Standards of instrumentation of EMG

Standards of instrumentation of EMG Hatice Tankisi a, ⇑ , David Burke b , Liying Cui c , Mamede de Carvalho d,e , Satoshi Kuwabara f , Sanjeev D. Nandedkar g , Seward Rutkove h , Erik Stålberg i , Michel J.A.M. van Putten j , Anders Fuglsang-Frederiksen a a Department of Clinical Neurophysiology, Aarhus University Hospital & Dept of Clinical Medicine, Aarhus University, Aarhus, Denmark b Royal Prince Alfred Hospital and University of Sydney, Australia c Department of Neurology, Peking Union Medical College Hospital, Chinese Academy of Medical Sciences, Beijing, China d Faculdade de Medicina-iMM, Universidade de Lisboa, Lisbon, Portugal e Department of Neurosciences, Centro Hospitalar Universitário de Lisboa, Portugal f Department of Neurology, Graduate School of Medicine, Chiba University, Japan g Natus Neurology, USA h Harvard Medical School, Boston, MA, USA i Department Clin Neurophysiology, Inst Neurosciences, Uppsala University, Sweden j Medisch Spectrum Twente & University of Twente, the Netherlands See Editorial, pages 235–236 article info Article history: Accepted 14 July 2019 Available online 5 November 2019 Keywords: EMG Instrumentation Filters Amplifiers Trigger and delay line EMG averaging EMG electrodes EMG artefacts Electrical safety Databases highlights Standard instrumentation ensures high quality recordings and enables comparison of results. This consensus document on ‘‘Standards of Instrumentation of EMG” is written by an expert panel. This report covers technical aspects as well as topics for optimal and standardized examinations. abstract Standardization of Electromyography (EMG) instrumentation is of particular importance to ensure high quality recordings. This consensus report on ‘‘Standards of Instrumentation of EMG” is an update and extension of the earlier IFCN Guidelines published in 1999. First, a panel of experts in different fields from different geographical distributions was invited to submit a section on their particular interest and expertise. Then, the merged document was circulated for comments and edits until a consensus emerged. The first sections in this document cover technical aspects such as instrumentation, EMG hardware and software including amplifiers and filters, digital signal analysis and instrumentation settings. Other sections cover the topics such as temporary storage, trigger and delay line, averaging, electrode types, stimulation techniques for optimal and standardised EMG examinations, and the artefacts electromyographers may face and safety rules they should follow. Finally, storage of data and databases, report generators and external communication are summarized. Ó2019 The Author(s). Published by Elsevier B.V. on behalf of International Federation of Clinical Neurophysiology. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/ licenses/by-nc-nd/4.0/). Contents 1. Introduction . . . ...................................................................................................... 244 2. Instrumentation ...................................................................................................... 244 https://doi.org/10.1016/j.clinph.2019.07.025 1388-2457/Ó2019 The Author(s). Published by Elsevier B.V. on behalf of International Federation of Clinical Neurophysiology. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). ⇑ Corresponding author at: Department of Clinical Neurophysiology, Aarhus University Hospital, Nørrebrogade 44, 8000 Aarhus C, Denmark. E-mail address: [email protected] (H. Tankisi). Clinical Neurophysiology 131 (2020) 243–258 Contents lists available at ScienceDirect Clinical Neurophysiology journal homepage: www.elsevier.com/locate/clinph 2.1. EMG hardware and software . . . . . . . . . . ............................................................................ 245 2.1.1. Amplifiers . . . . . . . . . . . . . ................................................................................. 246 2.1.2. Filters . . . . . . . . . . . . . . . . ................................................................................. 247 3. Digital instrumentation . . . . . . . ......................................................................................... 248 4. Digital signal analysis . . . . . . . . ......................................................................................... 248 4.1. Frequency analysis . . . . . . . . . . . . . . . . . . ............................................................................ 248 4.2. Turns-amplitude analysis . . . . . . . . . . . . . ............................................................................ 248 5. Instrumentation settings . . . . . . ......................................................................................... 249 6. Temporary storage, trigger and delay line . . . . . . . . . . . ...................................................................... 249 7. Signal averaging and noise reduction . . . . . . . . . . . . . . . ...................................................................... 250 7.1. Sensory nerve action potentials and somatosensory evoked potentials. . . . . . . . . . . . ......................................... 250 7.2. EMG potentials . . ............................................................................................... 251 7.3. EEG activity . . . . . ............................................................................................... 252 7.4. Practical issues when averaging signals. . ............................................................................ 252 8. Electrodes . . ......................................................................................................... 252 8.1. Electrode materials and other considerations . . . . . . . . . . . . . ............................................................ 253 8.2. Needles . . . . . . . . ............................................................................................... 253 8.3. Surface electrodes ............................................................................................... 254 9. Stimulation . ......................................................................................................... 254 10. Artefacts . . ......................................................................................................... 255 11. Safety in NCS and EMG . . . . . . ......................................................................................... 256 11.1. Electrical safety and leakage currents . . ............................................................................ 256 11.2. Implanted pacemakers, cardiac defibrillators and stimulators. . . . . . . . . . . . . . . . . . ......................................... 256 11.3. Recommendations on electrodiagnostic studies and anticoagulation . . . . . . . . . . . . ......................................... 256 12. Storage of data and databases. ......................................................................................... 256 13. Report generators. . . . . . . . . . . ......................................................................................... 257 14. External communication . . . . . ......................................................................................... 257 15. Concluding remarks . . . . . . . . . ......................................................................................... 257 Declaration of Competing Interest . . . . . . . . . . . . . . . . . ...................................................................... 257 Acknowledgements . . . . . . . . . . ......................................................................................... 257 References . ......................................................................................................... 257 1. Introduction Instrumentation and technical issues play an important role in an electromyographer ´s daily routine. At one level, everyone is familiar with artefacts and noise which may distort electrophysiological signals and can resemble results from nerves or muscles. High quality recordings are essential to the examination. In addition, it is suggested that increasingly, standardisation of instrumentation and recording will enable comparison of Electromyography (EMG) and nerve conduction studies (NCS) results within and between laboratories. Such standards and guidelines will allow a uniform practice and improve the selection of patients for research studies. Evidence-based documentation is, at present, sparse in electrodiagnostic medicine (FuglsangFrederiksen and Pugdahl, 2011). Standardisation is becoming more important in the health care system, with rapidly improving new technology enabling greater standardisation of instrumentation. Most EMG equipment is now digital and computer-based. Reports on electromyographic instrumentation have been published previously (Guld et al., 1970; Guld et al., 1974; Guld et al., 1983; Bischoff et al., 1999). The present consensus report is an update and extension of the latest report in Recommendations for the Practice of Clinical Neurophysiology: Guidelines of the International Federation of Clinical Neurophysiology (EEG Suppl. 52) (Bischoff et al., 1999). In the present report the technical aspects, i.e. amplifiers, filters and instrumentation have been covered in detail, much as in the report from 1999. However, these sections have been updated corresponding to the advanced knowledge in technology in the recent decades. Similarly, the sections on databases and storage of data, report generation and external communication have required considerable updating because of developments in computer science. In the former report, signal averaging and noise reduction were mentioned briefly, only for recordings of SNAPs. In the present document we added evoked potentials, EMG and EEG signals, as well as some practical aspects. In the previous document, the safety section was only mentioned briefly and was limited to electrical safety. In the present report we now also discuss the safety issues of implanted pacemakers, cardiac defibrillators and stimulators and classical and novel anticoagulants in the light of the recent literature. This report addresses instrumentation for common diagnostic tests – e.g., EMG, nerve conduction, evoked potentials. It was not intended to be a comprehensive review of techniques for which the instrumentation is used, but of necessity aspects of some techniques are considered, focusing on procedures of established clinical value. We therefore mention some of the tests in more detail and in the relevant sections some new developments in neurophysiology have also been covered. There have been considerable developments not only in the technological improvement of EMG equipment, but also in imaging methods and genetics. EMG remains the primary and most commonly used method in routine clinical practice, even in developed countries, but imaging and genetics can provide complementary data. Among the imaging methods, ultrasonography has gained an established place in EMG laboratories supplementing electrophysiological studies. For the generation of this consensus document, a panel of experts, in different fields and from different geographical regions, was invited to submit a section on their particular interest and expertise. Then, the merged document was circulated for comments and edits until this consensus was achieved. This report does not present clinical practice guidelines, and a search strategy in e.g. PubMed was not appropriate for this topic. 2. Instrumentation Contemporary EMG machines have a dedicated hardware unit with amplifiers, stimulators, control panel and a separate com244 H. Tankisi et al. / Clinical Neurophysiology 131 (2020) 243–258 puter. This configuration allows the computer to be upgraded or replaced, while keeping the EMG hardware unit. In the EMG hardware, the amplifiers and stimulators are the most important parts. Some hardware includes an ultrasound facility. Often ignored, but an important aspect of any machine is its ergonomics and control panel functions, including the foot switch and hand controls. When setting up a machine, it should be tuned optimally, preferably by a team including representatives from the users (physicians, technicians, local engineers) and manufacturers. When considering which machine to purchase, these details must be considered, since they influence medical quality, operability and, in the long term, cost efficiency. Computer hardware (A) Number of screens: Multiple screens allow EMG signals and data to be shown simultaneously with other data, such as the referral or radiology. (B) CPU speed and RAM memory size: The minimum limits are specified by the EMG equipment manufacturer. The requirements depend on the complexity of software. (C) Hard drive size: The minimum limits are specified by the EMG equipment manufacturer. If data is to be stored locally, the minimum size needs to be sufficient for the estimated number of stored patients. Greater memory is usually needed at some time, so anticipation of this is recommended. (D) Loudspeaker for replay of EMG signals without EMG hardware unit. A good sound quality is essential. (E) Printer: Local and/or network connected. Software (A) Analysis software (a) Available tests, workflow and various other features should optimize workflow and be compatible with local practice and reference limits. (b) Help function such as strategies and signal quality control. (c) Reference limits & Prepare for ongoing collection of reference material & Result presentation; Tables, signals, text, color & Compatibility with used algorithms (d) Voice control may be used for some functions (B) Remote viewer: View ongoing (live) recordings remotely from another computer. (C) Database support: Ensure the database engine used by the EMG equipment is supported and compatible with the server to be used. (D) Operating system The type and version options are specified by the EMG equipment manufacturer. (E) Reporting (a) Multiple designable templates that comply with requirements from all referring sources. (b) Digital delivery of text, signals, tables 2.1. EMG hardware and software The primary function of the electrodiagnostic system is to faithfully record and analyse various biological signals. It is important to have an optimal ‘signal to noise ratio’, i.e. amplify the neurophysiological signal voltage while attenuating background noise. This is done using analogue hardware and digital signal processing techniques (Fig. 1). The signal and noise are recorded by surface or needle electrodes. It is carried to the amplifier input via electrode leads or cable. These components behave like an antenna and may add more noise. A differential amplifier magnifies the signal while attenuating the unwanted noise, aided by analogue filters. The amplified signal is measured using an analogue-to-digital convertor (ADC) and the voltage values stored as an array of numbers. This digitised signal allows further computerized analysis. Some algorithms reduce the noise, e.g., digital filters, averaging, smoothing, etc., and others make measurements such as latency, amplitude and area in NCS. More sophisticated algorithms can detect MUPs in needle EMG. Signal characteristics are also assessed from the signal sound, generated either with analogue hardware or using the digital technology. An EMG machine also offers stimulation devices to excite nerves and muscles. These may generate electrical, visual or auditory stimuli. External devices providing other forms of stimulation, e.g. magnetic field, contact heat, reflex hammer, etc. can be interfaced to provide timing signals through so-called ‘triggers’. To achieve this, some instruments pass the digitized signals through a ‘digital-to-analogue convertor (DAC)’, to convert digital signals (with much less noise) into analogue form. This can be used for research where the investigator wants to re-sample the signals and develop algorithms for their own analysis. Fig. 1. The organization of various components and accessories to an electrodiagnostic system are shown schematically. H. Tankisi et al. / Clinical Neurophysiology 131 (2020) 243–258 245 The EMG machine displays signals, measurements, and the settings of the amplifier and stimulator. The latter can be changed using a dedicated control panel or using software commands via a mouse or computer keyboard. Data collection can be initiated using the foot switch. The software is responsible for signal processing and for generating reports. Databases can be created, and remote review used for second opinions or to help with interpretation. Constant changes in operatingsystems and in regulations governing patient information protection can make this a challenging task. The electrodiagnostic systems can also record nonneurophysiological events. Temperature, for instance, should be measured and recorded during NCS. A ‘patient response’ unit may be used to record the number of times the test subject acknowledges different types of stimuli in cognitive function assessments. Video cameras may be integrated into the system to observe the patient’s behaviour during a study, e.g., focusing on the checkerboard pattern in a visual evoked potential investigation. Recently we have seen the addition of ultrasound imaging probes to the device. The handling of these inputs is very different from that of the neurophysiological potentials, and is outside the scope of the current discussion. 2.1.1. Amplifiers The amplifier is perhaps the most critical component for the quality of the electrodiagnostic system. Selective amplification of a neurophysiological potential while attenuating background noise can be accomplished using a ‘differential’ amplifier (DA) (Webster, 1998). DA requires inputs or connections from three electrodes (Fig. 2). In past the electrodes were called ‘G1 0 , ‘G2 0 and ‘ground’. The terms G1 and G2refer to the grids of vacuumtubes used in old amplifiers which are no longer available. Later these inputs were called ‘active’, ‘reference’ and ‘ground’. The term ‘ground’ is confusing. The ‘ground’ in electrodiagnostic recording refers to a point on the amplifier circuit that is used as a point of reference for voltage measurement. Outside electrodiagnostics, it is also used to describe one of the connections in the power supply and wall outlets. The ‘reference’ electrode is presumed to be electrically silent but does record large volume-conducted potentials, such as the electrocardiogram (ECG). Given the confusion of terms and their origins, the terms ‘E1 0 , ‘E2 0 and ‘E0 0 are recommended for the three connections to the amplifier (Robinson et al., 2016). On most systems these inputs are colour-coded as black (E1), red (E2) and green (E0). The amplifier does not amplify the voltage at E1 or E2 inputs. It magnifies their difference, and hence it is called a differential amplifier. Fig. 2A shows the E1 electrode (a monopolar needle) recording a 50l V fibrillation potential. The E2 is a surface electrode placed on the skin surface and for simplicity it is assumed that it records no electrical activity, i.e. 0 l V. Their difference is amplified and the fibrillation potential is seen as a 50,000 l V signal. This amplification by 1,000 is the ‘differential’ gain of the amplifier. The ambient noise is also recorded by both electrodes and their cables. Here the ‘common’ noise is 1,000 l V, but the difference between signals at E1 and E2 is zero, and the noise will not be seen at the amplifier output. Similarly, the very large ECG potential can be eliminated by differential amplification. This enables the selective amplification of the small neurophysiologic signal in the presence of high-amplitude noise. The example reflects an ‘ideal’ DA. In practice the ‘common signal’ at E1 and E2 inputs is also amplified, but much less. In Fig. 2B, the noise voltage at the output is 500 l V. The ratio of output to input noise voltage, produces a ‘common mode gain’ of 0.5 for the amplifier. In this example, the signal-to-noise ratio at the amplifier input is 0.05 and would make it difficult to recognize the fibrillation potential. However, at the output of the amplifier the signal-tonoise ratio is 100, and this would allow it to be recognized quite easily. A DA should have a high differential gain and low common mode gain. These properties are defined in a single characteristic called the common mode rejection ratio (CMRR). It is reported in units of decibels and calculated as CMRR ðdBÞ¼20 Log ðDifferential Gain=Common mode gainÞ In our schematic amplifier, the CMRR is 66 dB. Modern electrodiagnostic systems have amplifiers with CMRR exceeding 100 dB. It is important to note that the CMRR decreases at higher frequencies. Most vendors specify the value at 50 or 60 Hz (i.e., power line frequency). Another characteristic of the amplifier is the input voltage range. As example, if the range is 50 to + 50 l V, then signals with amplitudes between these ranges can be handled without distortion. If the signal amplitude is outside the range, it will saturate the amplifier and the true signal amplitude cannot be measured. This is recognized from the ‘clipped’ peaks of the signals on the display. So, the amplifier range should be set higher than the amplitude of signals recorded in the test. In sensory NCS, the stimulus artefact can be much bigger than the nerve response. If the Fig. 2. The operation of the differential amplifier is illustrated. (A) Differential signal inputs are amplified. (B) The common signals (here noise in B) are attenuated. (S: Signal amplitude, N: Noise amplitude). 246 H. Tankisi et al. / Clinical Neurophysiology 131 (2020) 243–258 amplifier saturates, it can generate a long duration artefact that interferes with the sensory potential to be recorded. The range is controlled in the software to avoid such distortion. The electronic components of the amplifier also produce some noise This is addressed by the E1, E2 and E0 electrodes being connected together and the amplifier output measured. The peak-peak amplitude or root mean square (RMS) value of the signal are reported. RMS is usually less than 1 l V, but also depends on the amplifier range and the filter settings. High amplifier range (i.e. low gain setting) and short band width will give lower noise. The amplifier is also characterized by its ‘input impedance’. As the amplifier needs a tiny amount of current to measure a voltage, the input impedance needs to be several orders of magnitude larger than the input impedance of the generator of the voltage, i.e. muscle, nerve, body fluids. Essentially, the voltage measured (Vm) is given by: Vm = Vsource*(Ropamp)/(Rsource + Ropamp). With Rsource the ‘internal impedance’ is meant. If Ropamp = Rsource, then Vm = 0.5*Vsource. Without going into details of circuit analysis, the amplifier impedance should be high. Low impedance makes the system more sensitive to environmental noise. It may also underestimate signal amplitude. Fortunately, modern systems report impedances in excess of 100 to 1000 megaX . Just like CMRR, the impedance decreases at higher frequencies. Modern systems offer ‘switching amplifiers’. The unit provides a ‘head box’ with many input connections. The user can select the inputs in software to select any pair of inputs to make a recording. This facility is used mainly for evoked potential studies where a small set of electrodes is used to create multiple channels of recordings. These channels usually have a much lower CMRR. Such channels may not be suitable for recording signals with high frequencies (e.g., needle EMG) or when the electrodes differ in their impedances (surface versus needle). To facilitate appropriate settings of the amplifier, the signals characteristics of different test procedures are summarized in Table 1. The best strategy for high quality recordings is to reduce the ambient noise and to ensure that the noise is not different on E1 and E2 electrodes. 2.1.2. Filters Ideally, our measurement system should reproduce the signal of interest as exactly as possible while rejecting undesired signals. In clinical practice, however, we typically obtain a mixture of signals and ‘noise’, where the latter refers to any signal that does not contain relevant information for our diagnostic procedure. The undesired signal components can result from ambient power line noise (50 Hz or 60 Hz) or movement artefacts, but can also include EMG signals in a frequency range that is outside the region of interest for a particular procedure. For example, in recording singlefibre action potentials, the signal of interest is in the frequency range 0.5–5 kHz, where lower-frequency components (such as distant EMG potentials) can be safely suppressed. While filtering refers to any process where irrelevant signals are suppressed, in most clinical situations filtering is limited to attenuation of particular frequency components in the signal. The name of a filter is then defined by the frequency values that the filter attenuates (low or high frequency filters) or passes (high-pass filter, bandpass filter). The amount of attenuation depends on the ‘‘steepness” of the filter (expressed as the attenuation in dB/octave) and the cut-off frequency. The cut-off frequency is defined as the frequency where the original signal’s amplitude is attenuated by 3 dB. In today’s equipment, all filters (except for the anti-aliasing filter) are digital, which also allow filtering of frequency components with minimal phase distortion of the signals. We will discuss the most common frequency filters. Low-frequency filters (LFF; high pass filters) Low-frequency (or, synonymously, high-pass) filters attenuate low frequency components in the signal. An increase in the lowfrequency cut-off causes initial amplitude loss of slowly changing signals, waveform distortion, but more importantly it also decreases the latency to the peak of the waveform and can introduce artefacts (i.e. a tail of the motor unit action potential). When recording motor unit potentials (MUP), the duration as well as the amplitude decreases when the cut-off is increased up to 500 Hz. Using a 500 Hz cut-off the contribution from distant muscle fibres is attenuated because of the soft tissue itself acts as a highfrequency filter. Ideally the low limiting frequency should be one decade (factor 10) lower than the lowest frequency of the signal, to make sure that a phase shift, if present at all, does not affect latencies. Movement artefacts contain slow frequencies. In some cases, the only way to remove this artefact is to increase the lower limiting frequency. This is commonly done for surface EMG recording of movements (e.g., tremor recordings; gait) and may be necessary for motor evoked potential recordings to transcranial magnetic stimulation (TMS). It is usually required if EMG traces are rectified before averaging, e.g., for averaged F waves. High-frequency filters (HFF; low pass filters) High-frequency filters attenuate high frequencies. A decrease in the high-frequency cut-off reduces the amplitude and rise time. If using a high-frequency cut-off that is too low, the system will not be able to record adequately the rise of the potential (containing the highest frequencies of the signal), and this may (i) lower the amplitude, (ii) reduce the number of phases and (iii) prolong the duration of the main peak component of the signal. An inappropriate HFF can also affect the measurement of onset latency of a potential. It will be prolonged because the abrupt decline from the baseline is missed and more time elapses before the beginning of the potential can be appreciated. Band-pass filters Most filters used are band-pass filters, a combination of a high and low frequency filter. The effects of changing the frequency range of a bandpass filter on recording a MUP is illustrated in Fig. 3. Notch filter A notch filter is a special type of band stop filter. In electrophysiology it is normally designed to reduce power line interference (50 Hz or 60 Hz). Ideally, it should not be used because most neurophysiological signals contain significant components at this frequency, and their use may hide interesting components. In Table 1 Typical signal amplitudes, filter settings and sampling frequency, f s (Nilsson et al., 1993) in different electrodiagnostic test procedures. Filter settings are defined by the frequencies of the signal of interest. Recording Signal amplitude (peak-peak 2 ) Typical filters Sampling frequency, f s (kHz) Compound muscle action potential 0–50 mV 1 Hz–5 kHz 20 Sensory nerve action potential 0–100 uV 10 Hz–5 kHz 20 Needle EMG 0–30 mV 2 Hz–10 kHz 50 Single fibre EMG 0–50 mV 500 Hz–10 kHz 50 Surface EMG 0–10 mV 1 Hz–1 kHz 5 Somatosensory evoked potentials 0–50 l V 30–3000 Hz 20 Visual evoked potential 0–0.5 mV 1–100 Hz 1 Auditory evoked potentials 0–50 l V 100 Hz–3 kHz 10 Cognitive evoked potentials 0–50 l V 0.1–200 Hz 1 Sympathetic skin response 0–2 mV 0.1–100 Hz 1 Electrocardiogram 1 0–5 mV 10–100 Hz 2 Electroencephalogram 0–300 l V 1–200 Hz 500 Hz 1 For heart rate variability studies. 2 Peak to peak signal amplitude also includes the background signal level. H. Tankisi et al. / Clinical Neurophysiology 131 (2020) 243–258 247 addition, the phase changes abruptly back and forth around the notch frequency, and this may distort the waveform. 3. Digital instrumentation EMG equipment uses digital computers for data sampling, storage and signal processing. After the analogue signal has been amplified, the analogue-to-digital (AD) converter discretizes the signal in both time and amplitude, and assigns a digital value to the amplitude at defined time points. This assignment of the amplitude to a digital value is performed by using a finite number of digital amplitude values. In this conversion process, two important criteria must be satisfied. First, the sampling frequency should be sufficiently high to reliably represent the original analogue signal. Second, the digitization of the amplitude should be sufficiently fine to accurately represent the amplitude of the original signal in the digital domain. These two constraints are illustrated in Fig. 4, digitizing an analogue signal, consisting of a sine wave with frequency 1 Hz and amplitude 1 mV using different sampling rates and AD converters. The phenomenon illustrated in Fig. 4, panel B is known as ‘‘aliasing”: if the sampling frequency, fs, is lower than the highest frequency present in the signal of interest, fmax, the frequency obtained has an erroneous value, in this example 0.1 Hz. Using appropriate algorithms, it is possible to reconstruct the waveform in detail if the sampling rate is more than twice the highest-frequency component of the waveform (Nyquist theorem). In practice, the sampling frequency used is typically 2–5 times the highest frequency component in the signal of interest. In order to guarantee that the maximum frequency in the signal is known, an analogue ‘‘anti-aliasing” filter is used before the signal is digitized. Table 1 presents typical values. The required number of bits for the AD converter is defined by the desired amplitude resolution and the maximum amplitude of the signal. Current AD converters are 24 bits or more, which is more than sufficient for most applications (van Putten, 2009). 4. Digital signal analysis To complement visual analysis of the EMG, various quantitative tools can be used. It is noted, however, that an experienced neurophysiologist can interpret most findings for clinical diagnostic purposes with careful visual inspection and auditory assessment. For research purposes, however, quantitative EMG (qEMG) is extensively used. In some laboratories, qEMG is used routinely for clinical purposes. 4.1. Frequency analysis When the subject is exerting a constant force, the surface EMG can be used for spectral analysis, creating the power spectrum density. During continued muscle activity, mean and median frequencies typically decrease. Using needle EMG and constant force, frequency analysis of MUPs has shown that the frequency distribution is shifted towards higher frequencies in myopathies and towards lower values in patients with neurogenic disorders (Fuglsang-Frederiksen, 2000). 4.2. Turns-amplitude analysis In this technique, the amplitude of the interference pattern is plotted versus the number of turns for different levels of contraction. A ‘‘turn” is defined as a reversal of voltage of > 100 l V, and if the interference pattern has many turns it will look and sound spiky. By comparing the distribution with normal values, myopathic and neurogenic patterns can be differentiated (FuglsangFrederiksen et al., 2016). Fig. 3. Effect of different bandwidths of band-pass filters on shape, amplitude and duration of a motor unit action potential. From Bischoff et al., Standards of instrumentation of EMG, 1999. 248 H. Tankisi et al. / Clinical Neurophysiology 131 (2020) 243–258 5. Instrumentation settings The EMG machine setting is one of the important parts for acquiring data. It should be adjusted to the same setting as those when control data were collected by the laboratory. Gain and sweep speed enable display of waveforms on the screen, and should be adjusted during practice to ensure the waveforms display perfectly in the window, avoiding overlap, and of sufficient size to be viewed and measured clearly (Table 2). The gain, and to a lesser degree sweep speed, can affect the latency measurement of action potentials, including distal motor latency, duration of compound muscle action potential or MUP, etc. As gain is increased (high sensitivity), it will become apparent that where the potential really begins (i.e., the onset latency) is less than at low gain, and the duration will increase. So, when measuring the latency, the gain and sweep speed should be adjusted to the same special setting as used when recording normal control values (Figs. 5 and 6). In digital EMG machines automatic cursor placement algorithms are available. While these are very useful and provide standardization for measurements and usually high accuracy, it is mandatory to inspect visually the accuracy of cursor placement. 6. Temporary storage, trigger and delay line All modern commercially available EMG machines have the functions of temporary storage, and trigger and delay lines. The latter allows the isolation of a single action potential from other action potentials and so enables analysis and confirmation of the consistency of the shape. By delaying each action potential after it has triggered the sweep, it appears in the same position on the screen every time there is a discharge of that unit. This requires an electronic delay circuit and the temporary storing of the recorded MUP. The sweep is triggered by the potential in real time, but the display is delayed by the preset interval (Kimura et al., 1988; Czekajewski et al., 1969; Nissen-Petersen et al., 1969). With this arrangement, the potential studied can occur repetitively and in its entirety in the same spot on the screen for precise measurement of action potential parameters in a short time. Use of a delay line is essential for the analysis of spontaneous activities, of MUPs in the concentric needle EMG and of jitter in single fibre EMG. The trigger can be placed at the different levels on the action potential to discriminate between different units. When the signal amplitude is less than the trigger level, the signal is not displayed, and when the amplitude exceeds the trigger level, the system acquires and displays one sweep. However, any potentials larger than the trigger level will trigger the sweep. Some EMG machines can also acquire the potentials with an amplitude between two trigger levels (window trigger), thus allowing the clinician to select a potential that is not the largest. During continuous recordings, a series of action potentials with the same features will be displayed Fig. 4. Panel A: analog sinus signal with frequency 1 Hz and amplitude 1 mV. Panel B. Sampling with 0.9 Hz results in aliasing, where the resultant analog signal is a sinus of 0.1 Hz. Panel C: sampling at a frequency of 100 Hz, but with a 3 bits AD converter, resulting in 5 digital amplitude values (2 3 -1), as one bit is needed for the sign (positive or negative) of the signal. This significantly distorts the representation of the signal. Panel D shows the digital signal sampled at 100 Hz with an 8-bit AD-converter, where the analog signal is now reliably represented in the digital domain. Note the altered time bases, which are not given under A and B. Table 2 Recommended gain and sweep speed settings for the commonly used electrodiagnostic recordings. Test Gain (mV/division) Sweep speed (ms/division) Motor nerve conduction studies 2000 2–5 Sensory nerve conduction studies 10 1 F wave/H reflex studies 200 5–10 Repetitive nerve stimulation 2000 2 or 200 EMG At rest 100 5–10 Minimal contraction 200–1000 5–10 Maximal contraction 1000 100 Single-fibre EMG 200–1000 0.5–1 Sympathetic skin response 500–1000 500–1000 H. Tankisi et al. / Clinical Neurophysiology 131 (2020) 243–258 249 on the screen for analysis and be temporarily stored. The amplitude trigger allows the clinician to see the EMG signal after the trigger occurs, the delay line allows the clinician to see the activity that precedes the trigger, and temporary storage allows the signal display on the screen for a short time for analysis. These functions are important for observing the repeatability of the potential during MUP analysis of needle EMG, and for recording stable single fibre action potentials for jitter analysis in single fibre EMG. Another function of the trigger is to initiate the recording of an action potential. In motor NCS or sympathic skin response, the trigger occurs when the nerve is stimulated. When the computer displays the signals immediately after triggering, the whole waveform can be seen on the screen. In sensory NCS or evoked potential studies, the trigger is important for the averaging technique; only potentials that are time-locked to the trigger will be recorded (Pease et al., 2007). 7. Signal averaging and noise reduction The widespread adoption of signal averaging in the 1970s has greatly enhanced the precision of NCS, particularly those on sensory nerves, where the signal-to-noise ratio is lower than during motor conduction studies. Prior to the advent of averaging, a common method for defining small potentials was to superimpose multiple sweeps (as in Fig. 7B and the lowest traces in Fig. 8), but this does not allow latencies to be measured precisely. In clinical practice, signal averaging is indicated whenever there is a low signal-to-noise ratio, i.e., when the background ‘‘noise” obscures the potential to be recorded or the latencies of that potential. 7.1. Sensory nerve action potentials and somatosensory evoked potentials. In these recordings, the signal may be so small that the noise inherent in the recording obscures the potential (Fig. 7A). Sensory nerve action potentials (SNAPs) may be recorded with surface electrodes (or near-nerve needle electrodes (Buchthal and Rosenfalck, 1966)) but with both techniques the SNAP is often difficult to define in single sweeps; somatosensory evoked potentials (SSEPs) are always so. Signal averaging is recommended routinely, even when the potential can be visualised readily, because it is critical that onset latency be defined accurately. For example, in orthodromic recordings the sensory potentials of the median and ulnar nerves using surface electrodes may be some tens of l Vs at the Fig. 5. The distal motor latency (DML) increased with increased gain. A. the DML was 2.8 ms when measured at 5 mV/D, B. the DML was 2.6 ms when measured at 0.5 mV/D. The measurements were done manually. Fig. 6. The sweep speed should be changed according to different conditions. A. F wave studies recorded at routine sweep speed (5 ms/D) in a patient with Morvan syndrome. B. when the sweep speed was adjusted to 20 ms/D, the after-discharges were detected. 250 H. Tankisi et al. / Clinical Neurophysiology 131 (2020) 243–258 wrist, but will only be a few l V at proximal sites, such as elbow and axilla, because the sensory volley becomes increasingly dispersed the greater the distance. Dispersion with distance is greater for SNAPs than for CMAPs because of the duration of the unitary potentials that summate to give the compound response (1–2 ms for axonal potentials; 10–15 ms for EMG potentials), such that there is greater phase cancellation for the SNAP. When recorded from Erb’s point, the sensory potentials will be much smaller than when recorded in the upper limb, and the traces are more likely to be contaminated by EMG activity. In pathology, the SNAP may be difficult to identify in a single sweep, such that the signal-tonoise ratio is poor even in distal recordings. The ‘‘reference” electrode used for SSEPs should be chosen so that activity detected by the reference does not distort the activity recorded by the ‘‘active” electrode. For upper-limb nerves, many laboratories use a cephalic reference at F pz but this injects frontal activity into the recording and should be discouraged. Less problematic cephalic reference sites are either the contralateral side or the contralateral earlobe. A remote non-cephalic reference allows far-field potentials generated by peripheral and deep midline generators to be identified, but this is at the expense of greater noise. The contralateral reference will effectively remove these farfield potentials, so that the activity from the relevant sensory cortex can be visualised. With lower limb nerves, there is little frontal activity at F pz and a reference at this site is satisfactory. 7.2. EMG potentials Not all noise is ‘‘machine noise”; it may be biological, and it may even be necessary for the desired activity to appear (as in Fig. 7B). Evoked EMG potentials commonly require averaging for accurate definition when recorded during a voluntary contraction of the target muscle, e.g., when recording the motor evoked potential during a background voluntary contraction of the target muscle (e.g., (Rossini et al., 2015)), or when recording the H reflex from a muscle from which it cannot normally be recorded at rest (such as tibialis anterior, the thenar muscles, extensor carpi radialis; see (Burke, 2016). In Fig. 7B, the background ‘‘noise” is the voluntary EMG activity against which the H wave must be identified. Superimposing multiple sweeps may then allow the definition of the target Fig. 7. Averaging to define sensory nerve action potentials and muscle action potentials. A, a small sensory nerve action potential in response to weak stimulation of the index finger (at 11.3 mA) recorded using surface electrodes over the median nerve at the wrist. Eight successive single responses, with an indefinite response peaking at ~3ms (vertical dotted line). A clearer potential with an onset latency of 2.5 ms is apparent when 9 sweeps were averaged (vertical arrow). B, raster display of EMG activity during voluntary abduction of the thumb, with an uncertain potential, no bigger than the background EMG activity, at the vertical dotted line. Lowest trace:superimposition of the rastered traces at higher amplification, showing a consistent waveform, the H reflex, at the vertical arrow. Note the absence of a direct motor response (M wave) at this stimulus intensity (5.4 mA). To define the reflex latency would require signal averaging (not illustrated). Fig. 8. Improvement in the signal-to-noise ratio. Sixteen raw evoked potential responses are cascaded on the left, and averaged two by two in the second column, then successively to the right averages of 4, 8 and all 16 traces. The raw and averaged traces are superimposed in the bottom row. The potential is not clear in the raw sweeps but becomes clearer as more sweeps are averaged. From Stegeman and Van Putten (2017), with permission. H. Tankisi et al. / Clinical Neurophysiology 131 (2020) 243–258 251 Edelberg R. Local electrical response of the skin to deformation. J Appl Physiol 1973;34:334–40. Espiritu MG, Lin CS, Burke D. Motoneuron excitability and the F wave. Muscle Nerve 2003;27:720–7. Fuglsang-Frederiksen A. The utility of interference pattern analysis. Muscle Nerve 2000;23:18–36. Fuglsang-Frederiksen A, Pugdahl K, Tankisi H. Chapter 8. Quantiative EMG. In: Mills KR, editor. Oxford textbook of clinical neurophysiology. 1st ed. Oxford University Press; 2016. p. 81–96. Fuglsang-Frederiksen A, Pugdahl K. Current status on electrodiagnostic standards and guidelines in neuromuscular disorders. Clin Neurophysiol 2011;122:440–55. Gallina A, Pollock CL, Vieira TM, Ivanova TD, Garland SJ. Between-day reliability of triceps surae responses to standing perturbations in people post-stroke and healthy controls: A high-density surface EMG investigation. Gait Posture 2016;44:103–9. Guld C, Rosenfalck A, Willison RG. Standards of instrumentation and application of EMG. In: Recommendations for the practice of clinical neurophysiology, IFSECN. Elsevier, Amsterdam, 1983. p. 124–33. Guld C, Rosenfalck A, Willison RG. Report of the committee on EMG instrumentation. Electroencephalogr Clin Neurophysiol 1974;37:532–3. Guld C, Rosenfalck A, Willison RG. Technical factors in recording electrical activity of muscle and nerve in man. Electroencephalogr Clin Neurophysiol 1970;28:399–413. Huppertz HJ, Disselhorst-Klug C, Silny J, Rau G, Heimann G. Diagnostic yield of noninvasive high spatial resolution electromyography in neuromuscular diseases. Muscle Nerve 1997;20:1360–70. IEC. International electrotechnical commission (IEC). Safety of medical electrical equipment, publication 60601-1-1; 2005. Jabre JF, Stålberg EV, Bassi R. TeleMedicine and Internet EMG. Suppl Clin Neurophysiol 2000;53:163–7. Johnsen B, Fuglsang-Frederiksen A, Vingtoft S, Fawcett P, Liguori R, Nix W, et al. Differences in the handling of the EMG examination at seven European laboratories. Electroencephalogr Clin Neurophysiol 1994;93:155–8. Johnson EW. Practical electromygraphy Subsequent ed. 1988. Williams & Wilkins Press; 1988. Kimura J, Sakimura Y, Machida M, Fuchigami Y, Ishida T, Claus D, et al. Effect of desynchronized inputs on compound sensory and muscle action potentials. Muscle Nerve 1988;11:694–702. Klijn JA, Kloprogge MJ. Movement artefact suppressor during ECG monitoring. Cardiovasc Res 1974;8:149–52. Kokubun N, Sonoo M, Imai T, Arimura Y, Kuwabara S, Komori T, et al. Reference values for voluntary and stimulated single-fibre EMG using concentric needle electrodes: a multicentre prospective study. Clin Neurophysiol 2012;123:613–20. Kouyoumdjian JA, Stålberg EV. Reference jitter values for concentric needle electrodes in voluntarily activated extensor digitorum communis and orbicularis oculi muscles. Muscle Nerve 2008;37:694–9. Kural MA, Karlsson P, Pugdahl K, Isak B, Fuglsang-Frederiksen A, Tankisi H. Diagnostic utility of distal nerve conduction studies and sural near-nerve needle recording in polyneuropathy. Clin Neurophysiol 2017;128:1590–5. Kwon H, Di Cristina JF, Rutkove SB, Sanchez B. Recording characteristics of electrical impedance-electromyography needle electrodes. Physiol Meas 2018;39. 055005–6579/aabb8c. LaBan MM, Petty D, Hauser AM, Taylor RS. Peripheral nerve conduction stimulation: its effect on cardiac pacemakers. Arch Phys Med Rehabil 1988;69:358–62. Lapatki BG, Eiglsperger U, Schindler HJ, Radeke J, Holobar A, van Dijk JP. Threedimensional amplitude characteristics of masseter motor units and representativeness of extracted motor unit samples. Clin Neurophysiol 2019;130:388–95. Lee I, Kushlaf H. Needle electromyography practice patterns in patients taking novel oral anticoagulants: A survey-based study. Muscle Nerve 2018;58:307–9. London ZN. Safety and pain in electrodiagnostic studies. Muscle Nerve 2017;55:149–59. Martinez-Valdes E, Negro F, Laine CM, Falla D, Mayer F, Farina D. Tracking motor units longitudinally across experimental sessions with high-density surface electromyography. J Physiol 2017;595:1479–96. Matsumoto L, Hanajima R, Matsumoto H, Ohminami S, Terao Y, Tsuji S, et al. Supramaximal responses can be elicited in hand muscles by magnetic stimulation of the cervical motor roots. Brain Stimul 2010;3:153–60. McGill KC, Cummins KL, Dorfman LJ, Berlizot BB, Leutkemeyer K, Nishimura DG, et al. On the nature and elimination of stimulus artifact in nerve signals evoked and recorded using surface electrodes. IEEE Trans Biomed Eng 1982;29:129–37. Nilsson J, Panizza M, Hallett M. Principles of digital sampling of a physiologic signal. Electroencephalogr Clin Neurophysiol 1993;89:349–58. Nissen-Petersen H, Guld C, Buchthal F. A delay line to record random action potentials. Electroencephalogr Clin Neurophysiol 1969;26:100–6. Ohira M, Silcox J, Haygood D, Harper-King V, Alsharabati M, Lu L, et al. Electromyography tests in patients with implanted cardiac devices are safe regardless of magnet placement. Muscle Nerve 2013;47:17–22. Pease S, Lew L, Johnson WJ. Johnson’s practical electromyography 4th ed.. Philadelphia: Lippincott Williams & Wilkins; 2007. Pierrot-Deseilligny E, Burke D. The, circuitry of the human spinal cord. Its role in motor control and movement disorders. New York: Cambridge University Press; 2005. p. 642. Robinson LR, Christie M, Nandedkar S. A message from the ground electrode. Muscle Nerve 2016;54:1010–1. Rossini PM, Burke D, Chen R, Cohen LG, Daskalakis Z, Di Iorio R, et al. Non-invasive electrical and magnetic stimulation of the brain, spinal cord, roots and peripheral nerves: Basic principles and procedures for routine clinical and research application. An updated report from an I.F.C.N. Committee. Clin Neurophysiol 2015;126:1071–107. Sanchez B, Rutkove SB. Electrical impedance myography and its applications in neuromuscular disorders. Neurotherapeutics 2017;14:107–18. Schoeck AP, Mellion ML, Gilchrist JM, Christian FV. Safety of nerve conduction studies in patients with implanted cardiac devices. Muscle Nerve 2007;35:521–4. Sgro J, Emerson R, Pedley T. Methods for steadily updating the averaged responses during neuromonitoring. In: Desmedt J, editor. Neuromonitoring in surgery. Amsterdam: Elsevier Science; 1989. p. 49–60. Shibasaki H, Hallett M. What is the Bereitschaftspotential? Clin Neurophysiol 2006;117:2341–56. Shibasaki H, Hallett M. Electrophysiological studies of myoclonus. Muscle Nerve 2005;31:157–74. Sleutjes BTHM, Drenthen J, Boskovic E, van Schelven LJ, Kovalchuk MO, Lumens PGE, et al. Excitability tests using high-density surface-EMG: A novel approach to studying single motor units. Clin Neurophysiol 2018;129:1634–41. Stålberg Erik, van Dijk Hans, Falck Björn, Kimura Jun, Neuwirth Christoph, et al. Standards for quantification of EMG and neurography. Clin Neurophysiol 2019;130(9):1688–729. https://doi.org/10.1016/j.clinph.2019.05.008. Stålberg E, Stålberg S, Melander M, Arimura K. A personal computer based system used in electromyography for interpretation and reporting. Comput Methods Programs Biomed 1991;34:219–27. Stålberg E, Stålberg S. Regional network in clinical neurophysiology, tele-EMG. In: Wootton R, editor. European telemedicine 1998/99: Kensington Publications Ltd in conjuction with the European Health telematics observatory and the royal society of medicine; 1998. p. 101–3. Stålberg S. Small bits to big bites. Muscle Nerve Suppl 2002;11:S119–27. Tam HW, Webster JG. Minimizing electrode motion artifact by skin abrasion. IEEE Trans Biomed Eng 1977;24:134–9. Türker KS. Electromyography: some methodological problems and issues. Phys Ther 1993;73:698–710. Pugdahl K, Johnsen B, Tankisi H, Camdessanché JP, de Carvalho M, Fawcett PRW, et al. Added value of electromyography in the diagnosis of myopathy: A consensus exercise. Clin Neurophysiol 2017;128:697–701. van Putten M. Essentials of neurophysiology. Basic concepts and clinical applications for scientists and engineers. Springer Verlag; 2009. Stegeman DF, Van Putten MJAM. Recording of neural signals, neural activation and signal processing. In: Mills KR, editor. Oxford textbook of clinical neurophysiology. Oxford: Oxford University Press; 2017. p. 37–45. Vingtoft S, Fuglsang-Frederiksen A, Ronager J, Petrera J, Stigsby B, Willison RG, et al. KANDID–an EMG decision support system–evaluated in a European multicenter trial. Muscle Nerve 1993;16:520–9. Webster JG. Reducing motion artifacts and interference in biopotential recording. IEEE Trans Biomed Eng 1984;31:823–6. Webster J. Amplifiers and signal processing. In: Webster JG, editor. Medical instrumentation. Application and design (Third Edition). John Wiley and Sons Inc; 1998. p. 89–120. Wood SM, Jarratt JA, Barker AT, Brown BH. Surface electromyography using electrode arrays: a study of motor neuron disease. Muscle Nerve 2001;24:223–30. 258 H. Tankisi et al. / Clinical Neurophysiology 131 (2020) 243–258