The compensatory interaction between motor unit firing behavior and muscle force during fatigue.

Throughout the literature, different observations of motor unit firing behavior during muscle fatigue have been reported and explained with varieties of conjectures. The disagreement amongst previous studies has resulted, in part, from the limited number of available motor units and from the misleading practice of grouping motor unit data across different subjects, contractions, and force levels. To establish a more clear understanding of motor unit control during fatigue, we investigated the firing behavior of motor units from the vastus lateralis muscle of individual subjects during a fatigue protocol of repeated voluntary constant force isometric contractions. Surface electromyographic decomposition technology provided the firings of 1,890 motor unit firing trains. These data revealed that to sustain the contraction force as the muscle fatigued, the following occurred: 1) motor unit firing rates increased; 2) new motor units were recruited; and 3) motor unit recruitment thresholds decreased. Although the degree of these adaptations was subject specific, the behavior was consistent in all subjects. When we compared our empirical observations with those obtained from simulation, we found that the fatigue-induced changes in motor unit firing behavior can be explained by increasing excitation to the motoneuron pool that compensates for the fatigue-induced decrease in muscle force twitch reported in empirical studies. Yet, the fundamental motor unit control scheme remains invariant throughout the development of fatigue. These findings indicate that the central nervous system regulates motor unit firing behavior by adjusting the operating point of the excitation to the motoneuron pool to sustain the contraction force as the muscle fatigues.

Synchronization of motor unit firings: an epiphenomenon of firing rate characteristics not common inputs.

Synchronous motor unit firing instances have been attributed to anatomical inputs shared by motoneurons. Yet, there is a lack of empirical evidence confirming the notion that common inputs elicit synchronization under voluntary conditions. We tested this notion by measuring synchronization between motor unit action potential trains (MUAPTs) as their firing rates progressed within a contraction from a relatively low force level to a higher one. On average, the degree of synchronization decreased as the force increased. The common input notion provides no empirically supported explanation for the observed synchronization behavior. Therefore, we investigated a more probable explanation for synchronization. Our data set of 17,546 paired MUAPTs revealed that the degree of synchronization varies as a function of two characteristics of the motor unit firing rate: the similarity and the slope as a function of force. Both are measures of the excitation of the motoneurons. As the force generated by the muscle increases, the firing rate slope decreases, and the synchronization correspondingly decreases. Different muscles have motor units with different firing rate characteristics and display different amounts of synchronization. Although this association is not proof of causality, it consistently explains our observations and strongly suggests further investigation. So viewed, synchronization is likely an epiphenomenon, subject to countless unknown neural interactions. As such, synchronous firing instances may not be the product of a specific design and may not serve a specific physiological purpose. Our explanation for synchronization has the advantage of being supported by empirical evidence, whereas the common input does not.

Is the notion of central fatigue based on a solid foundation?

Exercise-induced muscle fatigue has been shown to be the consequence of peripheral factors that impair muscle fiber contractile mechanisms. Central factors arising within the central nervous system have also been hypothesized to induce muscle fatigue, but no direct empirical evidence that is causally associated to reduction of muscle force-generating capability has yet been reported. We developed a simulation model to investigate whether peripheral factors of muscle fatigue are sufficient to explain the muscle force behavior observed during empirical studies of fatiguing voluntary contractions, which is commonly attributed to central factors. Peripheral factors of muscle fatigue were included in the model as a time-dependent decrease in the amplitude of the motor unit force twitches. Our simulation study indicated that the force behavior commonly attributed to central fatigue could be explained solely by peripheral factors during simulated fatiguing submaximal voluntary contractions. It also revealed important flaws regarding the use of the interpolated twitch response from electrical stimulation of the muscle as a means for assessing central fatigue. Our analysis does not directly refute the concept of central fatigue. However, it raises important concerns about the manner in which it is measured and about the interpretation of the commonly accepted causes of central fatigue and questions the very need for the existence of central fatigue.

Decomposition of surface EMG signals from cyclic dynamic contractions.

Over the past 3 decades, various algorithms used to decompose the electromyographic (EMG) signal into its constituent motor unit action potentials (MUAPs) have been reported. All are limited to decomposing EMG signals from isometric contraction. In this report, we describe a successful approach to decomposing the surface EMG (sEMG) signal collected from cyclic (repeated concentric and eccentric) dynamic contractions during flexion/extension of the elbow and during gait. The increased signal complexity introduced by the changing shapes of the MUAPs due to relative movement of the electrodes and the lengthening/shortening of muscle fibers was managed by an incremental approach to enhancing our established algorithm for decomposing sEMG signals obtained from isometric contractions. We used machine-learning algorithms and time-varying MUAP shape discrimination to decompose the sEMG signal from an increasingly challenging sequence of pseudostatic and dynamic contractions. The accuracy of the decomposition results was assessed by two verification methods that have been independently evaluated. The firing instances of the motor units had an accuracy of ∼90% with a MUAP train yield as high as 25. Preliminary observations from the performance of motor units during cyclic contractions indicate that during repetitive dynamic contractions, the control of motor units is governed by the same rules as those evidenced during isometric contractions. Modifications in the control properties of motoneuron firings reported by previous studies were not confirmed. Instead, our data demonstrate that the common drive and hierarchical recruitment of motor units are preserved during concentric and eccentric contractions.

Biomechanical benefits of the Onion-Skin motor unit control scheme.

Muscle force is modulated by varying the number of active motor units and their firing rates. For the past five decades, the notion that the magnitude of the firing rates is directly related to motor unit size and recruitment threshold has been widely accepted. This construct, here named the After-hyperpolarization scheme evolved from observations in electrically stimulated cat motoneurons and from reported observations in voluntary contractions in humans. It supports the assumption that the firing rates of motor units match their mechanical properties to "optimize" force production, so that the firing rate range corresponds to that required for force-twitch fusion (tetanization) and effective graduation of muscle force. In contrast, we have shown that, at any time and force during isometric voluntary constant-force contractions in humans, the relationship between firing rate and recruitment threshold is inversely related. We refer to this construct as the Onion-Skin scheme because earlier-recruited motor units always have greater firing rates than latter-recruited ones. By applying a novel mathematical model that calculates the force produced by a muscle for the two schemes we found that the Onion-Skin scheme is more energy efficient, provides smoother muscle force at low to moderate force levels, and appears to be more conducive to evolutionary survival than the After-hyperpolarization scheme.

Statistically rigorous calculations do not support common input and long-term synchronization of motor-unit firings.

Over the past four decades, various methods have been implemented to measure synchronization of motor-unit firings. In this work, we provide evidence that prior reports of the existence of universal common inputs to all motoneurons and the presence of long-term synchronization are misleading, because they did not use sufficiently rigorous statistical tests to detect synchronization. We developed a statistically based method (SigMax) for computing synchronization and tested it with data from 17,736 motor-unit pairs containing 1,035,225 firing instances from the first dorsal interosseous and vastus lateralis muscles--a data set one order of magnitude greater than that reported in previous studies. Only firing data, obtained from surface electromyographic signal decomposition with >95% accuracy, were used in the study. The data were not subjectively selected in any manner. Because of the size of our data set and the statistical rigor inherent to SigMax, we have confidence that the synchronization values that we calculated provide an improved estimate of physiologically driven synchronization. Compared with three other commonly used techniques, ours revealed three types of discrepancies that result from failing to use sufficient statistical tests necessary to detect synchronization. 1) On average, the z-score method falsely detected synchronization at 16 separate latencies in each motor-unit pair. 2) The cumulative sum method missed one out of every four synchronization identifications found by SigMax. 3) The common input assumption method identified synchronization from 100% of motor-unit pairs studied. SigMax revealed that only 50% of motor-unit pairs actually manifested synchronization.

Error reduction in EMG signal decomposition.

Decomposition of the electromyographic (EMG) signal into constituent action potentials and the identification of individual firing instances of each motor unit in the presence of ambient noise are inherently probabilistic processes, whether performed manually or with automated algorithms. Consequently, they are subject to errors. We set out to classify and reduce these errors by analyzing 1,061 motor-unit action-potential trains (MUAPTs), obtained by decomposing surface EMG (sEMG) signals recorded during human voluntary contractions. Decomposition errors were classified into two general categories: location errors representing variability in the temporal localization of each motor-unit firing instance and identification errors consisting of falsely detected or missed firing instances. To mitigate these errors, we developed an error-reduction algorithm that combines multiple decomposition estimates to determine a more probable estimate of motor-unit firing instances with fewer errors. The performance of the algorithm is governed by a trade-off between the yield of MUAPTs obtained above a given accuracy level and the time required to perform the decomposition. When applied to a set of sEMG signals synthesized from real MUAPTs, the identification error was reduced by an average of 1.78%, improving the accuracy to 97.0%, and the location error was reduced by an average of 1.66 ms. The error-reduction algorithm in this study is not limited to any specific decomposition strategy. Rather, we propose it be used for other decomposition methods, especially when analyzing precise motor-unit firing instances, as occurs when measuring synchronization.

Transposed firing activation of motor units.

Muscles are composed of groups of muscle fibers, called motor units, each innervated by a single motoneuron originating in the spinal cord. During constant or linearly varying voluntary force contractions, motor units are activated in a hierarchical order, with the earlier-recruited motor units having greater firing rates than the later-recruited ones. We found that this normal pattern of firing activation can be altered during oscillatory contractions where the force oscillates at frequencies ≥2 Hz. During these high-frequency oscillations, the activation of the lower-threshold motor units effectively decreases and that of the higher-threshold motor units effectively increases. This transposition of firing activation provides means to activate higher-threshold motor units preferentially. Our results demonstrate that the hierarchical regulation of motor unit activation can be manipulated to activate specific motoneuron populations preferentially. This finding can be exploited to develop new forms of physical therapies and exercise programs that enhance muscle performance or that target the preferential atrophy of high-threshold motor units as a result of aging or motor disorders such as stroke and amyotrophic lateral sclerosis.

Dynamical learning and tracking of tremor and dyskinesia from wearable sensors.

We have developed and evaluated several dynamical machine-learning algorithms that were designed to track the presence and severity of tremor and dyskinesia with 1-s resolution by analyzing signals collected from Parkinson's disease (PD) patients wearing small numbers of hybrid sensors with both 3-D accelerometeric and surface-electromyographic modalities. We tested the algorithms on a 44-h signal database built from hybrid sensors worn by eight PD patients and four healthy subjects who carried out unscripted and unconstrained activities of daily living in an apartment-like environment. Comparison of the performance of our machine-learning algorithms against independent clinical annotations of disorder presence and severity demonstrates that, despite their differing approaches to dynamic pattern classification, dynamic neural networks, dynamic support vector machines, and hidden Markov models were equally effective in keeping error rates of the dynamic tracking well below 10%. A common set of experimentally derived signal features were used to train the algorithm without the need for subject-specific learning. We also found that error rates below 10% are achievable even when our algorithms are tested on data from a sensor location that is different from those used in algorithm training.

High-resolution tracking of motor disorders in Parkinson's disease during unconstrained activity.

Parkinson's disease (PD) can present with a variety of motor disorders that fluctuate throughout the day, making assessment a challenging task. Paper-based measurement tools can be burdensome to the patient and clinician and lack the temporal resolution needed to accurately and objectively track changes in motor symptom severity throughout the day. Wearable sensor-based systems that continuously monitor PD motor disorders may help to solve this problem, although critical shortcomings persist in identifying multiple disorders at high temporal resolution during unconstrained activity. The purpose of this study was to advance the current state of the art by (1) introducing hybrid sensor technology to concurrently acquire surface electromyographic (sEMG) and accelerometer data during unconstrained activity and (2) analyzing the data using dynamic neural network algorithms to capture the evolving temporal characteristics of the sensor data and improve motor disorder recognition of tremor and dyskinesia. Algorithms were trained (n=11 patients) and tested (n=8 patients; n=4 controls) to recognize tremor and dyskinesia at 1-second resolution based on sensor data features and expert annotation of video recording during 4-hour monitoring periods of unconstrained daily activity. The algorithms were able to make accurate distinctions between tremor, dyskinesia, and normal movement despite the presence of diverse voluntary activity. Motor disorder severity classifications averaged 94.9% sensitivity and 97.1% specificity based on 1 sensor per symptomatic limb. These initial findings indicate that new sensor technology and software algorithms can be effective in enhancing wearable sensor-based system performance for monitoring PD motor disorders during unconstrained activities.

Neural control of muscle force: indications from a simulation model.

We developed a model to investigate the influence of the muscle force twitch on the simulated firing behavior of motoneurons and muscle force production during voluntary isometric contractions. The input consists of an excitatory signal common to all the motor units in the pool of a muscle, consistent with the "common drive" property. Motor units respond with a hierarchically structured firing behavior wherein at any time and force, firing rates are inversely proportional to recruitment threshold, as described by the "onion skin" property. Time- and force-dependent changes in muscle force production are introduced by varying the motor unit force twitches as a function of time or by varying the number of active motor units. A force feedback adjusts the input excitation, maintaining the simulated force at a target level. The simulations replicate motor unit behavior characteristics similar to those reported in previous empirical studies of sustained contractions: 1) the initial decrease and subsequent increase of firing rates, 2) the derecruitment and recruitment of motor units throughout sustained contractions, and 3) the continual increase in the force fluctuation caused by the progressive recruitment of larger motor units. The model cautions the use of motor unit behavior at recruitment and derecruitment without consideration of changes in the muscle force generation capacity. It describes an alternative mechanism for the reserve capacity of motor units to generate extraordinary force. It supports the hypothesis that the control of motoneurons remains invariant during force-varying and sustained isometric contractions.

Preferred sensor sites for surface EMG signal decomposition.

Technologies for decomposing the electromyographic (EMG) signal into its constituent motor unit action potential trains have become more practical by the advent of a non-invasive methodology using surface EMG (sEMG) sensors placed on the skin above the muscle of interest (De Luca et al 2006 J. Neurophysiol. 96 1646-57 and Nawab et al 2010 Clin. Neurophysiol. 121 1602-15). This advancement has widespread appeal among researchers and clinicians because of the ease of use, reduced risk of infection, and the greater number of motor unit action potential trains obtained compared to needle sensor techniques. In this study we investigated the influence of the sensor site on the number of identified motor unit action potential trains in six lower limb muscles and one upper limb muscle with the intent of locating preferred sensor sites that provided the greatest number of decomposed motor unit action potential trains, or motor unit yield. Sensor sites rendered varying motor unit yields throughout the surface of a muscle. The preferred sites were located between the center and the tendinous areas of the muscle. The motor unit yield was positively correlated with the signal-to-noise ratio of the detected sEMG. The signal-to-noise ratio was inversely related to the thickness of the tissue between the sensor and the muscle fibers. A signal-to-noise ratio of 3 was found to be the minimum required to obtain a reliable motor unit yield.

Hierarchical control of motor units in voluntary contractions.

For the past five decades there has been wide acceptance of a relationship between the firing rate of motor units and the afterhyperpolarization of motoneurons. It has been promulgated that the higher-threshold, larger-soma, motoneurons fire faster than the lower-threshold, smaller-soma, motor units. This relationship was based on studies on anesthetized cats with electrically stimulated motoneurons. We questioned its applicability to motor unit control during voluntary contractions in humans. We found that during linearly force-increasing contractions, firing rates increased as exponential functions. At any time and force level, including at recruitment, the firing rate values were inversely related to the recruitment threshold of the motor unit. The time constants of the exponential functions were directly related to the recruitment threshold. From the Henneman size principle it follows that the characteristics of the firing rates are also related to the size of the soma. The "firing rate spectrum" presents a beautifully simple control scheme in which, at any given time or force, the firing rate value of earlier-recruited motor units is greater than that of later-recruited motor units. This hierarchical control scheme describes a mechanism that provides an effective economy of force generation for the earlier-recruited lower force-twitch motor units, and reduces the fatigue of later-recruited higher force-twitch motor units-both characteristics being well suited for generating and sustaining force during the fight-or-flight response.

Inter-electrode spacing of surface EMG sensors: reduction of crosstalk contamination during voluntary contractions.

We investigated the influence of inter-electrode spacing on the degree of crosstalk contamination in surface electromyographic (sEMG) signals in the tibialis anterior (target muscle), generated by the triceps surae (crosstalk muscle), using bar and disk electrode arrays. The degree of crosstalk contamination was assessed for voluntary constant-force isometric contractions and for dynamic contractions during walking. Single-differential signals were acquired with inter-electrode spacing ranging from 5 mm to 40 mm. Additionally, double differential signals were acquired at 10 mm spacing using the bar electrode array. Crosstalk contamination at the target muscle was expressed as the ratio of the detected crosstalk signal to that of the target muscle signal. The crosstalk contamination ratio approached a mean of 50% for the 40 mm spacing for triceps surae muscle contractions at 80% MVC and tibialis anterior muscle contractions at 10% MVC. For single differential recordings, the minimum crosstalk contamination was obtained from the 10 mm spacing. The results showed no significant differences between the bar and disk electrode arrays. During walking, the crosstalk contamination on the tibialis anterior muscle reached levels of 23% for a commonly used 22 mm spacing single-differential disk sensor, 17% for a 10 mm spacing single-differential bar sensor, and 8% for a 10 mm double-differential bar sensor. For both studies the effect of electrode spacing on crosstalk contamination was statistically significant. Crosstalk contamination and inter-electrode spacing should therefore be a serious concern in gait studies when the sEMG signal is collected with single differential sensors. The contamination can distort the target muscle signal and mislead the interpretation of its activation timing and force magnitude.

Resolving signal complexities for ambulatory monitoring of motor function in Parkinson's disease.

Automatic tracking of movement disorders in patients with Parkinson's disease (PD) is dependent on the ability of machine learning algorithms to resolve the complex and unpredictable characteristics of wearable sensor data. The challenge reflects the variety of movement disorders that fluctuate throughout the day which can be confounded by voluntary activities of daily life. Our approach is the development of multiple dynamic neural network (DNN) classifiers whose application are governed by a rule-based controller within the Integrated Processing and Understanding of Signals (IPUS) framework. Solutions are described for time-varying occurrences of tremor and dyskinesia, classified at 1 s resolution from surface electromyographic (sEMG) and tri-axial accelerometer (ACC) data acquired from patients with PD. The networks were trained and tested on separate datasets, respectively, while subjects performed unscripted and unconstrained activities in a home-like setting. Performance of the classifiers achieved an overall global error rate of less than 10%.

Resolving signal complexities for ambulatory monitoring of motor function in Parkinson's disease.

Automatic tracking of movement disorders in patients with Parkinson's disease (PD) is dependent on the ability of machine learning algorithms to resolve the complex and unpredictable characteristics of wearable sensor data. The challenge reflects the variety of movement disorders that fluctuate throughout the day which can be confounded by voluntary activities of daily life. Our approach is the development of multiple dynamic neural network (DNN) classifiers whose application are governed by a rule-based controller within the Integrated Processing and Understanding of Signals (IPUS) framework. Solutions are described for time-varying occurrences of tremor and dyskinesia, classified at 1 s resolution from surface electromyographic (sEMG) and tri-axial accelerometer (ACC) data acquired from patients with PD. The networks were trained and tested on separate datasets, respectively, while subjects performed unscripted and unconstrained activities in a home-like setting. Performance of the classifiers achieved an overall global error rate of less than 10%.

Dynamic neural network detection of tremor and dyskinesia from wearable sensor data.

We present a dynamic neural network (DNN) solution for detecting time-varying occurrences of tremor and dyskinesia at 1 s resolution from time series data acquired from surface electromyographic (sEMG) sensors and tri-axial accelerometers worn by patients with Parkinson's disease (PD). The networks were trained and tested on separate datasets, each containing approximately equal proportions of tremor, dyskinesia, and disorder-free data from 8 PD and 4 control subjects performing unscripted and unconstrained activities in an apartment-like environment. During DNN testing, tremor was detected with a sensitivity of 93% and a specificity of 95%, while dyskinesia was detected with a sensitivity of 91% and a specificity of 93%. Similar sensitivity and specificity levels were obtained when DNN testing was carried out on subjects who were not included in DNN training.

Relationship between firing rate and recruitment threshold of motoneurons in voluntary isometric contractions.

We used surface EMG signal decomposition technology to study the control properties of numerous simultaneously active motor units. Six healthy human subjects of comparable age (21 +/- 0.63 yr) and physical fitness were recruited to perform isometric contractions of the vastus lateralis (VL), first dorsal interosseous (FDI), and tibialis anterior (TA) muscles at the 20, 50, 80, and 100% maximum voluntary contraction force levels. EMG signals were collected with a five-pin surface array sensor that provided four channels of data. They were decomposed into the constituent action potentials with a new decomposition algorithm. The firings of a total of 1,273 motor unit action potential trains, 20-30 per contraction, were obtained. The recruitment thresholds and mean firing rates of the motor units were calculated, and mathematical equations were derived. The results describe a hierarchical inverse relationship between the recruitment thresholds and the firing rates, including the first and second derivatives, i.e., the velocity and the acceleration of the firing rates. This relationship describes an "operating point" for the motoneuron pool that remains consistent at all force levels and is modulated by the excitation. This relationship differs only slightly between subjects and more distinctly across muscles. These results support the "onion skin" property that suggests a basic control scheme encoded in the physical properties of motoneurons that responds consistently to a "common drive" to the motoneuron pool.