ED, Sparpaglione D, Pistorio A, Buonocore M. Myoelectric manifestations of muscle changes in stroke patients. Arch Phys Med Rehabil 2001;82:661–5. Twitchell TE. The restoration of motor function following hemiplegia in man. Brain 1951;74:443–80. Wakeling JM. Patterns of motor recruitment can be determined using surface EMG. J Electromyogr Kinesiol 2009;19:199–207. Weytjens JL, van Steenberghe D. The effects of motor unit synchronization on the power spectrum of the electromyogram. Biol Cybern 1984;51:71–7. Yao W, Fuglevand RJ, Enoka RM. Motor-unit synchronization increases EMG amplitude and decreases force steadiness of simulated contractions. J Neurophysiol 2000;83:441–52. Zaman SA, Macisaac DT, Parker PA. Repeatability of surface EMG-based single parameter muscle fatigue assessment strategies in static and cyclic contractions. Conf Proc IEEE Eng Med Biol Soc 2011;2011:3857–60. Zipp P. Effect of electrode parameters on the bandwidth of the surface e.m.g. powerdensity spectrum. Med Biol Eng Comput 1978;16:537–41.