Академический Документы
Профессиональный Документы
Культура Документы
References
Rapid response
Subject collections
http://rsta.royalsocietypublishing.org/content/367/1887/357.full.
html#ref-list-1
Receive free email alerts when new articles cite this article - sign up
in the box at the top right-hand corner of the article or click here
REVIEW
AND
D ARIO F ARINA 2
1. Introduction
The motor unit consists of a motor neuron and the muscle bres innervated by
its axonal branches. The number of muscle bres it contains ranges widely across
human muscles. The muscle bres of each motor unit are intermingled with bres
of other motor units so that bres belonging to several motor units are close to
each other. The bres of a motor unit occupy a portion of the cross section of a
muscle (motor unit territory), which has an irregular round shape (Sta
lberg et al.
1995; Sta
lberg & Falck 1997; Trontelj et al. 2004).
* Author for correspondence (roberto.merletti@polito.it).
One contribution of 13 to a Theme Issue Signal processing in vital rhythms and signs.
357
358
The electrical activity associated with the contraction of muscle bres in the
motor units can be recorded using either intramuscular or non-invasive (surface)
detection systems. The difference between the two detection modalities consists
in the effect of the volume conductor that separates the muscle bres from the
detecting electrodes; this effect can be modelled as a low-pass lter whose
selectivity depends on the distance between the electrodes and the source. As
compared with surface electromyographic (EMG) recording, the insertion of
electrodes into muscles allows the detection of electric potentials close to the
muscle bres, thus the effect of the volume conductor is limited. For this reason
the action potentials of individual motor units can be easily identied from the
interference signal, at least at moderate force levels.
The invasive recording of EMG signals also allows the measure of different
types of spontaneous activity, e.g. brillation potentials due to denervation, as
well as changes in the motor unit potentials related to disorders affecting the
muscle, the neuromuscular junction or the peripheral nerves. The traditional
clinical use of intramuscular electromyography is in the diagnosis of myopathies,
diseases of the neuromuscular junction and of the lower motor neuron through
the observation of features of the detected individual action potentials (single
bre electromyography) up to the interference signal. This analysis is based on
signal features (e.g. number of turns) that differ in healthy conditions and
diseases. Most of these features are lost in surface recordings owing to the severe
low-pass ltering and diffusion due to the volume conductor, which causes the
EMG signal bandwidth to reduce from more that 1 kHz, typical of the
intramuscular signals, to less than 400 Hz, typical of the surface recordings.
Needle or wire electrodes inserted in the contracting muscle record individual
motor unit action potentials. Depending on the type of electrode used and its
location (see gure 1), the recorded action potentials can be the result of the activity of a small (13), moderate (1520) or large (more than 20) number of muscle
bres. The action potentials generated by bres belonging to the same motor unit
are not time aligned owing to the different locations of the end-plates (gure 2).
In neurophysiologic investigations, intramuscular recordings are often used to
identify the discharge times of individual motor units (De Luca et al. 1982; Nawab
et al. 2008; Marateb & McGill in press). In this application, the shape of the action
potentials is used to identify the occurrences of the discharges of the active motor
units. Although surface EMG signals can also be decomposed into individual
motor unit activities (e.g. Gazzoni et al. 2004; Holobar & Zazula 2004), the analysis
of motor units is usually more accurate using intramuscular systems owing to the
narrower action potentials. Moreover, by contrast with surface techniques,
intramuscular electromyography allows recordings from deep muscles.
The greater selectivity of intramuscular recordings compared with surface
recordings also poses some limitations. Owing to its selectivity, intramuscular
EMG recordings reect the activity of only a small number of active motor units
whose bres are close to the position of the detection site. In addition, the identied
intramuscular action potentials are not representative of all the bres in the motor
unit, and it is difcult to extract parameters related to the membrane bre properties
by using intramuscular recordings only. Finally, intramuscular recording systems
usually have a very small detection volume and thus it is difcult to reposition a
needle to detect the same motor units in repeated insertions. Surface electromyography provides a more global assessment of muscle properties and thus overcomes
Phil. Trans. R. Soc. A (2009)
359
6.0
5.0
1 mm
4.0
0.5 mm
arb. units
(a)
3.0
(b)
detection volume
(c)
2.0
1.0
(d )
100
200
300
400 m
Figure 1. The different types of needle electrodes have different recording volumes covering different
portions of a motor unit territory. (a) Single bre electrode. (b) Concentric electrode whose recording
volume has a diameter of 0.51.5 mm. (c) Monopolar electrode. (d ) Macro electrode recording from the
single bre as well as from the cannula surface: the second recording is spike-triggered averaged by the
rst over a number of discharges. The amplitude (ordinate in arbitrary units) is shown versus
recording distance for simulated action potentials recorded with different types of leading-off surfaces.
Note the low amplitude obtained with large electrodes for short electrodebre distance. The curves
for point-shaped and 25 mm electrode practically overlap (modied from Ekstedt & Sta
lberg 1973;
Sta
lberg & Falck 1997, with permission). Filled circle, point-shaped; open square, circular 25 mm; lled
square, circular 100 mm; open circle, ellipsoid 580!150 mm (short diameter along the bre); triangle,
ellipsoid 580!150 mm (long diameter along the bre).
some of these limitations. For example, muscle bre conduction velocity can be
estimated during voluntary contractions with surface electromyography (Farina &
Merletti 2004), whereas it is usually assessed only with electrical stimulation of the
muscle bres with intramuscular systems (Troni et al. 1983).
This review describes the methods used for recording and processing
intramuscular EMG signals.
2. Electrodes for EMG signal detection
Since 1929 when Adrian & Bronk (1929) proposed the concentric needle
electrode, a number of systems for the detection of intramuscular EMG signals
have been developed. The concentric needle electrode detects signals between the
Phil. Trans. R. Soc. A (2009)
360
(a)
(b)
T
B
B
T
P
P
P
1 mV
1 ms
duration
Figure 2. (a) Single bre action potentials do not sum synchronously to form motor unit action
potentials, which thus present multiple phases (P) and turns (T) (reproduced from Sta
lberg &
Falck 1997, with permission). (b) A motor unit action potential showing an abnormal jiggle. Eight
traces are superimposed. Some of the components show an abnormal jitter, also resulting in
vertical shifts. Two components (marked with B) display intermittent blocking. A late component,
satellite (S), is generated by a muscle bre that is either innervated by a slowly conducting
terminal axon or has a slower conduction velocity due to atrophy (reproduced from Trontelj et al.
2004, with permission q (2004) IEEE).
tip of a wire insulated in the cannula and the cannula. Other needle electrodes
have been proposed in more recent times (gure 1) and have been adapted to the
specic algorithms for information extraction. For example, Buchthal et al.
(1957) extended the concentric needle by placing 12 insulated wires in a slot in
the cannula. With this system it was possible to record the motor unit electrical
activity from many points in the muscle, thus analysing the decrease in signal
amplitude with distance and measuring the size of the motor unit territory.
De Luca & Forrest (1972) described the quadripolar needle, consisting of
four detecting surfaces that provide different representations of the same motor
unit activity, thus adding information for the automatic decomposition of the
EMG signal.
The size of needles used to record intramuscular EMG signals is larger than
that of the nearby muscle bres (a concentric needle, for example, has a diameter
of 0.30.5 mm), thus the insertion of the needle damages some muscle bres
generating a small local oedema. In addition, in some cases, the insertion of
needles may unintentionally damage important structures. For example, in a
study on cadavers, Haig et al. (2003) showed that when a needle was inserted into
a muscle according to standard clinical procedures, the tip was in the intended
location in only 45 per cent of the cases and within 5 mm of important structures,
such as nerves, arteries or veins, in 17 per cent of the cases.
The potential distribution in space, and therefore the signal features, is
affected by the presence of the equipotential needle surface (Gootzen et al. 1989;
Theeuwen et al. 1993; Stegeman et al. 1994). The recorded signal amplitude
depends on whether the bre is located at the same or the opposite side of the
Phil. Trans. R. Soc. A (2009)
361
362
extremely exible substrate (Farina et al. 2008). These systems have the
advantage of allowing multi-channel intramuscular electromyography with a
single insertion and with a well-dened and reproducible spatial arrangement of
the detection sites.
A few techniques have been proposed to solve the problem of the very local
nature of the information extracted from intramuscular recordings. The macro
EMG electrode (Sta
lberg 1980; Sta
lberg & Fawcett 1982; g. 1D) consists in a
modied single bre EMG needle where a portion of the isolated cannula is
exposed and acts as a macro electrode. The signal recorded from the macro
electrode is averaged using single motor unit discharges as triggers, as extracted
from the single bre recording. Although the averaged action potentials recorded
using a macro needle electrode (MAEMG), should better represent the
electrical activity of a motor unit, there is a relatively high correlation between
features of the actions potentials detected from concentric needle electrodes
(CNEMG) and MAEMG (Finsterera & Fuglsang-Frederiksen 2000).
Scanning electromyography is another technique based on triggered averaging.
This method provides an electrophysiological cross section of a motor unit
(Sta
lberg & Antoni 1980). Two intramuscular electrodes are used: a single bre
electrode for triggering and a concentric electrode for acquisition, inserted at
least 2 cm away and in the direction of the bres. The muscle is voluntarily
activated and the concentric electrode is connected to a mechanical linear
actuator, controlled by the acquisition system; the concentric electrode is pulled
out in steps of 50 mm or multiples, so that a corridor of approximately 20 mm is
investigated. The signal from the concentric electrode in each position is
averaged using the single bre action potentials as trigger and therefore outlines
the thickness of the motor unit territory in each location and direction of pulling.
3. Intramuscular EMG signal decomposition
Indwelling EMG signal decomposition is the process of identication and
classication of action potentials of individual motor units from the interference
pattern (gure 3). This process is complex owing to the variability in shape of the
action potentials generated by the same motor unit and the overlapping in time
and frequency of action potentials generated by different motor units. The
changes in shape of action potentials are usually limited for isometric constant
force contractions of short duration. For longer recordings, however, it is usually
necessary to update the templates by averaging the last N identied action
potentials to account for changes due to fatigue.
Pioneer work in EMG signal decomposition was accomplished by De Luca and
co-workers (De Luca et al. 1982; LeFever & De Luca 1982; LeFever et al. 1982).
In the last two decades, these and other researchers worked towards a robust
solution to the problem of intramuscular EMG signal decomposition (McGill
et al. 1985, 2005; Stashuk & De Luca 1989; Wellig et al. 1998; Stashuk & Qu
1998; Stashuk 2001; Katsis et al. 2007; Nawab et al. 2008; Marateb & McGill
in press). Although the solutions vary greatly for the need of operators
assistance, complexity and accuracy of the results, and computational time, some
general considerations can be drawn. In general, it is not possible to accurately
decompose EMG signals detected during very high-force contractions (more than
Phil. Trans. R. Soc. A (2009)
363
decomposition algorithm
muscle
-motor neurons
50% of the maximal force). Moreover, the number of identied motor units is
limited to 38, with peaks of 1112 with recent algorithms (Nawab et al. 2008)
or multi-channel recordings (McGill et al. 2005). Despite this limitation,
intramuscular EMG signal decomposition currently offers the possibility of
accurate in vivo analysis of the neural drive to muscles and has allowed the
construction of conceptual models of motor unit control strategies.
In classic approaches, the decomposition of a composite EMG signal requires
two steps of (i) detecting action potentials (segmentation) and (ii) recognizing
detected action potentials as members of a class (classication). For
segmentation, the characteristics that differentiate the action potential shapes
from the background noise should be explored. Various manifestation variables
can be adopted for solving this detection problem, such as amplitude, signal
derivatives, coefcients of a transformation (e.g. wavelet transform). To classify
the detected action potentials, it is essential that the action potentials produced
by the same motor unit are more similar in shape than action potentials
generated by different motor units. The differences in shape should be quantied
(feature space) in order to compare action potentials. Various features have been
used for this purpose, including morphological statistics (e.g. peak-to-peak
voltage, number of phases, duration, number of turns), Fourier transformation
coefcients, coefcients obtained from other transformations (e.g. a wavelet
basis), the time samples of the band-pass ltered signal and the time samples of
the low-pass differentiated signal (for review, see Stashuk 2001).
In most decomposition methods, the rst classication step is performed only
on action potentials that do not overlap in time. The overlapping potentials,
which form compound waveforms that do not repeat (therefore forming classes
with single elements), should be decomposed into the basic single unit potentials.
This operation is usually the most difcult in EMG signal decomposition and
Phil. Trans. R. Soc. A (2009)
364
365
V1
V2
V3
V4
V5
V6
V7
V8
+ +
skin
muscle tissue
3 fibre
motor unit
uron
depolarized zone
tor ne
-mo
uninsulated
wire tips
(b)
innervation zone
intramuscular
signals
ch 1
MU1
ch 1
MU2
2 ms
ch 15
0.5 mV 5 ms
Figure 4. (a) Schematic (not in proper scale) of a motor unit, of a pair of intramuscular wires and of
a surface electrode array, (b) intramuscular signal from two motor units (each line is the average
of 20 discharges obtained from decomposition of the wire recordings) and spike-triggered averages
of signals recorded from an electrode array placed on the skin on one side of the innervation zone of
the muscle. Each set of surface signals is the spike-triggered average of 20 discharges (120, 221,
322, etc.) to show the moving average of the surface multichannel action potential during a
sustained contraction. Muscle: tibialis anterior. Contraction level: 25% MVC. Interelectrode
distance of the surface array: 5 mm ( modied from Farina et al. 2002, with permission).
366
(c) 4.3
4.2
4.1
4.0
3.9
3.8
3.7
3.6
3.5
(d)
CV (m s 1)
(a) 20
18
16
14
12
10
8
6
4
2
0
20
40
60 80 100 120
time (s)
20
40 60 80 100 120
time (s)
Figure 5. (ad ) Estimated discharge rates and conduction velocity of two motor units during 120 s
recordings using the technique described in gure 4. Discharge rates are obtained from wire signals,
whereas conduction velocity is estimated from the surface electrode array ( reproduced from Farina
et al. 2002, with permission).
maximum and minimum slope, the separation between turns, etc.) to distinguish
a true turn from noise. The portion of signal included between two subsequent
turns is referred to as a segment and described by its duration and amplitude.
Other parameters, such as the number of turns per second, mean segment
duration and amplitude, spectral variables (e.g. mean or median frequency and
spectral moments), and coefcients of wavelet expansions are also applied.
Number and amplitude of turns as well as other variables are often reported in
scatter plots versus time or force and compared with typical plots of healthy or
pathological subjects for diagnostic inference and classication of disorders.
This work was supported by grants provided by the European Community (CyberManS, contract
no. 016712), the European Space Agency (MESM, contract no. 15097/01/NL/SH), the Italian
Space Agency (OSMA), and the Bank Foundations Compagnia di San Paolo and Fondazione CRT.
References
Adrian, E. D. & Bronk, D. W. 1929 The discharge of impulses in motor nerve bres: part II. The
frequency of discharge in reex and voluntary contractions. J. Physiol. 67, 3151.
Basmajian, J. V. & De Luca, C. J. 1985 Muscles alive: their functions revealed by electromyography,
5th edn. Baltimore, MD: William and Wilkins.
Brownell, A. & Bromberg, M. B. 2007 Comparison of standard and pediatric size concentric needle
EMG electrodes. Clin. Neurophysiol. 118, 11621165. (doi:10.1016/j.clinph.2007.01.016)
Buchthal, F., Guld, C. & Rosenfalck, P. 1957 Multielectrode study of the territory of a motor unit.
Acta Physiol. Scand. 39, 83104.
Phil. Trans. R. Soc. A (2009)
367
De Luca, C. J. & Forrest, W. J. 1972 An electrode for recording single motor unit activity during strong
muscle contractions. IEEE Trans. Biomed. Eng. 19, 367372. (doi:10.1109/TBME.1972.324140)
De Luca, C. J., LeFever, R. S., McCue, M. P. & Xenakis, A. P. 1982 Control scheme governing
concurrently active human motor units during voluntary contractions. J. Physiol. 329, 129142.
Ekstedt, J. & Sta
lberg, E. 1973 How the size of the needle electrode leading-off surface inuences
the shape of the single muscle ber action potential in electromyography. Comput. Progr.
Biomed. 3, 204212. (doi:10.1016/0010-468X(73)90006-8)
Farina, D. & Merletti, R. 2004 Estimation of average muscle ber conduction velocity from
two-dimensional surface EMG recordings. J. Neurosci. Methods 134, 199208. (doi:10.1016/
j.jneumeth.2003.12.002)
Farina, D., Crosetti, A. & Merletti, R. 2001a A model for the generation of synthetic intramuscular EMG signals to test decomposition algorithms. IEEE Trans. Biomed. Eng. 48, 6677.
(doi:10.1109/10.900250)
Farina, D., Colombo, R., Merletti, R. & Baare Olsen, H. 2001b Evaluation of intra-muscular EMG
signal decomposition algorithms. J. Electromyogr. Kinesiol. 11, 175187. (doi:10.1016/S10506411(00)00051-1)
Farina, D., Arendt-Nielsen, L., Merletti, R. & Graven-Nielsen, T. 2002 Assessment of single motor unit
conduction velocity during sustained contraction of tibials anterior muscle with advanced spike
triggered averaging. J. Neurosci. Methods 115, 112. (doi:10.1016/S0165-0270(01)00510-6)
Farina, D., Yoshida, K., Stieglitz, T. & Koch, K. P. 2008 Multi-channel thin-lm electrode for
intramuscular electromyographic recordings. J. Appl. Physiol 104, 821827. (doi:10.1152/
japplphysiol.00788.2007)
Finsterera, J. & Fuglsang-Frederiksen, A. 2000 Concentric needle EMG versus macro EMG: relation
in healthy subjects. Clin. Neurophysiol. 111, 12111215. (doi:10.1016/S1388-2457(00)00310-2)
Fuglsang-Frederiksen, A. 1981 Electrical activity and force during voluntary contraction of normal
and diseased muscle. Acta Neurol. Scand. 63, 160.
Gazzoni, M., Farina, D. & Merletti, R. 2004 A new method for the extraction and classication of
single motor unit action potentials from surface EMG signals. J. Neurosci. Methods 136,
165177. (doi:10.1016/j.jneumeth.2004.01.002)
Gootzen, T., Theeuwen, M. & Stegeman, D. 1989 The calculated inuence of a metal electrode
shaft in needle EMG recordings. Proc. 11th Ann. Congress of IEEE Eng. Med. Biol. Soc 3,
986987. (doi:10.1109/IEMBS.1989.95747)
Gydikov, A., Kosarov, D., Kossev, A., Kostov, K., Trayanova, N. & Radicheva, N. 1986 Motor unit
potentials at high muscle activity recorded by selective electrodes. Biomed. Biochem. Acta 45,
S63S68.
Haig, A., Goodmurphy, C., Harris, A., Ruiz, A. & Etemad, J. 2003 The accuracy of needle
placement in lower limb muscles: a blinded study. Arch. Phys. Med. Rehab. 84, 877882.
(doi:10.1016/S0003-9993(03)00014-5)
Holobar, A. & Zazula, D. 2004 Correlation-based decomposition of surface electromyograms at low
contraction forces. Med. Biol. Eng. Comput. 42, 487495. (doi:10.1007/BF02350989)
Katsis, C., Exarchos, T., Papaloukas, C., Goletsis, Y., Fotiadis, D. & Sarma, I. 2007 A two stage
method for MUAP classication based on EMG decomposition. Comp. Biol. Med. 37,
12321240. (doi:10.1016/j.compbiomed.2006.11.010)
LeFever, R. S. & De Luca, C. J. 1982 A procedure for decomposing the myoelectric signal into its
constituent action potentials-part I: technique, theory, and implementation. IEEE Trans.
Biomed. Eng. 29, 149157. (doi:10.1109/TBME.1982.324881)
LeFever, R. S., Xenakis, A. P. & De Luca, C. J. 1982 A procedure for decomposing the myoelectric
signal into its constituent action potentials-part II: execution and test for accuracy. IEEE
Trans. Biomed. Eng. 29, 158164. (doi:10.1109/TBME.1982.324882)
Lefurge, T., Goodall, E., Horch, K., Stensaas, L. & Schoenberg, A. A. 1991 Chronically implanted
intrafascicular recording electrodes. Ann. Biomed. Eng. 19, 197207. (doi:10.1007/BF02368469)
Malagodi, M. S., Horch, K. W. & Schoenberg, A. A. 1989 An intrafascicular electrode for recording
action potentials in peripheral nerves. Ann. Biomed. Eng. 7, 397410. (doi:10.1007/BF02368058)
Phil. Trans. R. Soc. A (2009)
368
Marateb, H. & McGill, K. In press. Resolving superimposed MUAPs using particle swarm
optimization. IEEE Trans. BME. (doi:10.1109/TBME.2008.2005953)
McGill, K., Cummins, K. L. & Dorfman, L. J. 1985 Automatic decomposition of the clinical
electromyogram. IEEE Trans. Biomed. Eng. 32, 470477. (doi:10.1109/TBME.1985.325562)
McGill, K., Lateva, Z. & Marateb, H. 2005 EMGLAB: an interactive decomposition. J. Neurosci.
Methods 149, 121133. (doi:10.1016/j.jneumeth.2005.05.015)
McNaughton, T. G. & Horch, K. W. 1996 Metallized polymer bers as leadwires and intrafascicular
microelectrodes. J. Neurosci. Methods 70, 103110. (doi:10.1016/S0165-0270(96)00111-2)
Nandedkar, S. D., Barkhaus, P. E., Sanders, D. B. & Sta
lberg, E. V. 1988 Analysis of amplitude
and area of concentric needle EMG motor unit action potentials. EEG J. 69, 561567.
Nawab, H., Wotiz, R. & De Luca, C. 2008 Decomposition of indwelling EMG signals. J. Appl.
Physiol 105, 700710. (doi:10.1152/japplphysiol.00170.2007)
Sornmo, L. & Laguna, P. 2005 Bioelectrical signal processing in cardiac and neurological
applications. London, UK: Elsevier/Academic Press.
Sta
lberg, E. 1980 Macro EMG, a new recording technique. J. Neurol. Neurosurg. Psychiatry 43, 475483.
Sta
lberg, E. & Antoni, L. 1980 Electrophysiological cross section of the motor unit. J. Neurol.
Neurosurg. Psychiatry 43, 469474.
Sta
lberg, E. & Falck, B. 1997 The role of electromyography in neurology. Electroencephalogr. Clin.
Neurophysiol. 103, 579589. (doi:10.1016/S0013-4694(97)00138-7)
Sta
lberg, E. & Fawcett, P. R. W. 1982 Macro EMG in healthy subjects of different ages. J. Neurol.
Neurosurg. Psychiatry 45, 870878.
Sta
lberg, E., Falck, B., Sonoo, M., Stalberg, S. & Astrom, M. 1995 Multi MUP EMG analysisa
two years experience in daily clinical work. Electroencephalogr. Clin. Neurophysiol. 97, 145154.
(doi:10.1016/0924-980X(95)00007-8)
Stashuk, D. 2001 EMG signal decomposition: how can it be accomplished and used?
J. Electromyogr. Kinesiol. 11, 151173. (doi:10.1016/S1050-6411(00)00050-X)
Stashuk, D. & De Luca, C. J. 1989 Update on the decomposition and analysis of EMG signals. In
Computer-aided electromyography and expert systems (ed. J. E. Desmedt), pp. 3853. Holland,
The Netherlands: Elsevier.
Stashuk, D. W. & Qu, Y. 1998 Robust method for estimating motor unit ring pattern statistics.
Med. Biol. Eng. Comput. 35, 18.
Stashuk, D., Farina, D. & Sgaard, K. 2004 Decomposition of intramuscular EMG signals. In
Electromyography: physiology, engineering and non invasive applications (eds R. Merletti &
P. Parker), ch. 3, pp. 4777. Piscataway, NJ: IEEE Press.
Stegeman, D., Gootzen, T., Theeuwen, M. & Vingerhoets, H. 1994 Intramuscular potential changes
caused by the presence of the recording EMG needle electrode. Electroencephalogr. Clin.
Neurophysiol. 93, 8190. (doi:10.1016/0168-5597(94)90070-1)
Stein, R. B., French, A. S., Mannard, A. & Yemm, R. 1972 New methods for analyzing motor
function in man and animals. Brain Res. 40, 187192. (doi:10.1016/0006-8993(72)90126-6)
Strommen, J. A. & Daube, J. R. 2001 Determinants of pain in needle electromyography. Clin.
Neurophysiol. 112, 14141418. (doi:10.1016/S1388-2457(01)00552-1)
Taylor, A. M., Steege, J. W. & Enoka, R. M. 2002 Motor-unit synchronization alters spiketriggered average force in simulated contractions. J. Neurophysiol. 88, 265276.
Theeuwen, M., Gootzen, T. & Stegeman, D. 1993 Muscle electric activity: a model study of the
effect of needle electrodes on single ber action potentials. Ann. Biomed. Eng. 21, 377389.
(doi:10.1007/BF02368630)
Troni, W., Cantello, R. & Rainero, I. 1983 Conduction velocity along human muscle bers in situ.
Neurology 33, 14531459.
Trontelj, J., Jabre, J. & Mihelin, M. 2004 Needle and wire detection techniques. In
Electromyography: physiology, engineering and non invasive applications (eds R. Merletti &
P. Parker), ch. 2, pp. 2745. Piscataway, NJ: IEEE Press.
Wellig, P., Moschytz, G. S. & Liiubli, T. 1998 Decomposition of EMG signals using timefrequency
features. Proc. EMBS 3, 14971500. (doi:10.1109/IEMBS.1998.747170)