- Open Access
Estimating changes in metabolic power from EMG
© Blake and Wakeling; licensee Springer. 2013
Received: 16 January 2013
Accepted: 7 May 2013
Published: 19 May 2013
Metabolic rates can increase 21 times above resting levels during cycling with the majority attributed to muscular contractions. Metabolic estimates attained through gas exchange parameters are limited by the respiration rate and time delay with respect to these contractions. In contrast surface electromyography (EMG) contains instantaneous muscle contraction information at higher temporal resolutions. An adequate metabolic power-EMG relationship has not been established to use EMG as a metabolic estimate during dynamic activities. The purpose of this study was to establish a metabolic power-EMG relationship during non steady-state conditions. Participants cycled at workloads between 25 and 90% O2max while EMG and gas exchange were monitored. The EMG was resolved into intensities and total EMG intensity was calculated as the sum of intensities across all muscles for each pedal cycle. Metabolic power was estimated from gas exchange parameters and the mean total EMG intensity between breaths was calculated and used as breath-by-breath values. Comparisons were made between breath-by-breath resolutions of metabolic power and total EMG intensity. Different weighting coefficients were also applied to the EMG for each muscle to analyze the effects of different muscle weightings on metabolic power estimations. There was a significant correlation (r = 0.91) between estimates of metabolic power from EMG and gas exchange. Muscle weighting had a significant effect on metabolic power determination with the highest and lowest correlated estimates having the largest weightings on muscles proximal and distal to the knee respectively. This study demonstrates that EMG contains important information about the metabolic costs of muscle contractions and provides good predictions of metabolic changes during non steady-state conditions. Also, the importance of each muscle is workload dependent with inappropriate weightings reducing metabolic estimations. These findings have implications for future EMG applications as they provide more immediate, higher temporal resolution predictions of changes in metabolic power.
Contracting muscles require energy and more energy is needed as muscular force production increases. Metabolic rates can increase 21 times above resting levels in trained cyclists (Astrand & Rodahl1986), which is primarily attributed to the energy supplied to the contracting leg muscles by aerobic and anaerobic sources. Muscle contractions during dynamic activities must therefore contain considerable information about metabolic power.
Aerobic metabolism during isometric or dynamic activities is often estimated through indirect calorimetry using oxygen and carbon dioxide gas exchange where the energy utilized by the working muscles reflects the changes in pulmonary oxygen uptake (Poole et al.1992). Unfortunately measures of metabolic power based on gas exchange are unable to resolve metabolic costs to a resolution greater than the respiration rate. Consequently, information about muscular contractions that influence metabolic power is neglected since these contractions can occur more frequently than the respiration rate.
Changes in force during muscular contractions are primarily achieved by altering the number of active motor units and the motor unit firing rates. These changes can be detected using surface electromyography (EMG), which provides information about the active muscle by measuring the electrical signals of the motor unit action potentials. Gait cycles can take less than a second during high-speed movements and EMG fluctuates considerably within this time. EMG therefore must contain information about metabolic power at a higher temporal resolution than gas exchange measures, yet it is unknown if EMG can be used to estimate metabolic power changes during non steady-state dynamic activities.
Previous studies looking at the relationship between oxygen uptake (O2) and EMG have shown evidence of a linear relationship below the anaerobic threshold (Arnaud et al.1997;Bigland-Ritchie & Woods1976;Jammes et al.1998) and a non-linear relationship above the anaerobic threshold (Hug, Decherchi, et al.2004a). The non-linearity of this relationship is partially explained by a greater increase in EMG thanO2 due to a larger reliance on anaerobic metabolism. We recently found a significant monotonic increase in metabolic power (r = 0.86), estimated fromO2 that was associated with increased EMG intensity (Wakeling et al.2011). Mean values ofO2 and EMG intensity were calculated while cycling at or near steady-state at a range of workloads from 25-90% maximum oxygen uptake (O2max). Although this relationship was significant it only indicates that metabolic power is related to EMG intensity during steady-state activity when averaged over an extended time. The relationships at the workload transitions were not investigated and the higher temporal resolution of the EMG was ignored. Dynamic activities are rarely performed at constant workloads except during controlled experiments. It is therefore important to assess the relationship between metabolic power and EMG in non steady-state conditions.
Oxygen uptake kinetics in response to a stepwise change in workload can be modeled as exponential processes distinguished by changes in arterial and venous blood oxygen content (Jones & Poole2005). This provides good predictions of the oxygen uptake kinetics underlying the breath-by-breath fluctuations. Adequate changes in cardiac output andO2 to accommodate oxygen demands of the working muscles are delayed compared to the EMG signal since the altered venous blood oxygen content takes time to reach the lungs. For example, Whipp and co-workers (1982) showed that, in response to a stepwise increase in workload, there exists a time delay of approximately 20 seconds accounting for approximately 20% of the total increase inO2 before the primary exponential rise towards steady-state.
The purpose of this study was to further define metabolic changes based on EMG by establishing a significant relationship between metabolic power and EMG on a breath-by-breath basis in non steady-state conditions. It was hypothesized that a significant relationship exists at a greater temporal resolution than previously determined and good estimates of changes in metabolic power would be established from the EMG signal.
Time delay, transfer function coefficients ( τ and β) and correlations for each subject to estimate metabolic power from EMG
The primary finding of this study was the significant correlation (r = 0.91) between estimates of metabolic power from gas exchange and EMG intensity at a higher temporal resolution than the previously established relationship between metabolic power and EMG intensity (r = 0.86; (Wakeling et al.2011)). This was important as it included workload transitions as found in non steady-state conditions. This indicates that breath-by-breath EMG intensities provide a good estimate of changes in metabolic power.
The time-shifted correlation between metabolic power and EMG intensity (r = 0.85) revealed an improved relationship over the initial breath-by-breath correlation (r = 0.80). The t delay from the measured EMG signal to the subsequentO2 realized at the mouth was approximately 12 breaths or 28.33 ± 4.31 seconds. This is reasonable given the time delay of approximately 20 seconds attributed to the delay in the transport of blood from the working muscles to the lungs (Whipp et al.1982). The EMG signal therefore contains metabolic information closer in time to the actual energy use than gas exchange since the t delay is only apparent because the estimated metabolic power was measured at the mouth.
The maximum allowable value of τ in the transfer function (Eq. 1) was 100 breaths, yet six subjects had optimized values at this maximum (Table 1). Values greater than 100 breaths were also tested to ensure the resultant correlations were maximized. No subjects displayed appreciable gains in correlation (no difference in r values to the four significant digits reported) with values of τ above 100 breaths. The aim of this study was not to identify specific transfer function constants, but to show a significant relationship between metabolic power estimated from gas exchange and EMG. This was accomplished within the constraints imposed on the constants.
There were 3.19 ± 0.27 pedal cycles per breath across all workloads, which provides a small measure of smoothing, similar to a moving average, for the breath-by-breath EMG intensities when compared to the cycle-by-cycle values. Even at 25%O2max, with the lowest mean respiration rate of 23.83 ± 1.75 breaths per minute, there were only 4.15 ± 0.36 pedal cycles per breath. This level of data reduction maintains the important information of the EMG signal while reducing the amount of noise (Figure 4). Determination of an instantaneous measure of metabolic cost remains elusive, but these data and the small number of pedal cycles between breathes suggests that cycle-by-cycle EMG intensity with a time resolution of approximately 0.70 seconds would also provide a reasonable estimate of theO2 kinetics and therefore the metabolic power.
Similar to previous findings (Ericson1986;Hug & Dorel2009;Hug, Bendahan, et al.2004b;Lawrence & De Luca1983), some muscles displayed non-linear relationships between the EMG signal and force or workload (Figure 3). This shows that the importance of each muscle is workload dependent. Different weightings were applied to each muscle in order to evaluate its significance to the estimation of metabolic power. Given the importance of the larger knee extensor muscles in cycling (Ericson1986;Hug & Dorel2009), it was predicted that a reduced muscle set would provide as accurate predictions of metabolic changes as the entire set of muscles. In fact combinations of all muscles resulted in the best estimates and inappropriate weighting resulted in significantly reduced estimations (Figure 2). These results provide insight into the most important muscles for accurate metabolic predictions and imply that each muscle has a different role in metabolic power depending on the relative workload intensity. Therefore, when cycling involves a range of workloads, as is the case in this study, it is particularly important to include EMG from many muscles to attain an accurate account of metabolic power.
Muscles proximal to the knee, which include the muscles considered to be the primary power producers in cycling (VL and VM, (Ericson1986)), were more heavily weighted in Hcor. Eliminating muscles distal to the knee with the lowest weights (TA and Sol) did not improve the metabolic estimations as these muscles had high relative EMG intensities at lower workloads (Figure 3). The EMG intensities for the 60%O2max condition are inflated for all muscles (Figure 3) since they occurred after the 75% or 90% conditions for some participants. The 75% and 90% conditions occurred with respiration exchange ratios greater than one (Blake et al.2012) indicating a greater contribution of anaerobic energy sources. With the 60% condition taking place after these workloads and no rest between conditions, there was likely increased fatigue, which is associated with higher levels of EMG intensity (Edwards & Lippold1956;Petrofsky1979). These inflated values decrease the curvature of the best-fit relationship of EMG intensity across all workloads, especially for those muscles proximal to the knee (Figure 3). Despite these limitations the EMG signals contained substantial information about the metabolic power as evidenced by the significant correlation between the metabolic power estimations (r = 0.91).
Normalization of the EMG signals to the mean across all trials for each subject and each muscle presents a limitation to the interpretation of the EMG-metabolic cost relationship. This method weights all muscles the same and ignores metabolic differences of the structural and contractile properties of the individual muscles. Interestingly, imposing 5000 different weighting combinations to the muscle EMG signals only increased the metabolic power estimation correlation from r = 0.91 to r = 0.93 (Hcor). Changing the weightings is equivalent to using different normalization values for each muscle, and so the 5000 weighting combinations represented 5000 ways to normalize the muscles. The normalization method chosen for this study resulted in close to the best possible normalization for predicting metabolic power from EMG for this cycling task. However, an inappropriate set of normalization could result in poor results as shown by the significantly reduced correlation of Lcor (r = 0.77; Figure 2).
This study shows that breath-by-breath changes in EMG across ten leg muscles can be used as a good estimate of changes in metabolic power in non steady-state conditions. Estimations of metabolic changes on a cycle-by-cycle basis may also be reasonable given that breath-by-breath resolution acts as a moving average of approximately three to four pedal cycles, thereby maintaining the important features of the data. This has implications for future studies and applications involving EMG from large muscle groups during dynamic activities as it provides higher temporal resolution and more real-time predictions of changes in metabolic power than gas exchange measures. Also, metabolic power estimated from gas exchange is used in calculations of mechanical efficiency, which implies that breath-by-breath changes in EMG can also be used to investigate the relationship of mechanical work and metabolic power.
Nine competitive male cyclists (age = 41.8 ± 2.7 yr, mass = 77.2 ± 2.2 kg, height = 1.81 ± 0.01 m,O2max = 64.65 mL kg - 1 min - 1 , yearly mileage = 9428 ± 1913 km; mean ± SEM) participated in the study. The participants gave their informed written consent, and the ethics committee in accordance with the Office of Research Ethics at Simon Fraser University approved all procedures.
The cycling protocol and data collection was completed as described previously (Blake et al.2012). Briefly, participants cycled at a freely chosen cadence (94.7 ± 1.1 rpm) continuously for 18 minutes in three minute intervals at 25, 40 and 55% of the power output atO2max in random order followed by 60, 75 and 90%O2max, also in random order, while breath-by-breath oxygen and carbon dioxide gas exchange (Vmax 229 metabolic cart, Sensormedics, Yorba Linda, California) and EMG from 10 leg muscles (tibialis anterior (TA, MG, LG, Sol, VM, RF, VL, ST, BF and GM)) were recorded. The EMG recording was completed using bipolar Ag/AgCl surface electrodes (10mm diameter, 21mm spacing) and was amplified (factor of 1000), band-pass filtered (bandwidth 10-500 Hz; Biovision, Wehrheim, Germany) and recorded at 2000 Hz through a 16-bit A/D converter (USB-6210, National Instruments, Austin, TX). The EMG was resolved into intensities using wavelets (Von Tscharner2000) and normalized to the mean intensity for each participant for each muscle across all conditions. The EMG intensity is a measure of the power of the EMG signal in the frequency range 10-500 Hz. Total EMG intensity was calculated as the sum of the EMG intensities from each muscle for each pedal cycle and thus was termed the cycle-by-cycle value (Figure 4).
Variation in the structural and contractile properties of different muscles may be reflected in energy requirement differences not identified in the EMG signals since each muscle is normalized to its own mean for each subject across all trials. Therefore, in a subsequent analysis, different weighting coefficients were applied to the EMG intensity from each muscle to analyze the effect of muscle weighting on the correlation between the estimations of metabolic power from the EMG and gas exchange. Random weightings were assigned to the individual muscles in 5000 different combinations for each subject. These were then used in the transfer function, re-optimizing for the coefficients (τ, β and t delay), to predict metabolic power and re-calculate the estimated metabolic power correlations. In order to evaluate the differences in the individual muscle weighting coefficients that produced the best and worst metabolic power correlations, mean weightings of the 100 highest (Hcor) and lowest (Lcor) correlated muscle combinations were compared. Two-sided t-tests were conducted between Hcor and Lcor weighting coefficients for each muscle to determine if the mean values were significantly different. Statistical tests were considered significant at p ≤ 0.05 and results are reported as mean ± SEM.
We thank NSERC for financial support.
- Arnaud S, Zattara-Hartmann MC, Tomei C, Jammes Y: Correlation between muscle metabolism and changes in M-wave and surface electromyogram: dynamic constant load leg exercise in untrained subjects. Muscle Nerve 1997, 20(9):1197-1199. 10.1002/(SICI)1097-4598(199709)20:9<1197::AID-MUS20>3.0.CO;2-PView ArticleGoogle Scholar
- Astrand PO, Rodahl K: Textbook of work physiology: physiological bases of exercise. 3rd edition. New York: McGraw-Hill, Inc.; 1986.Google Scholar
- Bigland-Ritchie B, Woods JJ: Integrated electromyogram and oxygen uptake during positive and negative work. J Physiol 1976, 260(2):267-277.View ArticleGoogle Scholar
- Blake OM, Champoux Y, Wakeling JM: Muscle coordination patterns for efficient cycling. Med Sci Sports Exerc 2012, 44(5):926-938. 10.1249/MSS.0b013e3182404d4bView ArticleGoogle Scholar
- Edwards RG, Lippold OCJ: The relation between force and integrated electrical activity in fatigued muscle. J Physiol 1956, 132(3):677.View ArticleGoogle Scholar
- Ericson MO: On the biomechanics of cycling. A study of joint and muscle load during exercise on the bicycle ergometer. Scand J Rehabil Med Suppl 1986, 16: 1-43.Google Scholar
- Foss ML, Keteyian SJ, Fox EL: Fox’s physiological basis for exercise and sport (6th edition). Dubuque (IA): McGraw-Hill; 1998.Google Scholar
- Hug F, Bendahan D, Le Fur Y, Cozzone PJ, Grelot L: Heterogeneity of muscle recruitment pattern during pedaling in professional road cyclists: a magnetic resonance imaging and electromyography study. Eur J Appl Physiol 2004, 92(3):334-342.View ArticleGoogle Scholar
- Hug F, Decherchi P, Marqueste T, Jammes Y: EMG versus oxygen uptake during cycling exercise in trained and untrained subjects. J Electromyogr Kinesiol 2004, 14(2):187-195. 10.1016/S1050-6411(03)00081-6View ArticleGoogle Scholar
- Hug F, Dorel S: Electromyographic analysis of pedaling: a review. J Electromyogr Kinesiol 2009, 19(2):182-198. 10.1016/j.jelekin.2007.10.010View ArticleGoogle Scholar
- Jammes Y, Caquelard F, Badier M: Correlation between surface electromyogram, oxygen uptake and blood lactate concentration during dynamic leg exercises. Resp Physiol 1998, 112(2):167-174. 10.1016/S0034-5687(98)00023-1View ArticleGoogle Scholar
- Jones AM, Poole DC: Oxygen uptake kinetics in sport, exercise and medicine. London: Routledge; 2005.Google Scholar
- Lamarra N, Whipp BJ, Ward SA, Wasserman K: Effect of interbreath fluctuations on characterizing exercise gas exchange kinetics. J Appl Physiol 1987, 62(5):2003-2012. 10.1063/1.339541View ArticleGoogle Scholar
- Lawrence JH, De Luca CJ: Myoelectric signal versus force relationship in different human muscles. J Appl Physiol 1983, 54(6):1653-1659.Google Scholar
- Lee SSM, de Boef MM, Arnold AS, Biewener AA, Wakeling JM: EMG analysis tuned for determining the timing and level of activation in different motor units. J Electromyogr Kinesiol 2011, 21(4):557-565. 10.1016/j.jelekin.2011.04.003View ArticleGoogle Scholar
- Petrofsky JS: Frequency and amplitude analysis of the EMG during exercise on the bicycle ergometer. Eur J Appl Physiol 1979, 41(1):1-15. 10.1007/BF00424464View ArticleGoogle Scholar
- Poole DC, Gaesser GA, Hogan MC, Knight DR, Wagner PD: Pulmonary and leg VO2 during submaximal exercise: implications for muscular efficiency. J Appl Physiol 1992, 72(2):805-810.Google Scholar
- Von Tscharner V: Intensity analysis in time-frequency space of surface myoelectric signals by wavelets of specified resolution. J Electromyogr Kinesiol 2000, 10(6):433-445. 10.1016/S1050-6411(00)00030-4View ArticleGoogle Scholar
- Wakeling JM, Blake OM, Wong I, Rana M, Lee SSM: Movement mechanics as a determinate of muscle structure, recruitment and coordination. Phil Trans R Soc B 2011, 366(1570):1554-1564. 10.1098/rstb.2010.0294View ArticleGoogle Scholar
- Whipp BJ, Ward SA, Lamarra N, Davis JA, Wasserman K: Parameters of ventilatory and gas exchange dynamics during exercise. J Appl Physiol 1982, 52(6):1506-1513.Google Scholar
- Zajac FE: Muscle and tendon: properties, models, scaling, and application to biomechanics and motor control. Crit Rev Biomed Eng 1989, 17(4):359-411.Google Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.