Electromyography: some methodological problems and issues.

Key Words: Biofeedback, Cross talk, Electromyography, Mechanical artifact, Stimulus artifact.

Electromyographic (EMG) activity, or the electrical activity of skeletal muscles, is of considerable interest to physical therapists and other clinicians because it is a direct representation of the outflow of motoneurons in the spinal cord to the muscle as a result of voluntary or reflex activation.

Electromyography can be used by physical therapists to study biofeedback,[1] functional anatomy of muscles,[2,3] firing characteristics of motor units,[4] and excitability of motoneurons.[5] Electromyography can also be related to the amount of force developed by a muscle[6] and the reflex connection of muscles.[7] This information can then be used to guide treatment. Treatment may be aimed at either increasing the contraction force of muscles or relaxing muscles so that abnormalities in the patient's posture or gait can be corrected.[1,2]

In order to be able to obtain correct information from EMG recordings, the technique used has to be designed so as to minimize artifacts and noise. A number of textbooks[2,8] and reviews and reports[9-11] describe the basis of the EMG techniques.

The purpose of this review is to discuss some of the issues involved in recording and interpreting EMG data. A major focus of this article is to compare and contrast the advantages and disadvantages of surface electromyography (SEMG) and intramuscular EMG. Nine major aspects of the topic are covered in this article: (1) instrumentation, (2) elimination of electrical noise, (3) mechanical artifacts, (4) stimulus artifacts, (5) volume conduction (cross talk), (6) biofeedback, (7) accuracy, (8) quantification, and (9) standardization.

Instrumentation

Decisions about the instrumentation to be used when recording EMG activity include those related to electrodes, amplifiers, filters, recording and displaying devices, and equipment noise.

Electrodes

The choice of electrode depends on the muscle under study. For large superficial muscles, surface electrodes should be used. For small superficial muscles and for those that are situated under other muscles, intramuscular electrodes should be used.[2]

Surface electrodes. These electrodes can be either active or passive. Active surface electrodes (ie, surface electrodes with a built-in preamplifier) are electrodes that have the following characteristics: (1) high input resistance (1,012 [omega]) (thus, they are not sensitive to changes in the electrical resistance of the electrode-skin interface, and some people believe there is no need for skin preparation[2,10,11]), (2) built-in amplifiers that amplify the signal several times before connecting to the next amplifier, and (3) characteristics that minimize movement-related artifacts and resistance-related changes in the EMG signal.

The distinguishing feature of passive surface electrodes is that they do not have high electrical input resistance, and therefore they are affected by changes in skin resistance. The skin surface must be cleaned of dead cells, dirt, and natural oil using 70% alcohol in order to reduce the electrical resistance of the skin.[12] Skin resistance can be further reduced by rubbing the skin with an abrasive gel preparation in order to remove electrically nonconducting elements between the electrode and the muscle itself, thus improving the electrical "contact."[9,11,12]

Intramuscular electrodes. Needle or wire-based intramuscular electrodes are used in order to detect activity from muscles that are small or that are located deep within the body. Cleaning the skin surface is also essential for the insertion of a sterile wire or needle electrode into the muscle, as with all subcutaneous or intramuscular injections. Although needle electrodes are still used, wire electrodes[12,13] are most popular for kinesiological studies because they are very fine and easily implanted and removed.[2,8] In an intramuscular electrode, the extent of the active (uninsulated) area determines the extent of the intramuscular recordings.[2] For example, electrodes with large active areas using up to about 15 mm of exposed needle[14] Can record from numerous muscle fibers and, therefore, are referred to as "macro" electrodes. A small active area, achieved by exposing only the cut end of the insulated wire or tip of the needle electrode, however, is preferred when recording electrical activity from single muscle and nerve fibers.[2,15,16]

Number of electrodes to be used.

The EMG electrodes are used nearly always in a bipolar configuration.[9] In bipolar recording, two electrodes are placed on or in the muscle, and the potential between them is recorded.[2,17] Bipolar recording enables the noise-suppressing capacity of the amplifier to be fully utilized and avoids the disadvantages present in monopolar recording, and is therefore the preferred configuration.[2,9,16]

In the less popular monopolar configuration, one electrode is placed on or in the muscle, and an indifferent electrode is placed on a nonmuscular part of the body. The potential between the two electrodes is then recorded. The disadvantages of monopolar recording are (1) that it detects all electrical signals, including the noise from the vicinity of the detection area, and (2) that it is less specific when recording from a small area of the muscle.[2,16]

The Electromyography Amplifier

The amplifier should be capable of amplifying the original signal with minimal distortion; therefore, a differential amplifier with the following characteristics is recommended[2,11,9:] 1. The capacity to amplify the input

data by at least 1,000 times (up to

10,000 times if a chart recorder is

used). 2. A frequency bandwidth appropriate

for all EMG signals (ie, 20-10,000

Hz).[2] 3. An amplifier input electrical resistance

at least 10 times as high as

the electrode-to-skin electrical

resistance. For SEMG, this resistance

should be 1 M[omega] or larger.[9-11] 4. A common mode rejection ratio

(CMRR) of about 1,000:1 (equal to

60 dB) or preferably higher.[9-11]

The CMRR determines the ratio by

which the common signals are

rejected. The common inputs to

the electrodes are generally caused

by artifacts or cross talk; therefore,

the higher the CMRR ratio, the

lower the noise in an EMG

record.[2] The CMRR function of the

amplifier is never fully realized,

however, because of the unequal

source resistance in each input

terminal, or "source impedance

imbalance."[9] The problems that are

associated with this imbalance can

be improved by increasing the

input resistance of the amplifier.[9]

Principles for Filtering Electromyographic Signals

Because the frequency characteristics of SEMG and single motor unit recordings are different, filtering limits should also be different. The filter settings should not eliminate the predominant frequencies one is attempting to record in EMG.

Surface electromyography data.

For most muscles, the frequency spectrum of the SEMG data is about 20 to 500 Hz.[2,18] The frequency values may vary depending on the thickness of soft tissue covering the muscle and the fiber-type composition of the muscle.[2] The upper (ie, high-cut or low-pass) and lower (ie, low-cut or high-pass) limits of the filters, therefore, should not eliminate any of the frequencies within this range.

Intramuscular electromyography data. The frequency spectrum of motor unit potentials is much higher than that of the SEMG. Although it is suggested that a bandwidth of 20 to 10,000 Hz be used for motor unit studies,[2] it is possible to filter motor unit signals at about 500 Hz (low-cut) without significantly altering the singularities of the single motor unit action potentials. This approach eliminates most of the electrical and movement-related artifacts and does not significantly alter the recognition of the unit potentials.[19]

It is important to remember that the value that is set on the filter does not eliminate frequencies above (for high-cut filter settings) or below (for low-cut filter settings) the frequency spectrum. For example, a 30-Hz low-cut filter setting will not totally eliminate all 30-Hz signals because this filter setting can only reduce the amplitude of 30-Hz frequency component of the signal by about 70%.[8] To ensure that all 30-Hz and lower frequencies are eliminated from the EMG record, much higher low-cut filter values may have to be used.[8] Usage of higher filter settings will, however, distort the SEMG records.[8]

It has been reported that low-cut filter settings of up to 500 Hz do not alter the recognition of single motor unit potentials.[19] With such low-cut filter settings, the interference by mechanical sources and electrical noise can be totally eliminated from intramuscular EMG records.

During recording, however, the filter setting for the system should not exceed 20 Hz to ensure that relevant information is not lost.[2] The 30-Hz or higher low-cut filters can be used when monitoring the EMG signals during an experiment (on-line) or after the completion of an experiment (off-line).

Recording and Displaying Devices

During recording, the electrical signals from the muscles must be free of mechanical artifacts, electrical noise, and cross talk. If these signals are not eliminated and the permanent records contain these unwanted signals, it will be difficult to separate them after the experiment. It is possible to reduce the electrical and mechanical artifacts by filtering the recorded EMG data, but the cross talk cannot be eliminated from the record. Permanent records of EMG data are desirable because they are a means of verifying the quality of the record, should this be necessary. There are at least three methods for creating permanent records of the data: (1) use of low-frequency-response recording devices, (2) use of high-frequency-response recording devices, and (3) use of computer programs.

Low-frequency-response recording devices. These devices record low-frequency signals, and they include pen-and-chart recorders. When using such devices, the exact frequency response (ie, how fast the pen can move) of the recorder must be known because these characteristics of the recorder determine the shape of the final EMG record.

If the pen recorder is the only form of recording medium that is available, most experts agree that the EMG signal should be modified before the recording begins.[2,20] This modification procedure involves full-wave rectification and high-cut filtering[2,9,20] so that the high-frequency components of the EMG record are not lost (Fig. 1). The aim of such modification is to reduce the frequency of the signal without distorting the data significantly. For example, most pen recorders have a response frequency of around 60 Hz (ie, they will not properly record frequencies8 that are over 60 Hz). Because 60 Hz is close to the middle of the frequency power of the SEMG, the signal will lose most of its characteristic shape and will be distorted unless it is modified. Once these modifications are done, the SEMG can then be recorded with a pen recorder without the resultant distortions.[10]

High-frequency--response recording devices. These devices are capable of recording very-high-frequency signals, and they include cathode ray oscilloscopes, audio recorders, video recorders, and computer programs with high sampling frequency. Such devices are necessary for recording unmodified (ie, real) EMG data.

Due to the fact that the shapes of the motor unit action potentials have very sharp components (ie, they contain some very high frequencies), the recording medium for registering these signals should be able to record frequencies at or above 10 kHz.

Computer programs. There are many computer programs that can record and analyze EMG data, and it is not the intent of this review to describe or discuss these computer programs and their uses. The sampling rate characteristics of the recording channels of the computer system should be considered. For each channel, the sampling rate of the analog-to-digital conversion component should be at least two times the highest expected EMG frequency.[10] To record the highest SEMG components (around 500 Hz), the sampling rate for the SEMG channel should be higher than 1,000 Hz.[10] It follows, therefore, that to record the motor unit potentials from intramuscular electrodes in a different channel, the sampling rate of that channel should be higher than 20,000 Hz.[2] Unless this condition is met, the recorded data can be distorted by the computer processing.[21]

Equipment Noise

Equipment noise is generated by the very nature of the recording electrodes and the characteristics of the recording amplifier. Electrodes introduce a "thermal noise" caused by the physical property of the metals used in the electrodes.[2] This property of metals is proportional to the square root of the resistance of the detection surface and cannot be totally eliminated.[2] This noise, however, can be reduced by cleaning the electrode contacts.

The amplifiers also generate some noise because of the physical properties of the semiconductors used in amplifiers. Again, this noise cannot be totally eliminated,[2] although it can be reduced to very low levels by the use of battery-operated low-power amplifiers.[9]

The amount of noise in the recording system should be determined before using the system. This can be done by connecting the recording electrodes to an electrically nonactive tissue (eg, skin over the patella)[22] or to a known resistance and then recording the amount of noise on a recording device such as an oscilloscope, which has high-frequency-response characteristics.[9,22] The International Society of Electrophysiological Kinesiology recommends that amplifier noise be less than 5 [mu]V RMS (root mean square), measured with a source resistance of 100 k[omega] and a bandwidth of 0.1 to 1,000 Hz.[9]

Eliminating Electrical Noise

A 50- or 60-Hz noise is often observed in SEMG and intramuscular EMG records, especially when the skin surface is not properly prepared, the electrodes are not firmly attached to the skin, or monopolar recording is used.[2] This noise may appear as regular sinusoidal waves in the raw EMG traces (small arrows in Figs. 2A and 3Aa) or regularly occurring spikes when the EMG records are low-cut filtered (small arrows in Fig. 2Ba). This noise can be eliminated using ground electrodes (see Figs. 2C and 2D) and appropriate preparation of the skin surface when active electrodes are not used.

Surface Electromyography

In order to reduce the noise in SEMG, the following steps should be taken: 1. Use active surface electrodes.[10] 2. Choose a bipolar recording

technique.[9] 3. Ground the subject.[8] 4. Use very short leads.[10] The length

of the leads connecting the recording

electrodes to the amplifiers

should be kept as short as possible[10] to minimize the amount of

electrical noise being picked up

from the power sources around

the subject.

Electrical noise can be a major problem in SEMG recordings because the frequency of the noise may overlap the frequency bandwidth (ie, the range of frequencies) of the SEMG signal. The electrical activity coming from the muscle under investigation, therefore, will include noise signals unless the source of the noise is established and eliminated.

Intramuscular Electromyography

When recording intramuscularly, the frequency overlap between the actual signal and noise does not create a major problem because the single motor unit signals are largely composed of high frequencies and low-cut filtering of about 500 HZ does not significantly alter the shape of the signals.[2,4,18]

If the subject is not properly grounded, the intramuscular EMG records may contain electrical noise unless bipolar recording is used or the records are low-cut filtered to about 500 Hz.[19] Slow-wave electrical noise may interfere with the recognition of the single motor unit signal.[23] In most studies, single motor unit potentials are recognized by their amplitudes; therefore, the presence of noise can alter the size of the signal in a random manner, thereby resulting in a failure to recognize some of the correct signal potentials.[23]

Two Methods for Grounding the Subject

Whichever recording technique is used, it is desirable to minimize or preferably eliminate the noise from the EMG records. This can be done by grounding the subject, using a ground electrode that is usually placed on an indifferent (ie, nonmuscular) site.[13,17,24]

Conventional ground electrode.

These electrodes are attached usually to the skin over an area in which there are no muscles. To achieve good grounding with this type of electrode, the skin under the surface electrode should be prepared so that the electrical resistance of the skin-ground contact falls below the electrical resistance of other electrodes connected to the subject. This is achieved by thoroughly cleaning and scrubbing the skin to reduce the ground's electrical resistance, preferably to around 3 k[omega].[12] Furthermore, the ground electrode should be larger in active skin-contact area than the other electrodes on the subject.[17] Selection of the location of the ground electrode is also believed to be of some importance.[17] It has been recommended that the ground electrode be placed close to the recording electrodes[2] and preferably on a site not overlying a muscle (eg, center of the forehead, wrist, elbow, ear lobe).[17]

The use of the ground electrode becomes even more important when electrical stimulation is used. Unless the subject is properly grounded, a large stimulus artifact may distort the EMG records, even when the stimulator is isolated from the ground.[24]It has been suggested that a large electrode should be placed between the stimulating and recording electrodes to minimize the stimulus artifact.[25] Like mechanical artifacts and electrical noise, the stimulus artifact is generally a problem for SEMG recording only. The recordings obtained by intramuscular electrodes are not significantly affected by these factors.

Lip-clip electrode. A new ground electrode that readily gives a low electrical resistance contact without skin preparation is the lip-clip electrode.[24] This electrode is simply clipped onto the subject's lip and makes close contact with the mucosa, which has a very low electrical resistance. This electrode has been proven to be superior to the conventional ground electrode in several ways, including the following: (1) It does not need meticulous and often abrasive skin preparation because it clips onto the mucosa; (2) it is easily applied, and its location is independent of the position of the recording electrodes; and (3) it minimizes electrical noise and stimulus artifact at least as well as other conventional electrodes (Fig. 3).[24]

Mechanical Artifacts in Electromyographic Recording

Generally, any movement of electrodes on the skin surface relative to the skin or the wires connecting the electrodes to the amplifier generates small electrical potentials that are superimposed on the electrical signal of the underlying muscle.[17] In many instances, EMG activity is recorded in which movement of the electrodes or wires is unavoidable (eg, during movement of limbs or during study of reflex responses to mechanical perturbations delivered by a reflex hammer26 or by a moving platform[27]). The extent of the mechanical artifact should be determined for each experimental paradigm and should be minimized.

Surface Electromyography

Electrical potentials that are produced as a result of mechanical factors in the system may have slow or fast components, but usually do not exceed 30 Hz due to the fact that body tissues cannot oscillate at higher frequencies.[2] Because 30 Hz is a common frequency for the SEMG records,[9,18] the artifacts that are produced by rapid displacement of limbs may become indistinguishable from SEMG signals, especially when they occur simultaneously. Therefore, I believe these artifacts should be recognized and eliminated on the raw EMG record before any modification of the EMG is performed (Fig. 4). This is because of the fact that the genuine muscle signals and artifacts will become indistinguishable once the signal is amplified, rectified, integrated, high-cut filtered, and averaged. These modification procedures eliminate most of the details, which may help to distinguish between a genuine muscle signal and the signal due to movement artifact.

The existence and the extent of movement artifacts in SEMG records often can be shown by moving the limb passively, examining the raw (ie, unaltered) EMG signal on an oscilloscope,[10] or using electrically insulated electrodes.[28] Despite the use of these techniques, however, most artifacts may remain.

There are a number of methods for minimizing movement artifacts including taping the electrodes and their wires to the skin surface and hence limiting the movement of the electrodes,[2] placing the amplifier (or the preamplifier) within 10 cm of the recording electrodes,[2] using the shortest possible leads for connecting the recording electrodes to the amplifiers,[29] and using active electrodes.[30]

If some mechanical artifacts still remain after these measures have been taken, filtering procedures should be considered to further reduce or eliminate the artifacts. For filtering to be effective, it is necessary to know the extent of the artifact and its frequency spectrum.

Intramuscular Electromyography

Mechanical artifacts and motor unit action potentials recorded using needle or wire electrodes have distinctly different frequency ranges,[2] and therefore separating artifacts from the real signal is easier. The movement artifacts in the intramuscular EMG record can be filtered out (using low-cut 500 Hz), and amplitude discrimination can be used to recognize motor unit potentials.[19]

Stimulus Artifacts in Electromyographic Recording

A problem with some types of EMG records is the artifact caused by the mechanical and electrical stimuli that may be used to evoke reflex responses (at time zero in Figs. 3A and 3B). It is important to determine the amplitude and the duration of the stimulus artifact when the muscle under investigation is completely relaxed. Only when this is achieved is it possible to take the precautions discussed in this section. The same approach for reducing stimulus artifacts should be used in all experiments so that EMG records can be compared.

Mechanical stimuli (taps and perturbations) usually cause some movement of the patient's limbs or the electrodes. Unless the precautions described previously are taken, these perturbations may introduce artifacts into the EMG records.

To avoid electrically induced stimulus artifacts, several options are available[25]: (1) Use a system that can disconnect the input signal from the amplifier during the application of the stimulus,[31] (2) place a large ground electrode with low electrical resistance between the stimulating and recording electrodes,[25] or (3) use a lip-clip ground electrode[24] to reduce the stimulus artifacts from the EMG records.

Volume Conduction (Cross Talk) in Electromyographic Recording

Surface Electromyography

Surface electromyographic records often contain some electrical activity originating from muscles other than the muscle under investigation. This phenomenon is known as cross talk and is primarily due to volume conduction of electrical activity. The existence of cross talk in the recordings can be confirmed using muscle stimulation experiments,[32] studies in which simultaneous EMG recordings are obtained from the surface and intramuscular electrodes,[33] double differential recording techniques (Fig. 5)[32] or insulting sheets inserted between muscles (in animals).[34]

Unfortunately, the possible existence of cross talk has often been ignored. As a consequence, volume conduction in SEMG records has led to several controversies in the literature (eg, the claim that low-intensity stimulation of the posterior tibial nerve elicits an excitatory response at H-reflex latency in the SEMG activity of the tibialis anterior muscle). Some groups have suggested that this reflex is genuine,[35] whereas others have shown that cross talk is responsible for this "reflex."[33,36]

Once the extent of the cross talk and the conditions (eg, stimulus intensity, contraction strength) that initiate cross talk are known, SEMG may be used without the need for an intramuscular electrode so long as the experimental conditions stay constant. For example, if the high stimulus intensities used in a study cause cross talk (as for the H-reflex in the limb), the researchers may decide to use only low stimulus intensities. If the researchers, however, choose to use high stimulus intensities, they must use double differential recording or intramuscular electrodes to record the EMG activity with minimal cross talk.[33] Double differential recording is illustrated in Figure 5. This technique allows the cancellation of signals that arrive simultaneously to every electrode pair. Such common signals come from distant sources (ie, other muscles)[32] and are not wanted in EMG records. Muscle signals originating from the muscle of interest, however, conduct below the electrodes and generate similar, but time-delayed, EMG signals, which are not cancelled by the double differential procedure.[32] This technique, therefore, discriminates between volume-conducted signals and action potentials that propagate along the muscle fiber below the surface electrodes.[32]

Intramuscular Electromyography

The motor unit potentials are usually free from volume conduction and therefore reflect only the activity of the muscle in which the electrode is placed.[32,33]

Biofeedback

It has long been recognized that the level of activity of a muscle can be controlled by the subject if he or she is provided with feedback on the level of activity in the muscle under study (biofeedback). In most EMG studies, attempts have been made to control the level of contraction of the muscle so that voluntary relaxation, forceful muscle contractions, or controlled contractions can be obtained.[1,15,20,37]

Surface Electromyography as Biofeedback

The SEMG signal usually needs to be full-wave rectified and filtered (0-3 Hz) before it can be used as a source of feedback to help the subject maintain a constant level of muscle activation.[37] Full-wave rectification followed by low-pass filtering converts the raw SEMG signal to a quantifiable and voluntarily controllable signal that is a reasonable, but not absolute, representation of the total electrical activity in a given muscle (Fig. 1). Subjects can be given a target line on an oscilloscope screen and asked to keep the EMG signal on that line. The contention is that if the subject keeps the EMG level on target throughout the duration of training, the level of activity of the muscle will be constant. This approach has been found to be very useful and effective in many studies, for example, when reeducating and strengthening the accessory breathing muscles in an individual with high-level complete quadriplegia,[38] when correcting muscle imbalance around a joint and helping patients with hemiparesis to perform reciprocal lower-extremity exercise,[39] when promoting the use of upper-extremity muscles in patients with neurological disorders,[40] when facilitating the rate of recovery of quadriceps femoris muscle function following anterior cruciate ligament reconstruction,[1] and when improving head control and symmetrical standing in children with cerebral palsy.[41]

Three points should be considered when using the SEMG as biofeedback: (1) both the amplitude and the frequency components of the SEMG records fall at maximal voluntary contraction levels,[42] (2) the amplitude component of SEMG signals increases and the frequency component decreases with time at submaximal force levels,[43] and (3) the signals of other muscles and various artifacts may contaminate the EMG signal.

The significance of these points is that even when afl the artifacts and cross talk are minimized, the SEMG record is only a close approximation of the level of electrical activity of the muscle, and it may not directly represent the number and firing frequency of the motor units in the muscle of interest. As long as one recognizes these limitations, it is possible to use the SEMG records of large superficial muscles for the purpose of biofeedback.

Single Motor Unit Frequency as Feedback

It has recently been suggested by our group that the firing frequency of a single motor unit may be used as a form of feedback at least for the motor unit of interest.[44] Single motor unit potentials from the muscle under study are recorded using wire electrodes. The subject is then provided with the auditory and visual feedback from the firing frequency of the motor unit and asked to keep the firing of the unit at a certain level. When this was done in our study, we noticed that the firing frequencies of other motor units also stayed remarkably constant, whereas the SEMG and force levels changed quite dramatically (Fig. 6).

Although this technique requires sophisticated apparatuses and is not practical for the clinical use of EMG, it is included in this discussion to illustrate the limitations of SEMG that were mentioned previously. The frequencies of the motor units in the muscle of interest cannot be controlled reliably by SEMG feedback. Furthermore, SEMG feedback and its ability to control the level of the drive to the muscle is only approximate.

Despite its limitations, however, I believe the preferred method of biofeedback is still SEMG. The SEMG signal should be free from artifacts and noise before rectification and filtering (0-3 Hz). It is important to remember that the frequency and the amplitude of the SEMG signals change over time.[2,18,42,43] The time effect can be minimized by randomizing the experimental procedure.[45]

How Accurately Do Electromyographic Records Reflect the Activity of the Underlying Motor Units?

Surface Electromyography

The SEMG activity of a muscle may contain electrical activity coming from other muscles (cross talk) that may be synergistic or antagonistic to the muscle under investigation. Indeed, it may even include electrical activity originating from the heart and the brain.[46] Even after electrical noise, mechanical artifacts, stimulus artifacts, and cross talk are eliminated, the EMG signal may still contain some noise that cannot be eliminated.[2,9] The sources of this noise may be the electrodes and the amplifier in the recording system. The amplitude of this noise varies depending on the characteristics of the electrodes and the amplifiers used,[9] and can be confused with the genuine activity in the muscle.[46] Some researchers, therefore, have claimed that SEMG cannot be used to determine the start of the activity of a muscle[2] because the EMG level that represents the activation of a muscle is decided subjectively,[46] and hence such information may be misleading.[47]

Others, however, claim that the start or finish of the muscle activity can be determined by SEMG[48] so long as the investigator is aware of the differences in the waveforms of genuine muscle signals and the artifacts.[8] It is well known that the waveform of the genuine muscle signal is different from the waveform of the noise that is introduced by the instruments. The SEMG signals usually have a narrow frequency range, whereas noise signals have either very fast components or slow, regular undulations.[8] It is important to be aware of the range of distortions of the original waveform produced by the amplifiers, filters, and display devices.[2,8] It is claimed that even when an investigator cannot visually differentiate the waveforms, an audio monitor may help the investigator discriminate the differences and hence use the SEMG for determining the beginning or the end of the genuine electrical activity of the muscle.[8] It is also possible to use special computer programs to identify reliably the start of muscle activity.[49]

Intramuscular Electromyography

The cross-talk phenomenon does not interfere with single motor unit records.[2] Furthermore, the responses of single motor units are not contaminated by the responses of other motor units of the same muscle. Single motor unit records, therefore, are the most direct representation of connections of muscle fibers with motoneurons.

The minimal activity in a muscle can be determined by intramuscular electrodes, because the action potentials are all-or-nothing events. Two interrelated conditions, however, must be met in order to be able to use intramuscular EMG when the start of muscle activity is studied. First, the gain of the amplifier should be high enough to monitor even the smallest action potentials from distant parts of the muscle. Second, the insulation of the electrodes should be stripped by at least 3 mm from their tips[13] to maximize the area from which the action potentials can be recorded.[2,16] The active (uninsulated) area in an intramuscular electrode determines the area of recording. Therefore, in studies in which the start of muscle activity is investigated, the insulation of an intramuscular electrode should be stripped so that it can record from a very large area of the muscle.

Quantification of Electromyographic Activity

The EMG trace has many measurable properties including the amplitude or the duration of the waveforms, the frequency spectrum of the record, the shape of the motor unit action potentials, and the firing intervals of the motor unit action potentials. The question is: Which one of these properties should we use so that the activity of muscle can be quantified?

Surface Electromyography

It is not particularly straightforward to make accurate quantitative measurements from raw SEMG records. The amplitude, the duration, and the frequency of the SEMG signals are important characteristics of the EMG record. The raw SEMG signals need to be modified using full-wave rectification and high-cut filtering[11,20,50-52] (Fig. 1) before they can be used quantitatively. This modification envelops the amplitude and the frequency characteristics of the SEMG signals so that a good estimate of the number and the frequency of firing of underlying motor units is represented on the final record.[8,9,11,15]

It is possible to make quantitative measurements on full-wave-rectified traces by using integration to measure the area under the curve (Fig. 1).[8-11] The area under the curve displays a linear or near-linear relationship with the force that is produced by the muscle.[6,8]and, therefore, is the preferred method for quantifying SEMG records. There are a number of other methods that are used for quantifying SEMG, and these methods are discussed fully elsewhere.[8,9,11]

Intramuscular Electromyography

Because the action potentials from motor units are all-or-nothing events, there is no ambiguity about the measurements of the latency or the duration of the response as long as the signals are amplified at a gain that makes them clearly visible. Moreover, quantification is much easier and does not need several preparatory steps.

The sign, strength, and shape of the synaptic potential between the stimulated afferent fiber and motoneuron can also be calculated using the firing intervals of the single motor unit action potentials. We have recently shown that in a regularly firing motor unit, a reduction in the interspike interval as a result of a stimulus indicates that the stimulated afferent must have an excitatory connection with the motor unit under study. Conversely, if the interspike interval becomes longer than expected as a result of the stimulus, then the synaptic connection must be inhibitory.[15,53,54]

Normalization of Electromyographic Data

To compare EMG data obtained from different subjects or from the same subject on different days, a normalization procedure is usually considered necessary for both recording and quantifying the EMG data. Several attempts have been made to achieve such normalization, with varying degrees of success.[21,49,51,52,55,56]

Surface Electromyography

Most researchers agree that to be able to compare SEMG records, the raw SEMG records should be full-wave rectified and smoothed as shown in Figures 1 and 6.[20,50,51] This approach is based on the findings that the rectified and smoothed SEMG signal (often misleadingly referred to as an "integrated EMG"[11] signal) displays a linear or near-linear relationship to the isometric force developed by the muscle[57] and that normalizing the SEMG level to the force is a useful method for comparing EMG data.[58] For example, if the mean SEMG level for 10% maximal voluntary contraction is 90 [mu]V in one experiment and 140 [mu]V in another experiment, they can both be taken as 100% and the changes that take place in the SEMG level in each experiment can be compared quantitatively as the percentage of change.[44]

In general, the number of active motor units increases with increasing force at low force levels, whereas the frequency of firing of motor units increases at high force levels.[59] Therefore, it is reasonable to expect that the electrical activity recorded over a muscle may represent the level of force developed by the muscle. In addition to the rectified and smoothed EMG record,[60] many other characteristics of the SEMG signal have been found to relate linearly to the force developed by the muscle, including the peak-to-peak amplitude of the EMG signal[59] and integrated EMG.[61]

Researchers should be cautious about using the EMG information as the absolute measure of the force developed by the muscle, because there are many conditions in which such a relationship is not perfect.[62] The slow isometric contraction (type of contraction in which the length of the muscle stays constant) seems to be the best contraction type for illustrating the linear relationship.[60,63] Under isometric conditions, the slope of the linear relationship has been found to vary with the speed of contraction.[64]

Furthermore, this relationship has been found to be affected by muscle type.[57] According to this work, the muscles that display predominantly uniform fiber composition and the muscles that depend on frequency coding for force modulation give linear responses. Frequency coding refers to increasing the frequency of firing of the motor units that are already active rather than recruiting new motor units (recruitment coding) when the force output of the muscle is to be increased.[65,66] In contrast, the muscles that have an even fiber mix and the muscles that depend on recruitment of new motor units for force modulation (recruitment coding) give nonlinear responses.[57]

It has been suggested that time averaging (ie, taking the mean SEMG level over a set time period as 100%) of the rectified and smoothed SEMG signal reduces the intrasubject variability.[52] Before the averaging process, the time period for each predetermined cycle (eg, heel contact to heel contact) must be normalized to 100%. This averaging process reduces the noise due to the stochastic nature of SEMG from about 10% to about 3% over a 15-cycle period.[52] Normalizing the SEMG levels to the percentage of each subject's maximal voluntary isometric contraction for each muscle is also a commonly used and accepted approach for reducing variability of the EMG records.[67]

A similar normalization process is also necessary in order to make intersubject comparisons possible. This normalization process is performed by setting the mean value of each subject's SEMG activity over the predetermined cycle period to 100%. This approach has the advantage of reducing the amplitude variations among subjects so that each subject's SEMG value has the same weight in the final intersubject ensemble average.[50-52]

It is important to remember that the amplitude of the EMG data can be different from one experiment to the next, even though similar percentages of maximal voluntary contraction are used and similar recording procedures are followed. If the SEMG data are normalized, however, then it is possible to make quantitative comparisons within and among subjects.

Intramuscular Electromyography

Comparison of single motor unit data among subjects and within the same subject on different days poses a very different problem: To be able to record from the same motor unit of a muscle in the same subject on different occasions is highly unlikely. One can, however, compare the firing and synaptic characteristics of motor units that have similar recruitment thresholds under various experimental conditions.[6-70]

Summary

This article compared the properties of SEMG and intramuscular EMG. It is suggested that SEMG recordings display signal frequencies that are lower than those of intramuscular EMG recordings; therefore, this difference must be considered in the filtering, recording, and reproduction of these signals.

It is further suggested that SEMG is usually more prone to electrical artifacts, mechanical artifacts, and contamination from the activity of other muscles (both agonists and antagonists) than intramuscular EMG.

It is recommended that unless contamination by artifacts and other muscles is eliminated, SEMG records may not represent the true activity of the muscle under investigation. Surface electromyographic activity can be quantified either by rectifying and smoothing the data or by integrating the data. Intramuscular EMG data, however, can be quantified by processing the firing intervals of the single motor unit potentials.

Finally, it is suggested that in order to compare EMG records within and among subjects, the records should be normalized to a known variable. The most commonly used normalization procedure is to normalize the EMG to the level of contraction of the muscle under study.

Acknowledgment

This laboratory is supported by the National Health and Medical Research Council of Australia.

References

[1] Draper V. Electromyographic biofeedback and recovery of quadriceps femoris muscle function following anterior cruciate ligament reconstruction. Phys Ther. 1990;70:11-17. [2] Basmajian JV, De Luca CJ, eds. Muscles Alive. Baltimore, Md: Williams & Wilkins; 1985. [3] Basmajian JV. Electromyography-dynamic gross anatomy: a review. J Am Anat. 1980; 159: 245-260. [4] Nelson RM, Soderberg GL, Urbscheit NL. Alteration of motor-unit discharge characteristics in aged humans. Phys Ther. 1984;64:29-34. [5] Robichaud JA, Agostinucci J, Vander Linden DW. Effect of air-splint application on soleus muscle motoneuron reflex excitability in non-disabled subjects and subjects with cerebrovascular accidents. Phys Ther. 1992;72:176-185. [6] Bigland-Ritchie B. Electromyography/force relations and fatigue of human voluntary contractions. Exerc Sport Sci Rev. 1981;9:75-119. [7] Wolf SL, Segal RL. Conditioning of the spinal stretch reflex: implications for rehabilitation. Phys Ther. 1990;70:652-656. [8] Loeb GE, Gans C. Electromyography for Experimentalists. Chicago, Ill: The University of Chicago Press; 1986. [9] Soderberg GL. Selected Topics in Surface Electromyography for Use in the Occupational Setting.. Expert Perspectives. Washington, DC: US Department of Health and Human Services, National Institute for Occupational Safety and Health; 1992. [10] Soderberg GL, Cook TM. Electromyography in biomechanics. Phys Ther. 1984;64:1813-1820. [11] Winter DA, Rau G, Kadefors R, et al. Units, Terms, and Standards in the Recording of Electromyographic Research. Montreal, Quebec, Canada: International Society of Electrophysiological Kinesiology, Department of Medical Research, Rehabilitation Institute of Montreal; 1980. [12] Basmajian JV. Electrodes and electrode connectors. In: Desmedt JE, ed. New Developments in Electromyography and Clinical Neurophysiology New York, NY: S Karger Publishers Inc; 1973:502-510. [13] Basmajian JV, Stecko GA A new bipolar indwelling electrode for electromyography. J Appl Physiol. 1962; 17:849. [14] Stalberg E. Macro electromyography: a new recording technique. J Neurol Neurosurg Psychiatry. 1980;43:475-482. [15] Turker KS, Miles TS. Inhibitory reflexes in human masseter muscle. in: van Steenberghe D, De Laat A, eds. Electromyography of Jaw Reflexes in Man. Leuven, Belgium: Leuven University Press; 1989:237-256. [16] Andreassen S, Rosenfalck A. Recording from a single motor unit during strong effort. IEEE Trans Biomed Eng. 1978;25:501-508. [17] Davis JF. Manual of Surface Electromyography. Dayton, Ohio: Wright Patterson Air Force Base; 1959. Wright Air Development Center (WADC) Technical Report 59-184. [18] Mills KR. Power spectral analysis of electromyogram and compound muscle action potential during muscle fatigue and recovery. J Physiol (Lond). 1982;326:401-409. [19] Fuglsang-Frederiksen A, Ronegar J. The motor unit firing rate and power spectrum of electromyography in humans. Electroencephalogr Clin Neurophysiol. 1988;70:68-72. [20] Miles TS, Turker KS. Decomposition of human electromyogramme in an inhibitory reflex. Exp Brain Res. 1987;65:337-342. [21] Kleissen RFM. Effects of electromyographic processing methods on computer-averaged surface electromyographic profiles for the gluteus medius muscle. Phys Ther. 1990;70:716-722. [22] Joseph J, Nightingale A, Williams PL. A detailed study of the electrical potentials recorded over some postural muscles while relaxed and standing. J Physiol (Lond). 1955; 127: 617-625. [23] Turker KS, Miles TS, Smith N, Nordstrom MA. On-line discrimination of unit potentials on the basis of their waveforms: a new approach implemented on a personal computer. In: van Steenberghe D, De Laat A, eds. Electromyography of Jaw Reflexes in Man. Leuven, Belgium: Leuven University Press; 1989:445-451. [24] Turker KS, Miles TS, Le TH. The lip dip: a simple, low-electrical resistance ground electrode for use in human electrophysiology. Brain Res Bull. 1988;21:139-141. [25] McGill KC, Cummins KL, Dorfman LJ, et al. On the nature and elimination of stimulus artifact in nerve signals evoked and recorded using surface electrodes. IEEE Trans Biomed Eng. 1982;29:129-136. [26] Lund JP, Lamarre Y, Lavigne G, Duquet G. Human jaw reflexes. In: Desmedt JE, ed. Motor Control Mechanisms in Health and Disease. New York, NY: S Karger Publishers Inc; 1983: 739-755. [27] Woollacott M. The assessment of postural control using a moveable platform. In: Torode M, Balnave R, eds. Motor Disturbances. Sydney, New South Wales, Australia: Cumberland College of Health Sciences; 1989:55-61. [28] Miles TS, Browne E, Wilkinson TM. Identification of movement artifacts in electromyography recordings. Electromyogr Clin Neurophysiol 1982;22:245-249. [29] Soderberg GL, Cook TM. An electromyographic analysis of quadriceps femoris muscle sitting and straight leg raising. Phys Ther. 1983; 63:1434-1438. [30] Accornero N. Active surface electromyography probe and contour follower. Electroencephalogr Clin Neurophysiol. 1981;51:331-332. [31] Roby RJ, Lettich E. A simplified circuit for stimulus artifact suppression. Electroencephalogr Clin Neurophysiol. 1975;39:85-87, [32] De Luca CJ, Merletti R. Surface myoelectric signal cross-talk among muscles of the leg. Electroencephalogr Clin Neurophysiol. 1988;69: 568-575. [33] Turker KS, Miles TS. Cross-talk from other muscles can contaminate electromyographic signals in reflex studies of the human leg. Neurosci Lett. 1990; 111:164-169. [34] Loeb GE, Yee WJ, Pratt CA, et al. Cross-correlation of electromyography reveals widespread synchronization of motor units during some slow movements in intact cats. J Neurosci Methods 1987;21:239-249. [35] Gottlieb GL, Myklebust BM, Penn RD, et al. Reciprocal excitation of muscle antagonist by the primary afferent pathway. Exp Brain Res. 1982;46:454-456. [36] Etnyre BR, Abraham ID. Antagonist muscle activity during stretching: a paradox reassessed. Med Sci Sports Exerc. 1988;20:285-289. [37] van der Glas HW, Weytjens A, De Laat A, et al. The influence of clenching level on the post-stimulus electromyography complex, including silent periods, of the masseter muscles in man. Arch Oral Biol. 1984;29:51-58. [38] Morrison SA. Biofeedback to facilitate unassisted ventilation in individuals with high-level quadriplegia. Phys Ther. 1988;68:1378-1380. [39] Brown DA, DeBacher GA Bicycle ergometer and electromyographic feedback for treatment of muscle imbalance in patients with spastic hemiparesis. Phys Ther. 1987;67:1715-1719. [40] Wolf SL, LeCraw DE, Barton LA. Comparison of motor copy and target biofeedback training techniques for restitution of upper extremity function among patients with neurological disorders. Phys Ther. 1989;69:719-735. [41] Bertoti DB, Gross AL. Evaluation of biofeedback seat insert for improving active sitting posture in children with cerebral palsy. Phys Ther. 1988;68:1109-1113. [42] Bigland-Ritchie B, Johansson R, Lippold OC, et al. Contractile speed and electromyographic changes during fatigue of sustained maximal voluntary contraction. J Neurophysiol. 1983;50:313-325. [43] Moritani T, Muro M, Nagata A. Intramuscular and surface electromyogram changes during muscle fatigue. J Appl Physiol. 1986;60: 1179-1185. [44] Turker KS, Miles TS. Surface electromyography, force and single motor-unit data for inhibitory reflex responses in human masseter muscle at two levels of excitatory drive. Arch Oral Biol. 1989;34:731-737. [45] Brodin P, Turker KS, Miles TS. Mechanoreceptors around the tooth evoke inhibitory and excitatory reflexes in the human masseter muscle. J Physiol (Lond). In press. [46] Yemm R. The role of tissue elasticity in the control of mandibular resting posture. In: Anderson DJ, Matthews B, eds. Mastication. Bristol, England: John Wright & Sons Ltd; 1976:81-89. [47] Winter D. Pathologic gait diagnosis with computer-averaged electromyographic profiles. Arch Phys Med Rehabil. 1984;65:393-395. [48] Moller E. Evidence that the rest position is subject to servo-control. In: Anderson DJ, Matthews B, eds. Mastication. Bristol, England: John Wright & Sons Ltd; 1976:72-80. [49] Di Fabio RP. Reliability of computerized surface electromyography for determining the onset of muscle activity. Phys Ther. 1987;67:43-48. [50] Yang JF, Winter DA. Electromyography reliability in maximal and submaximal isometric contractions. Arch Phys Med Rehabil. 1983;64: 417-420. [51] Yang JF, Winter DA. Electromyographic amplitude normalization methods: improving their sensitivity as diagnostic tools in gait analysis. Arch Phys Med Rehabil. 1984;65:517-521. [52] Winter DA, Yack HJ. Electromyographic profiles during normal human walking: stride-to-stride and inter-subject variability. Electroencephalogr Clin Neurophystol. 1987;67:402-411. [53] Miles TS, Turker KS, Le TH. Ia reflexes and EPSPs in human soleus motor neurones. Exp Brain Res. 1989;77:628-636. [54] Miles TS, Le TH, Turker KS. Biphasic inhibitory responses and their IPSPs in soleus motor neurons in man. Exp Brain Res, 1989;77: 637-646. [55] Barker GR, Wastell DG, Duxbury AJ. Spectral analysis of masseter and anterior temporalis: an assessment of reliability for use in the clinical situation. J Oral Rehabil 1989; 16:309-313. [56] Komi PV, Buskirk ER. Reproducibility of electromyographic measurements with inserted wire electrodes and surface electrodes. Electromyography. 1970;4:357-367. [57] Woods JJ, Bigland-Ritchie B. Linear and non-linear surface electromyography/force relationships in human muscles. Am J Phys Med, 1983;62:287-299. [58] Brown SHC, Cooke JD. Amplitude- and instruction-dependent modulation of movement-related electromyogram activity in humans. J Physiol (Lond). 1981;316:97-107. [59] Milner-Brown HS, Stein RB. The relation between the surface electromyogram and muscular force. J Physiol (Lond), 1975;246:549-569. [60] Inman BT, Ralston HJ, Saunders J, et al. Relation of human electromyogram to muscular tension. Electroencephalogr Clin Neurophysiol 1952;4:187-194. [61] Soechting JF, Roberts WJ. Transfer characteristics between electromyographic activity and muscle tension under isometric conditions in man. J Physiol (Paris) 1975;70:779-793, [62] Eric H, Hans S. Electromyography, force and relaxation time during and after continuous electrical stimulation of human skeletal muscle in situ. J Physiol (Lond). 1983;39:33-40. [63] Patla AE, Hudgins BS, Parker PA, et al. Myoelectric signal as a quantitative measure of muscle mechanical output. Med Biol Eng Comput. 1982;20:319-328. [64] Kawazoe Y, Kotani H, Maetani T, et al. Integrated electromyographic activity and biting force during rapid isometric contraction of fatigued masseter muscle in man. Arch Oral Biol. 1981;26:795-801. [65] Kernell D. Recruitment, rate modulation and the tonic stretch reflex. Prog Brain Res 1976;44:257-265. [66] Petajan JH. Motor unit control in movement disorders. Adv Neurol. 1983;39:897-905. [67] Soderberg GL, Cook TM, Rider SC, Stephenitch BL. Electromyographic activity of selected leg musculature in subjects with normal and chronically sprained ankles performing on a BAPS board. Phys Ther. 1991;71:514-522. [68] Miles TS, Turker KS. Does reflex inhibition of motor units follow the "Size Principle"? Exp Brain Res, 1986;62:443-445. [69] Turker KS, Seguin JJ, Miles TS. Modulation of an inhibitory reflex in single motor unit in human masseter at different joint angles. Neurosci Lett. 1989;100:157-163. [70] Tokizane T, Shimazu H. Functional Differentiation of Human skeletal muscle. Springfield, Ill: Charles C Thomas, Publisher; 1964.