Neuroscience Letters 556 (2013) 84–88
Contents lists available at ScienceDirect
Neuroscience Letters journal homepage: www.elsevier.com/locate/neulet
The impact of rectification on the electrically evoked long-latency reflex of the biceps brachii muscle Ssuhir Alaid ∗ , Malte E. Kornhuber Neurologic Hospital, Martin-Luther University Halle-Wittenberg, Ernst-Grube-Str. 40, 06120 Halle (Saale), Germany
h i g h l i g h t s • • • • •
The long latency reflex (LLR) is presumed to take a transcortical pathway. The impact of rectification on electrical LLR was studied over the biceps brachii. Rectified LLR amplitudes made up only 30% of non-rectified values. Thereby influences from stimulus strength or motor unit recruitment were obscured. Increased frequency contents after rectification result in excess phase cancellation.
a r t i c l e
i n f o
Article history: Received 21 January 2013 Received in revised form 14 June 2013 Accepted 4 October 2013 Keywords: Electrical train stimuli Long latency reflex Motor unit Rectification Upper arm muscle Weight load
a b s t r a c t Long latency reflexes (LLR) were elicited electrically and obtained by full wave rectified and non-rectified data recordings in 10 healthy subjects. After single or train stimuli (sensory radial nerve; interstimulus interval 3 ms) amplitude and peak latency values were measured over the bent biceps brachii (BB) muscle, either without or with 1.5 kg weight load. After rectification, mean LLR amplitude values made up 30% of the non-rectified data, independent from the stimulus type and weight load. In the non-rectified data, a significant gain in amplitude resulted from train stimuli compared with single stimuli, and from weight load compared to no weight load. No such significant difference was detected when rectified data were analysed. Furthermore, average amplitude values of rectified and non-rectified curves were studied using 11 sine waves and damped sine waves with equal phase intervals that were varied from 0◦ up to 34.4◦ . Phase shifts ranging from 10◦ to 25◦ resulted in excess amplitude decline of rectified data compared with non-rectified data. The long and polysynaptic course that LLR information takes leads to considerable overlap of responses to subsequent stimuli. This overlap of motor unit potentials forming the LLR obviously results in excess amplitude cancellation after rectification as shown for sine and damped sine waves. Rectification leads to an increase in the frequency content of the data that renders it prone to phase cancellation. In the present study, this cancellation was harmful as it prevented detection of important factors of influence such as stimulus strength and motor unit recruitment level. © 2013 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Stimulus evoked electromyographic (EMG) activity may present with low amplitudes in the microvolt range. The signal to noise ratio can be improved by averaging. Usually a preceding step with full wave rectification is recommended for this purpose [4]. Reliability of the results obtained after rectification depends on several assumptions [7] of which a constant delay between stimulus and
Abbreviations: BB, biceps brachii; LLR, long latency reflex. ∗ Corresponding author at: Department of Neurology, Martin-Luther University Halle-Wittenberg, Ernst-Grube-Str. 40, 06120 Halle (Saale), Germany. Tel.: +49 0345 557 3340; fax: +49 0345 557 3335. E-mail address: [email protected] (S. Alaid). 0304-3940/$ – see front matter © 2013 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.neulet.2013.10.009
response is of major importance. When intact motor units are stimulated in the periphery, this precondition can be regarded as given. However, in the case of polysynaptic reflex responses like the long latency reflex (LLR) considerable jitter of reflex responses may occur. Full wave rectification has previously been recommended [3], although it had not been based on comparative analyses of averaged rectified and non-rectified data. To follow this recommendation is potentially hazardous since each step in the processing of a bioelectrical signal may be associated with a loss of information content. This study used synthetic and original bioelectric signals to assess the effect of rectification on the amplitude and latency of the EMG signal. The aim of the present study was to clarify, whether full wave rectification is a useful and necessary step prior to averaging LLR responses or represents an unjustified alteration.
S. Alaid, M.E. Kornhuber / Neuroscience Letters 556 (2013) 84–88
85
2. Materials and methods
2.3. Theoretical considerations
2.1. Subjects
In order to assess the influence of overlap on amplitude values after averaging, 2 synthetic curves were studied as substitutes for single LLR signals. The first curve was a sine wave. In addition, a damped sine wave was taken as it better reflects the configuration of motor unit potentials. Calculations were performed with MATLAB version 7.0 (The MathWorks, Natick, MA, USA).
Ten healthy subjects (hospital staff, 24–45 years old, 6 females) volunteered in the following study. Experiments started at 10 a.m. and did not last longer than 1.5 h. The subjects gave their informed consent to participate in the investigation. Ethical approval for the study was received from the Martin Luther University of HalleWittenberg.
2.2. Data acquisition and evaluation Subjects were seated comfortably to an adjustable chair. The subject’s left forearm was positioned in the horizontal plane. In order to obtain rectified and non-rectified curves simultaneously, 2 channels were used to record LLR over the short head of the biceps brachii (BB) muscle. Thus, 2 “active” electrodes were set at the middle of the belly of the short head of the BB. The 2 reference electrodes were set over the olecranon and the wet ground electrode was tied to the middle of the left forearm. Electromyographic (EMG) recordings were made with a Multiliner EMG machine (Viasys, Hoechberg, Germany) during slight voluntary contraction either without weight or after loading 1.5 kg on the hand of the investigated arm. Two different muscle contraction levels were chosen since it has been demonstrated that phase cancellation in averages of rectified surface EMG curves may depend on the motor unit recruiting level [7]. Subjects were asked to fully supinate their left forearm in order to maintain a constant level of muscle contraction. The latter was carefully controlled during each experimental condition by auditory feedback and by eye. The electrical stimuli consisted of square wave pulses with duration of 0.2 ms over the left N. radialis superficialis. Supramaximum rectangular current pulses (9–12 mA) were delivered through 2 surface electrodes with a fixed distance of 2.5 cm fastened slightly proximal to the tabatière over the radial bone at the left wrist, the cathode set proximal to the anode. Time window: 500 ms, 50 ms before and 450 ms following the stimulus. Amplification: 200 V/division. Filter settings 2 Hz, 1 kHz. Sample rate 4 kHz. LLR were analysed from an average of 120 traces after single stimuli and trains of 3 stimuli with a fixed interstimulus interval (ISI) of 3 ms delivered at a rate of 1 Hz (Fig. 1). The investigated left arm was held actively bent at 90◦ . LLR may consist of up to three waves that follow the H reflex (short latency reflex, SLR), namely LLR1, LLR2 and LLR3. The second response (LLR2) has been reported as the most persistent one in normal subjects with an onset latency of about 50 ms [2,3,6,11]. In the present investigation, maximum LLR2 amplitude values were determined peak-to-peak on screen. Furthermore, peak latency values were determined on screen. Onset latency values were not determined, as they can be less accurately measured compared with peak latency values due to the interference with background EMG, especially after weight load as in the present investigation. After rectification, LLR2 could appear as one peak or as a bisected peak. Therefore, peak latency was defined as the half-interval between the initial ascending and the final descending parts of the peak. In a pilot study, the experiments including the above-described conditions were performed twice with the same electrode positions but interchanged input to the recording channels. As equal results were obtained in 3 subjects independent of electrode position, further subjects were investigated only once and without alternating the channel input. The mean value, the median and the standard deviation (SD) were calculated throughout. Wilcoxon test was used for statistical analyses.
1. Sine wave: y(t) = A · sin(ωt + ϕ0 ) 2. Damped sine wave z(t) = A · sin(ωt + ϕ0 ) · (e−ıt ) where A: zero to peak amplitude of undamped wave t: time ω = 2/T: angular frequency T: cycle duration ϕ0 : phase angle at t = 0 ı: damping constant of damped sine wave For sine calculations, values of A1 = 1 mV, T = 1 s For damped sine wave calculations, values of A2 = 1.55 mV, T = 1 s, ı = 1 s−1 Overlap was studied systematically by averaging 11 of these curves with equal amplitudes and phases shifted from one curve to the next by a constant angle ϕ (Fig. 2A and B), which was varied in 13 steps from 0◦ to 34.4◦ (Fig. 2C–H). A relatively wide span of overlap was chosen in order to cover the respective range that may be expected to occur when stimulus evoked motor units like those contributing to the LLR are investigated. Before averaging of single curves, the signal was either full wave rectified (by reflecting the negative parts of the signal to positive values) or left unchanged. The peak-to-peak amplitude of the resulting signal after averaging was taken from the difference of maximum and minimum values as calculated by MATLAB. 3. Results LLR2 data of 10 subjects were included in the statistical analyses. Nine of the subjects showed responses under all investigated experimental conditions. Both, rectified and non-rectified recordings from one subject are given in Fig. 1C and D. In general, LLR2 was better discernable in the non-rectified recordings than after rectification. Mean LLR2 amplitude values measured in rectified curves made up 29–34% as compared with the corresponding non-rectified LLR2 amplitude values, i.e. roughly 30%, independent of the stimulus strength and of loading weight (Fig. 1C and D). The decline in LLR2 amplitude values after rectification as compared with the according non-rectified LLR2 amplitude data (multiplied by 0.5) reached statistical significance when no weight was loaded (single and train stimuli, p < 0.05), while statistical significance was nearly reached after loading weight (single and train stimuli, p < 0.1). Train stimuli resulted in a significant increase (p < 0.005) in LLR2 amplitude values as compared with single stimuli only when non-rectified data were analysed, independent of loading weight (Fig. 1E). After rectification, no such difference was seen. When the BB muscle contraction level was increased by loading weight, LLR2 gained in amplitude both in non-rectified and rectified curves, independent of the stimulus type (p < 0.05) (Fig. 1E). Statistical significance was, however, only reached in non-rectified data. LLR2 peak latency values were significantly longer (p at least < 0.05) after rectification as compared to non-rectified data (Fig. 1F). Both, with and without rectification, LLR2 peak latency values were longer after train stimuli as compared to single stimuli. These differences reached significance only after rectification and when weight was loaded (p < 0.05; Fig. 1F). LLR2 peak latency values
86
S. Alaid, M.E. Kornhuber / Neuroscience Letters 556 (2013) 84–88
Fig. 1. Schematic illustration of averaging LLR2 responses from 4 blocks, each with 30 single curves (A, LLR2 derived from non-rectified data and B, LLR2 derived from rectified data). Representative original recordings of LLR2 over the biceps brachii muscle ipsilateral after single stimuli (C) and trains of 3 stimuli (D) delivered over the superficial radial nerve at a repetition rate of 1.0 Hz. Rect., rectified. (E) Box plots of LLR2-amplitudes obtained with single and trains of 3 stimuli obtained after rectification or without rectification after loading weight of 0.0 kg or 1.5 kg. Note that amplitudes of the rectified LLR2 waves were significantly smaller (p < 0.05) than the non-rectified LLR2 waves (multiplied by 0.5) when no weight was loaded (single and train stimuli). Train stimuli resulted in a significant increase (p < 0.005) in LLR2 amplitude values as compared with single stimuli only when non-rectified data were analysed, independent of loading weight. Box design: average (square); box area, 25–75%, and whisker interval, 5–95%. (F) Mean peak latency values of non-rectified and rectified LLR2 responses. Note the delay in latency values for rectified LLR2 waves in comparison with non-rectified LLR2 data.
obtained in rectified curves after single stimuli were significantly longer after weight load compared with no weight load (p < 0.05; Fig. 1F). For the sine wave and the damped sine wave, the relationship between average amplitude values and the correlated phase overlap (ϕ) is shown in Fig. 2I and J, both for the non-rectified curves and for the derived rectified curves. To facilitate comparison, a curve displaying 50% of the peak-to-peak amplitude values for the non-rectified data was introduced (Fig. 2I and J). With small or moderate amounts of overlap in the range of ϕ = 10–25◦ the rectified half-waves show an excess amount of cancellation as compared with the non-rectified sine waves. In fact, the rectified curves may be erased or even show a phase reversal while the shape of the corresponding non-rectified sine wave is still preserved, yet with reduced amplitude (Fig. 2E–H).
4. Discussion 4.1. Rectification and LLR2 amplitude The results of the present study show that LLR expression in averaged surface EMG recordings is strongly influenced by full wave rectification. The main and most reproducible LLR wave, i.e. LLR2 was analysed to address this question. When LLR2 was simultaneously recorded, rectification was associated with a stronger amplitude decline than without rectification. This amplitude decline was independent of both, the stimulus strength (single or train) and the muscle contraction level. The possible influence of rectification can be better understood by using synthetic curves. For this purpose, a sine curve and a damped sine curve were taken as examples. When a sine wave is
S. Alaid, M.E. Kornhuber / Neuroscience Letters 556 (2013) 84–88
Sine Wave 1 0 −1 0
0.2
0.4
87
Damped Sine Wave
0.6
0.8
1
1 0 −1 0
0.2
0.4
A 1 0 −1 0
0.2
0.4
0.6
0.8
1
1 0 −1 0
0.2
0.4
C 1 0 −1 0
0.2
0.4
0.2
0.6
0.4
0.8
1
1 0 −1 0
0.2
0.4
0.6
0.8
1
1.5 1 0.5 0 0 2
1
1
0
0 10
15
0.6
0.8
1
0.6
0.8
1
0.2
0.4
0.6
0.8
1
H
2
5
1
F
G
0
0.8
D
E 1.5 1 0.5 0 0
0.6 B
20
25
30
35
I
0
5
10
15
20
25
30
35
J
Fig. 2. The influence of phase shift on average peak-to-peak amplitude values of a sine wave and a damped sine wave. Eleven identical curves each with a constant phase shift to the neighbouring curve (A, B) were averaged either as non-rectified (thin line) or rectified waves (bold line) (C, D). The mean phase shift was modified in 13 steps from 0◦ to 34.4◦ . Thereby the 13 average curves display a prominent amplitude cancellation, both for non-rectified (E, F) and rectified curves (G, H). The average peak-to-peak amplitude data derived for the different phase shift conditions is summarized (I, J). Squares, non-rectified; stars, half-amplitude values of non-rectified data; diamond, rectified values. Note that rectification is associated with excess cancellation in the vast majority of phase shift conditions. The impact on cancellation is less prominent with the damped sine compared to the sine. For the sake of comparability, the amplitude of the damped sine was increased to A2 = 1.55 mV (sine wave, A1 = 1.00 mV).
rectified, the resulting 2 “positive” waves are expected to display half the amplitude and half the duration of the original sine wave. When after stimulus repetition the time interval between stimulus and response remains constant, averaging would result in no further change. However, when the delay between stimulus and response is variable, some overlap to subsequent stimuli will occur as in the case of the LLR. The influence of phase cancellation due to such overlap has been studied using averages of 11 identical waves with the same phase shift from one curve to the next (Fig. 2). Average amplitude values of the sine wave and the 2 positively directed half waves of the rectified sine wave are affected differently by phase shift. With small or moderate amounts of overlap, the rectified 2 half-waves show an excess amount of amplitude cancellation as compared with the non-rectified sine waves (Fig. 2). In fact, the rectified curves may be erased or show a phase reversal, while the shape of the original (non-rectified) sine wave is still preserved (Fig. 2C–H). The influence of phase shift on amplitude cancellation depends on the amount of overlap ϕ (Fig. 2I and J). Similar to these results, sensory action potentials have been shown to display a larger amount of phase cancellation after averaging than compound motor action potentials which span a broader time interval than the sensory ones [8]. However, in the case of the 2 half-waves derived by rectification from a sine wave, it is not the shortening of the half-waves compared to the original curve. Rather more or less doubling the frequency content within the same time interval leads to the excess in phase cancellation. This increase in the frequency is clearly noticeable in the rectified original recordings as compared with the non-rectified ones (Fig. 1A and B). Overlap simulation of the damped sine wave also shows phase cancellation, however, less prominently as compared to the ideal sine wave (Fig. 2). In contrast to the hypothetical situation of sinusoid curves, the shape of real motor unit action potentials usually is more complex. The influence of the actual wave form on the results
obtained after rectification is not easy to predict. Therefore, LLR measurements have been done to address this question. As a result, mean LLR2 amplitude values amounted to only 30% of those without rectification. It seems reasonable to assume that overlap from one stimulus to the next resulted in similar LLR2 amplitude cancellation as demonstrated for the hypothetical curves discussed above. On its long pathway, the LLR is processed transsynaptically by several neurons. Considerable differences in conduction times from stimulus to stimulus are consequently expected to occur. Onset values of individual motor unit potentials taking part in the LLR response were reported to cover a time interval of almost 10 ms in post stimulus time histograms (e.g. Fig. 1 in Palmer and Ashby [12]). This variability between stimulus and motor unit potentials may be associated with considerable overlap. Like in the case of the hypothetical sine wave, the increase in the frequency content presumably results in the excess decrease in LLR amplitudes after rectification, i.e. due to excess overlap of the 2 or more positive half-waves of the motor unit action potential as compared with the non-rectified signal. With overlap corresponding to ϕ = 10◦ up to ϕ = 25◦ , excess amplitude cancellation beyond the maximum value of 50% at ϕ = 0◦ was seen after rectification as compared without. Overlap in this order is not confined to LLR measurements. It may be observed as well when, e.g. startle reflexes or corticomuscular coherence is investigated. In the latter case, the usefulness of rectification has been questioned already previously [9]. Phase cancellation may also depend on the EMG background. Theoretically, overlap is to be expected between the stimulus evoked motor unit potentials that contribute to the LLR and those that are randomly recruited by voluntary drive. In a simulation study [7], it was shown that cancellation of rectified responses critically depends on motor unit recruiting levels, while no similarly strong cancellation was identified in averages of non-rectified curves. In the present investigation, the motor unit discharge rate
88
S. Alaid, M.E. Kornhuber / Neuroscience Letters 556 (2013) 84–88
was varied by weight load. Contrary to what could have been expected, LLR2 amplitudes obtained from rectified data did not show an extra decline after weight load as compared with no weight load. While Keenan and co-workers studied the influence of overlap with the background EMG activity for but 1 motor unit potential at a time [7], the number of motor units contributing to the LLR cannot be assumed to be constant when muscle force or stimulus strength is varied. In line with previously reported data [1], LLR amplitudes obtained from non-rectified data showed a significant increase in amplitude after weight load as compared with no weight load (Fig. 1E and F). Thus, the number of motor units that contribute to the LLR response displays an increase when the EMG background increases. The effect of additional cancellation due to a higher EMG background activity seems to be counteracted and even equalized by the increase in the number of motor units that contribute to the LLR. After rectification, the weight-load associated increase in LLR2 amplitudes was clearly present, yet not statistically significant. It seems possible that rectification may obscure the detection of physiological influences. This assumption is further supported by the finding that following rectification the true impact of train stimuli on LLR amplitudes could not be detected any more. In fact, LLR2 amplitude values measured in the non-rectified data were significantly larger after train stimuli as compared with single stimuli. This finding confirms previously reported results [1]. However, after averaging rectified data, no such LLR2 amplitude increase was detected (Fig. 1E and F). The train stimulus with an interstimulus interval of 3 ms presumably leads to a stepwise depolarization of the target neurons. Part of their population will receive suprathreshold excitation with each subsequent stimulus in the train. Due to such firing of motor units in the order of 3 ms intervals, an according overlap is expected to occur. Presumably, this overlap resulted in stronger amplitude cancellation after averaging rectified data compared to non-rectified data as outlined above. The influence of rectification on LLR recordings is so strong that it would be astonishing if over decades it would have escaped detection. In fact, several authors have recognized some influence of rectification on their LLR results previously. Thus, Naumann and Reiners [10] felt that “usually, peaks were more consistent on non-rectified recordings”. In a similar sense, Hallett and co-workers [5] have recommended to record both non-rectified data and rectified data. Driven by this suggestion, Sartucci and colleagues [13] reported on differences in onset latency values between non-rectified data and rectified curves (see below).
4.2. Rectification and LLR2 latency Rectification led to a statistically significant increase in the LLR2 peak latency data, regardless of the experimental condition. At a first glance, these results are in line with previous findings that a statistically significant increase in LLR2 onset latency values after rectification as compared with non-rectified data [13]. Peculiarities of the phase reversal that resulted in a latency prolongation of LLR2 were discussed [13]. No such latency prolongation was identified for the short latency reflex, however [13]. Subtle changes that are associated with phase reversal need not necessarily lead to a latency prolongation. A latency shortening would also be possible. In case of the peak latency, the situation is different. As discussed above, rectification leads to the doubling of the peak. Due to phase shifts, the 2 positive peaks frequently merge. As a consequence, the peak latency is prone to prolongate. In fact, depending on the way of peak latency measurement (see methods), the latency prolongation associated with rectification was considerably longer for the peak values obtained in the present investigation (ca. 8 ms on average) as compared with the onset values (4.4 ms on average) measured by Sartucci and co-workers [13].
5. Conclusion Long latency reflexes have been obtained in general by way of averaging rectified curves over the last decades. In the present investigation, averaging rectified LLR curves was associated with a loss of amplitude values as compared with non-rectified curves. This is presumably attributable to an excess in cancellation of motor unit potentials. Importantly, this loss of LLR amplitudes was associated with a loss of information content. Thus, after rectification, the impact of train stimuli or of weight load was missed. LLR may even completely evade detection after rectification due to a reduced signal-to-noise ratio. As one criterium for abnormality is the absence of LLR, the rectification step might lead to false diagnostic judgments in patient studies. However, averaging rectified data in principle provides the advantage to detect inhibitory influences that would not be visible when averages derived from raw data are taken. The sensitivity of the rectification step to identify inhibitory phenomena remains to be specified. Nevertheless, the choice how recorded data is treated best may depend on the aims of the study. When LLR amplitudes are in the focus, which is the rule, the use of rectification seems to be unjustified and even harmful. This conclusion is based on theoretical considerations as well as on bioelectrical measurements. Acknowledgements S.A. was supported by a stipend from the German Academic Exchange Service (PKZ: A/09/98838). We gratefully acknowledge that the data acquisition was done by the help of Olga Kolotilina. Furthermore, we like to thank Dr. rer. nat. Christine Kornhuber for valuable advice. References [1] S. Alaid, A. Zawierucha, M. Kornhuber, The electrically evoked long latency reflex of the biceps brachii muscle: the impact of train stimuli, preceding stimuli, and voluntary muscle contraction, Neurosci. Lett. 526 (2012) 91–95. [2] G. Deuschl, A. Ludolph, E. Schenck, C.H. Lücking, The relations between long-latency reflexes in hand muscles, somatosensory evoked potentials and transcranial stimulation of motor tracts, Electroencephalogr. Clin. Neurophysiol. 74 (1989) 425–430. [3] G. Deuschl, A. Eisen, Long latency reflexes following electrical nerve stimulation, Electroencephalogr. Clin. Neurophysiol. Suppl. 52 (1999) 263–268. [4] M.M. Gassel, K.H. Ott, Motoneuron excitability in man: a novel method of evaluation by modulation of tonic muscle activity, Electroencephalogr. Clin. Neurophysiol. 29 (1970) 190–195. [5] M. Hallett, A. Berardelli, P. Delwaide, H.J. Freund, J. Kimura, C.H. Lücking, J.C. Rothwell, B.T. Shahani, N. Yanagisawa, Central EMG and tests of motor control. Report of 1st IFCN committee, Electroencephalogr. Clin. Neurophysiol. 90 (1994) 404–432. [6] P.H. Hammond, Involuntary activity in biceps following the sudden application of velocity to the abducted forearm, J. Physiol. 127 (1955) 23–25. [7] K. Keenan, D. Farina, R. Merletti, R.M. Enoka, Amplitude cancellation reduces the size of motor unit potentials averaged from the surface EMG, J. Appl. Physiol. 100 (2006) 1928–1937. [8] J. Kimura, Y. Sakimura, M. Machida, Y. Fuchigami, T. Ishida, D. Claus, S. Kamejama, Y. Nakazumi, Y. Wang, T. Yamada, Effect of desynchronized inputs on compound sensory and muscle action potentials, Muscle Nerve 11 (1988) 694–702. [9] V.M. McClelland, Z. Cvetkovic, K.R. Mills, Rectification of the EMG is an unnecessary and inappropriate step in the calculation of corticomuscular coherence, J. Neurosci. Methods 205 (2012) 190–201. [10] M. Naumann, K. Reiners, Long-latency reflexes of hand muscles in idiopathic focal dystonia and their modification by botulinum toxin, Brain 120 (1997) 409–416. [11] J. Noth, K. Podoll, H.H. Friedemann, Long-loop reflexes in small hand muscles studied in normal subjects and in patients with Huntington’s disease, Brain 108 (1985) 65–80. [12] E. Palmer, P. Ashby, Evidence that a long latency stretch reflex in humans is transcortical, J. Physiol. 449 (1992) 429–440. [13] F. Sartucci, L. Bonfiglio, F. Logi, A. Pellegrinetti, L. Murri, Changes in long-latency reflexes onset latencies across full-wave rectified and non-rectified recordings, Clin. Neurophysiol. 110 (1999) 1975–1977.
Contents lists available at ScienceDirect
Neuroscience Letters journal homepage: www.elsevier.com/locate/neulet
The impact of rectification on the electrically evoked long-latency reflex of the biceps brachii muscle Ssuhir Alaid ∗ , Malte E. Kornhuber Neurologic Hospital, Martin-Luther University Halle-Wittenberg, Ernst-Grube-Str. 40, 06120 Halle (Saale), Germany
h i g h l i g h t s • • • • •
The long latency reflex (LLR) is presumed to take a transcortical pathway. The impact of rectification on electrical LLR was studied over the biceps brachii. Rectified LLR amplitudes made up only 30% of non-rectified values. Thereby influences from stimulus strength or motor unit recruitment were obscured. Increased frequency contents after rectification result in excess phase cancellation.
a r t i c l e
i n f o
Article history: Received 21 January 2013 Received in revised form 14 June 2013 Accepted 4 October 2013 Keywords: Electrical train stimuli Long latency reflex Motor unit Rectification Upper arm muscle Weight load
a b s t r a c t Long latency reflexes (LLR) were elicited electrically and obtained by full wave rectified and non-rectified data recordings in 10 healthy subjects. After single or train stimuli (sensory radial nerve; interstimulus interval 3 ms) amplitude and peak latency values were measured over the bent biceps brachii (BB) muscle, either without or with 1.5 kg weight load. After rectification, mean LLR amplitude values made up 30% of the non-rectified data, independent from the stimulus type and weight load. In the non-rectified data, a significant gain in amplitude resulted from train stimuli compared with single stimuli, and from weight load compared to no weight load. No such significant difference was detected when rectified data were analysed. Furthermore, average amplitude values of rectified and non-rectified curves were studied using 11 sine waves and damped sine waves with equal phase intervals that were varied from 0◦ up to 34.4◦ . Phase shifts ranging from 10◦ to 25◦ resulted in excess amplitude decline of rectified data compared with non-rectified data. The long and polysynaptic course that LLR information takes leads to considerable overlap of responses to subsequent stimuli. This overlap of motor unit potentials forming the LLR obviously results in excess amplitude cancellation after rectification as shown for sine and damped sine waves. Rectification leads to an increase in the frequency content of the data that renders it prone to phase cancellation. In the present study, this cancellation was harmful as it prevented detection of important factors of influence such as stimulus strength and motor unit recruitment level. © 2013 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Stimulus evoked electromyographic (EMG) activity may present with low amplitudes in the microvolt range. The signal to noise ratio can be improved by averaging. Usually a preceding step with full wave rectification is recommended for this purpose [4]. Reliability of the results obtained after rectification depends on several assumptions [7] of which a constant delay between stimulus and
Abbreviations: BB, biceps brachii; LLR, long latency reflex. ∗ Corresponding author at: Department of Neurology, Martin-Luther University Halle-Wittenberg, Ernst-Grube-Str. 40, 06120 Halle (Saale), Germany. Tel.: +49 0345 557 3340; fax: +49 0345 557 3335. E-mail address: [email protected] (S. Alaid). 0304-3940/$ – see front matter © 2013 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.neulet.2013.10.009
response is of major importance. When intact motor units are stimulated in the periphery, this precondition can be regarded as given. However, in the case of polysynaptic reflex responses like the long latency reflex (LLR) considerable jitter of reflex responses may occur. Full wave rectification has previously been recommended [3], although it had not been based on comparative analyses of averaged rectified and non-rectified data. To follow this recommendation is potentially hazardous since each step in the processing of a bioelectrical signal may be associated with a loss of information content. This study used synthetic and original bioelectric signals to assess the effect of rectification on the amplitude and latency of the EMG signal. The aim of the present study was to clarify, whether full wave rectification is a useful and necessary step prior to averaging LLR responses or represents an unjustified alteration.
S. Alaid, M.E. Kornhuber / Neuroscience Letters 556 (2013) 84–88
85
2. Materials and methods
2.3. Theoretical considerations
2.1. Subjects
In order to assess the influence of overlap on amplitude values after averaging, 2 synthetic curves were studied as substitutes for single LLR signals. The first curve was a sine wave. In addition, a damped sine wave was taken as it better reflects the configuration of motor unit potentials. Calculations were performed with MATLAB version 7.0 (The MathWorks, Natick, MA, USA).
Ten healthy subjects (hospital staff, 24–45 years old, 6 females) volunteered in the following study. Experiments started at 10 a.m. and did not last longer than 1.5 h. The subjects gave their informed consent to participate in the investigation. Ethical approval for the study was received from the Martin Luther University of HalleWittenberg.
2.2. Data acquisition and evaluation Subjects were seated comfortably to an adjustable chair. The subject’s left forearm was positioned in the horizontal plane. In order to obtain rectified and non-rectified curves simultaneously, 2 channels were used to record LLR over the short head of the biceps brachii (BB) muscle. Thus, 2 “active” electrodes were set at the middle of the belly of the short head of the BB. The 2 reference electrodes were set over the olecranon and the wet ground electrode was tied to the middle of the left forearm. Electromyographic (EMG) recordings were made with a Multiliner EMG machine (Viasys, Hoechberg, Germany) during slight voluntary contraction either without weight or after loading 1.5 kg on the hand of the investigated arm. Two different muscle contraction levels were chosen since it has been demonstrated that phase cancellation in averages of rectified surface EMG curves may depend on the motor unit recruiting level [7]. Subjects were asked to fully supinate their left forearm in order to maintain a constant level of muscle contraction. The latter was carefully controlled during each experimental condition by auditory feedback and by eye. The electrical stimuli consisted of square wave pulses with duration of 0.2 ms over the left N. radialis superficialis. Supramaximum rectangular current pulses (9–12 mA) were delivered through 2 surface electrodes with a fixed distance of 2.5 cm fastened slightly proximal to the tabatière over the radial bone at the left wrist, the cathode set proximal to the anode. Time window: 500 ms, 50 ms before and 450 ms following the stimulus. Amplification: 200 V/division. Filter settings 2 Hz, 1 kHz. Sample rate 4 kHz. LLR were analysed from an average of 120 traces after single stimuli and trains of 3 stimuli with a fixed interstimulus interval (ISI) of 3 ms delivered at a rate of 1 Hz (Fig. 1). The investigated left arm was held actively bent at 90◦ . LLR may consist of up to three waves that follow the H reflex (short latency reflex, SLR), namely LLR1, LLR2 and LLR3. The second response (LLR2) has been reported as the most persistent one in normal subjects with an onset latency of about 50 ms [2,3,6,11]. In the present investigation, maximum LLR2 amplitude values were determined peak-to-peak on screen. Furthermore, peak latency values were determined on screen. Onset latency values were not determined, as they can be less accurately measured compared with peak latency values due to the interference with background EMG, especially after weight load as in the present investigation. After rectification, LLR2 could appear as one peak or as a bisected peak. Therefore, peak latency was defined as the half-interval between the initial ascending and the final descending parts of the peak. In a pilot study, the experiments including the above-described conditions were performed twice with the same electrode positions but interchanged input to the recording channels. As equal results were obtained in 3 subjects independent of electrode position, further subjects were investigated only once and without alternating the channel input. The mean value, the median and the standard deviation (SD) were calculated throughout. Wilcoxon test was used for statistical analyses.
1. Sine wave: y(t) = A · sin(ωt + ϕ0 ) 2. Damped sine wave z(t) = A · sin(ωt + ϕ0 ) · (e−ıt ) where A: zero to peak amplitude of undamped wave t: time ω = 2/T: angular frequency T: cycle duration ϕ0 : phase angle at t = 0 ı: damping constant of damped sine wave For sine calculations, values of A1 = 1 mV, T = 1 s For damped sine wave calculations, values of A2 = 1.55 mV, T = 1 s, ı = 1 s−1 Overlap was studied systematically by averaging 11 of these curves with equal amplitudes and phases shifted from one curve to the next by a constant angle ϕ (Fig. 2A and B), which was varied in 13 steps from 0◦ to 34.4◦ (Fig. 2C–H). A relatively wide span of overlap was chosen in order to cover the respective range that may be expected to occur when stimulus evoked motor units like those contributing to the LLR are investigated. Before averaging of single curves, the signal was either full wave rectified (by reflecting the negative parts of the signal to positive values) or left unchanged. The peak-to-peak amplitude of the resulting signal after averaging was taken from the difference of maximum and minimum values as calculated by MATLAB. 3. Results LLR2 data of 10 subjects were included in the statistical analyses. Nine of the subjects showed responses under all investigated experimental conditions. Both, rectified and non-rectified recordings from one subject are given in Fig. 1C and D. In general, LLR2 was better discernable in the non-rectified recordings than after rectification. Mean LLR2 amplitude values measured in rectified curves made up 29–34% as compared with the corresponding non-rectified LLR2 amplitude values, i.e. roughly 30%, independent of the stimulus strength and of loading weight (Fig. 1C and D). The decline in LLR2 amplitude values after rectification as compared with the according non-rectified LLR2 amplitude data (multiplied by 0.5) reached statistical significance when no weight was loaded (single and train stimuli, p < 0.05), while statistical significance was nearly reached after loading weight (single and train stimuli, p < 0.1). Train stimuli resulted in a significant increase (p < 0.005) in LLR2 amplitude values as compared with single stimuli only when non-rectified data were analysed, independent of loading weight (Fig. 1E). After rectification, no such difference was seen. When the BB muscle contraction level was increased by loading weight, LLR2 gained in amplitude both in non-rectified and rectified curves, independent of the stimulus type (p < 0.05) (Fig. 1E). Statistical significance was, however, only reached in non-rectified data. LLR2 peak latency values were significantly longer (p at least < 0.05) after rectification as compared to non-rectified data (Fig. 1F). Both, with and without rectification, LLR2 peak latency values were longer after train stimuli as compared to single stimuli. These differences reached significance only after rectification and when weight was loaded (p < 0.05; Fig. 1F). LLR2 peak latency values
86
S. Alaid, M.E. Kornhuber / Neuroscience Letters 556 (2013) 84–88
Fig. 1. Schematic illustration of averaging LLR2 responses from 4 blocks, each with 30 single curves (A, LLR2 derived from non-rectified data and B, LLR2 derived from rectified data). Representative original recordings of LLR2 over the biceps brachii muscle ipsilateral after single stimuli (C) and trains of 3 stimuli (D) delivered over the superficial radial nerve at a repetition rate of 1.0 Hz. Rect., rectified. (E) Box plots of LLR2-amplitudes obtained with single and trains of 3 stimuli obtained after rectification or without rectification after loading weight of 0.0 kg or 1.5 kg. Note that amplitudes of the rectified LLR2 waves were significantly smaller (p < 0.05) than the non-rectified LLR2 waves (multiplied by 0.5) when no weight was loaded (single and train stimuli). Train stimuli resulted in a significant increase (p < 0.005) in LLR2 amplitude values as compared with single stimuli only when non-rectified data were analysed, independent of loading weight. Box design: average (square); box area, 25–75%, and whisker interval, 5–95%. (F) Mean peak latency values of non-rectified and rectified LLR2 responses. Note the delay in latency values for rectified LLR2 waves in comparison with non-rectified LLR2 data.
obtained in rectified curves after single stimuli were significantly longer after weight load compared with no weight load (p < 0.05; Fig. 1F). For the sine wave and the damped sine wave, the relationship between average amplitude values and the correlated phase overlap (ϕ) is shown in Fig. 2I and J, both for the non-rectified curves and for the derived rectified curves. To facilitate comparison, a curve displaying 50% of the peak-to-peak amplitude values for the non-rectified data was introduced (Fig. 2I and J). With small or moderate amounts of overlap in the range of ϕ = 10–25◦ the rectified half-waves show an excess amount of cancellation as compared with the non-rectified sine waves. In fact, the rectified curves may be erased or even show a phase reversal while the shape of the corresponding non-rectified sine wave is still preserved, yet with reduced amplitude (Fig. 2E–H).
4. Discussion 4.1. Rectification and LLR2 amplitude The results of the present study show that LLR expression in averaged surface EMG recordings is strongly influenced by full wave rectification. The main and most reproducible LLR wave, i.e. LLR2 was analysed to address this question. When LLR2 was simultaneously recorded, rectification was associated with a stronger amplitude decline than without rectification. This amplitude decline was independent of both, the stimulus strength (single or train) and the muscle contraction level. The possible influence of rectification can be better understood by using synthetic curves. For this purpose, a sine curve and a damped sine curve were taken as examples. When a sine wave is
S. Alaid, M.E. Kornhuber / Neuroscience Letters 556 (2013) 84–88
Sine Wave 1 0 −1 0
0.2
0.4
87
Damped Sine Wave
0.6
0.8
1
1 0 −1 0
0.2
0.4
A 1 0 −1 0
0.2
0.4
0.6
0.8
1
1 0 −1 0
0.2
0.4
C 1 0 −1 0
0.2
0.4
0.2
0.6
0.4
0.8
1
1 0 −1 0
0.2
0.4
0.6
0.8
1
1.5 1 0.5 0 0 2
1
1
0
0 10
15
0.6
0.8
1
0.6
0.8
1
0.2
0.4
0.6
0.8
1
H
2
5
1
F
G
0
0.8
D
E 1.5 1 0.5 0 0
0.6 B
20
25
30
35
I
0
5
10
15
20
25
30
35
J
Fig. 2. The influence of phase shift on average peak-to-peak amplitude values of a sine wave and a damped sine wave. Eleven identical curves each with a constant phase shift to the neighbouring curve (A, B) were averaged either as non-rectified (thin line) or rectified waves (bold line) (C, D). The mean phase shift was modified in 13 steps from 0◦ to 34.4◦ . Thereby the 13 average curves display a prominent amplitude cancellation, both for non-rectified (E, F) and rectified curves (G, H). The average peak-to-peak amplitude data derived for the different phase shift conditions is summarized (I, J). Squares, non-rectified; stars, half-amplitude values of non-rectified data; diamond, rectified values. Note that rectification is associated with excess cancellation in the vast majority of phase shift conditions. The impact on cancellation is less prominent with the damped sine compared to the sine. For the sake of comparability, the amplitude of the damped sine was increased to A2 = 1.55 mV (sine wave, A1 = 1.00 mV).
rectified, the resulting 2 “positive” waves are expected to display half the amplitude and half the duration of the original sine wave. When after stimulus repetition the time interval between stimulus and response remains constant, averaging would result in no further change. However, when the delay between stimulus and response is variable, some overlap to subsequent stimuli will occur as in the case of the LLR. The influence of phase cancellation due to such overlap has been studied using averages of 11 identical waves with the same phase shift from one curve to the next (Fig. 2). Average amplitude values of the sine wave and the 2 positively directed half waves of the rectified sine wave are affected differently by phase shift. With small or moderate amounts of overlap, the rectified 2 half-waves show an excess amount of amplitude cancellation as compared with the non-rectified sine waves (Fig. 2). In fact, the rectified curves may be erased or show a phase reversal, while the shape of the original (non-rectified) sine wave is still preserved (Fig. 2C–H). The influence of phase shift on amplitude cancellation depends on the amount of overlap ϕ (Fig. 2I and J). Similar to these results, sensory action potentials have been shown to display a larger amount of phase cancellation after averaging than compound motor action potentials which span a broader time interval than the sensory ones [8]. However, in the case of the 2 half-waves derived by rectification from a sine wave, it is not the shortening of the half-waves compared to the original curve. Rather more or less doubling the frequency content within the same time interval leads to the excess in phase cancellation. This increase in the frequency is clearly noticeable in the rectified original recordings as compared with the non-rectified ones (Fig. 1A and B). Overlap simulation of the damped sine wave also shows phase cancellation, however, less prominently as compared to the ideal sine wave (Fig. 2). In contrast to the hypothetical situation of sinusoid curves, the shape of real motor unit action potentials usually is more complex. The influence of the actual wave form on the results
obtained after rectification is not easy to predict. Therefore, LLR measurements have been done to address this question. As a result, mean LLR2 amplitude values amounted to only 30% of those without rectification. It seems reasonable to assume that overlap from one stimulus to the next resulted in similar LLR2 amplitude cancellation as demonstrated for the hypothetical curves discussed above. On its long pathway, the LLR is processed transsynaptically by several neurons. Considerable differences in conduction times from stimulus to stimulus are consequently expected to occur. Onset values of individual motor unit potentials taking part in the LLR response were reported to cover a time interval of almost 10 ms in post stimulus time histograms (e.g. Fig. 1 in Palmer and Ashby [12]). This variability between stimulus and motor unit potentials may be associated with considerable overlap. Like in the case of the hypothetical sine wave, the increase in the frequency content presumably results in the excess decrease in LLR amplitudes after rectification, i.e. due to excess overlap of the 2 or more positive half-waves of the motor unit action potential as compared with the non-rectified signal. With overlap corresponding to ϕ = 10◦ up to ϕ = 25◦ , excess amplitude cancellation beyond the maximum value of 50% at ϕ = 0◦ was seen after rectification as compared without. Overlap in this order is not confined to LLR measurements. It may be observed as well when, e.g. startle reflexes or corticomuscular coherence is investigated. In the latter case, the usefulness of rectification has been questioned already previously [9]. Phase cancellation may also depend on the EMG background. Theoretically, overlap is to be expected between the stimulus evoked motor unit potentials that contribute to the LLR and those that are randomly recruited by voluntary drive. In a simulation study [7], it was shown that cancellation of rectified responses critically depends on motor unit recruiting levels, while no similarly strong cancellation was identified in averages of non-rectified curves. In the present investigation, the motor unit discharge rate
88
S. Alaid, M.E. Kornhuber / Neuroscience Letters 556 (2013) 84–88
was varied by weight load. Contrary to what could have been expected, LLR2 amplitudes obtained from rectified data did not show an extra decline after weight load as compared with no weight load. While Keenan and co-workers studied the influence of overlap with the background EMG activity for but 1 motor unit potential at a time [7], the number of motor units contributing to the LLR cannot be assumed to be constant when muscle force or stimulus strength is varied. In line with previously reported data [1], LLR amplitudes obtained from non-rectified data showed a significant increase in amplitude after weight load as compared with no weight load (Fig. 1E and F). Thus, the number of motor units that contribute to the LLR response displays an increase when the EMG background increases. The effect of additional cancellation due to a higher EMG background activity seems to be counteracted and even equalized by the increase in the number of motor units that contribute to the LLR. After rectification, the weight-load associated increase in LLR2 amplitudes was clearly present, yet not statistically significant. It seems possible that rectification may obscure the detection of physiological influences. This assumption is further supported by the finding that following rectification the true impact of train stimuli on LLR amplitudes could not be detected any more. In fact, LLR2 amplitude values measured in the non-rectified data were significantly larger after train stimuli as compared with single stimuli. This finding confirms previously reported results [1]. However, after averaging rectified data, no such LLR2 amplitude increase was detected (Fig. 1E and F). The train stimulus with an interstimulus interval of 3 ms presumably leads to a stepwise depolarization of the target neurons. Part of their population will receive suprathreshold excitation with each subsequent stimulus in the train. Due to such firing of motor units in the order of 3 ms intervals, an according overlap is expected to occur. Presumably, this overlap resulted in stronger amplitude cancellation after averaging rectified data compared to non-rectified data as outlined above. The influence of rectification on LLR recordings is so strong that it would be astonishing if over decades it would have escaped detection. In fact, several authors have recognized some influence of rectification on their LLR results previously. Thus, Naumann and Reiners [10] felt that “usually, peaks were more consistent on non-rectified recordings”. In a similar sense, Hallett and co-workers [5] have recommended to record both non-rectified data and rectified data. Driven by this suggestion, Sartucci and colleagues [13] reported on differences in onset latency values between non-rectified data and rectified curves (see below).
4.2. Rectification and LLR2 latency Rectification led to a statistically significant increase in the LLR2 peak latency data, regardless of the experimental condition. At a first glance, these results are in line with previous findings that a statistically significant increase in LLR2 onset latency values after rectification as compared with non-rectified data [13]. Peculiarities of the phase reversal that resulted in a latency prolongation of LLR2 were discussed [13]. No such latency prolongation was identified for the short latency reflex, however [13]. Subtle changes that are associated with phase reversal need not necessarily lead to a latency prolongation. A latency shortening would also be possible. In case of the peak latency, the situation is different. As discussed above, rectification leads to the doubling of the peak. Due to phase shifts, the 2 positive peaks frequently merge. As a consequence, the peak latency is prone to prolongate. In fact, depending on the way of peak latency measurement (see methods), the latency prolongation associated with rectification was considerably longer for the peak values obtained in the present investigation (ca. 8 ms on average) as compared with the onset values (4.4 ms on average) measured by Sartucci and co-workers [13].
5. Conclusion Long latency reflexes have been obtained in general by way of averaging rectified curves over the last decades. In the present investigation, averaging rectified LLR curves was associated with a loss of amplitude values as compared with non-rectified curves. This is presumably attributable to an excess in cancellation of motor unit potentials. Importantly, this loss of LLR amplitudes was associated with a loss of information content. Thus, after rectification, the impact of train stimuli or of weight load was missed. LLR may even completely evade detection after rectification due to a reduced signal-to-noise ratio. As one criterium for abnormality is the absence of LLR, the rectification step might lead to false diagnostic judgments in patient studies. However, averaging rectified data in principle provides the advantage to detect inhibitory influences that would not be visible when averages derived from raw data are taken. The sensitivity of the rectification step to identify inhibitory phenomena remains to be specified. Nevertheless, the choice how recorded data is treated best may depend on the aims of the study. When LLR amplitudes are in the focus, which is the rule, the use of rectification seems to be unjustified and even harmful. This conclusion is based on theoretical considerations as well as on bioelectrical measurements. Acknowledgements S.A. was supported by a stipend from the German Academic Exchange Service (PKZ: A/09/98838). We gratefully acknowledge that the data acquisition was done by the help of Olga Kolotilina. Furthermore, we like to thank Dr. rer. nat. Christine Kornhuber for valuable advice. References [1] S. Alaid, A. Zawierucha, M. Kornhuber, The electrically evoked long latency reflex of the biceps brachii muscle: the impact of train stimuli, preceding stimuli, and voluntary muscle contraction, Neurosci. Lett. 526 (2012) 91–95. [2] G. Deuschl, A. Ludolph, E. Schenck, C.H. Lücking, The relations between long-latency reflexes in hand muscles, somatosensory evoked potentials and transcranial stimulation of motor tracts, Electroencephalogr. Clin. Neurophysiol. 74 (1989) 425–430. [3] G. Deuschl, A. Eisen, Long latency reflexes following electrical nerve stimulation, Electroencephalogr. Clin. Neurophysiol. Suppl. 52 (1999) 263–268. [4] M.M. Gassel, K.H. Ott, Motoneuron excitability in man: a novel method of evaluation by modulation of tonic muscle activity, Electroencephalogr. Clin. Neurophysiol. 29 (1970) 190–195. [5] M. Hallett, A. Berardelli, P. Delwaide, H.J. Freund, J. Kimura, C.H. Lücking, J.C. Rothwell, B.T. Shahani, N. Yanagisawa, Central EMG and tests of motor control. Report of 1st IFCN committee, Electroencephalogr. Clin. Neurophysiol. 90 (1994) 404–432. [6] P.H. Hammond, Involuntary activity in biceps following the sudden application of velocity to the abducted forearm, J. Physiol. 127 (1955) 23–25. [7] K. Keenan, D. Farina, R. Merletti, R.M. Enoka, Amplitude cancellation reduces the size of motor unit potentials averaged from the surface EMG, J. Appl. Physiol. 100 (2006) 1928–1937. [8] J. Kimura, Y. Sakimura, M. Machida, Y. Fuchigami, T. Ishida, D. Claus, S. Kamejama, Y. Nakazumi, Y. Wang, T. Yamada, Effect of desynchronized inputs on compound sensory and muscle action potentials, Muscle Nerve 11 (1988) 694–702. [9] V.M. McClelland, Z. Cvetkovic, K.R. Mills, Rectification of the EMG is an unnecessary and inappropriate step in the calculation of corticomuscular coherence, J. Neurosci. Methods 205 (2012) 190–201. [10] M. Naumann, K. Reiners, Long-latency reflexes of hand muscles in idiopathic focal dystonia and their modification by botulinum toxin, Brain 120 (1997) 409–416. [11] J. Noth, K. Podoll, H.H. Friedemann, Long-loop reflexes in small hand muscles studied in normal subjects and in patients with Huntington’s disease, Brain 108 (1985) 65–80. [12] E. Palmer, P. Ashby, Evidence that a long latency stretch reflex in humans is transcortical, J. Physiol. 449 (1992) 429–440. [13] F. Sartucci, L. Bonfiglio, F. Logi, A. Pellegrinetti, L. Murri, Changes in long-latency reflexes onset latencies across full-wave rectified and non-rectified recordings, Clin. Neurophysiol. 110 (1999) 1975–1977.
