Analysis and Biophysics of Surface EMG for Physiotherapists and Kinesiologists: Toward a Common Language With Rehabilitation Engineers

Abstract

Recent decades have seen a move toward evidence-based medicine to inform the clinical decision-making process with reproducible findings from high-quality research studies. There is a need for objective, quantitative measurement tools to increase the reliability and reproducibility of studies evaluating the efficacy of healthcare interventions, particularly in the field of physical and rehabilitative medicine. Surface electromyography (sEMG) is a non-invasive measure of muscle activity that is widely used in research but is under-utilized as a clinical tool in rehabilitative medicine. Other types of electrophysiological signals (e.g., electrocardiography, electroencephalography, intramuscular EMG) are commonly recorded by healthcare practitioners, however, sEMG has yet to successfully transition to clinical practice. Surface EMG has clear clinical potential as an indicator of muscle activation, however reliable extraction of information requires knowledge of the appropriate methods for recording and analyzing sEMG and an understanding of the underlying biophysics. These concepts are generally not covered in sufficient depth in the standard curriculum for physiotherapists and kinesiologists to encourage a confident use of sEMG in clinical practice. In addition, the common perception of sEMG as a specialized topic means that the clinical potential of sEMG and the pathways to application in practice are often not apparent. The aim of this paper is to address barriers to the translation of sEMG by emphasizing its benefits as an objective clinical tool and by overcoming its perceived complexity. The many useful clinical applications of sEMG are highlighted and examples provided to illustrate how it can be implemented in practice. The paper outlines how fundamental biophysics and EMG signal processing concepts could be presented to a non-technical audience. An accompanying tutorial with sample data and code is provided which could be used as a tool for teaching or self-guided learning. The importance of observing sEMG in routine use in clinic is identified as an essential part of the effective communication of sEMG recording and signal analysis methods. Highlighting the advantages of sEMG as a clinical tool and reducing its perceived complexity could bridge the gap between theoretical knowledge and practical application and provide the impetus for the widespread use of sEMG in clinic.

Keywords: clinical application; kinesiology; physiotherapy; rehabilitation; surface EMG tutorial; surface electromography.

Copyright © 2020 McManus, De Vito and Lowery.

Figures

Figure 1

Example of a surface EMG signal at a low force level (10% of maximum voluntary contraction, MVC) (A) and a higher force level, 40% MVC (C), in the first dorsal interosseous muscle. (B) Single motor unit action potential trains with different inter-spike intervals (ISIs), i.e., different motor unit firing rates and (D) an illustration of the increase in action potential duration that can occur with a decline in muscle fiber conduction velocity. (E) A schematic to illustrate a motor unit, and how a surface EMG signal could be recorded from a muscle using a bipolar electrode (two electrode contacts).

Figure 2

(A) Surface EMG can be used to provide real-time feedback of muscle activation during neuromuscular assessments. It can show the relative timing of activation from selected muscles in different tasks. (B) Visual biofeedback from surface EMG can also aid in muscle training to ensure that rehabilitation tasks are optimally performed (and the correct muscles are “relaxed” or “activated” as required by the task).

Figure 3

Signals describing the variation of a measurable quantity over time, e.g., (A) temperature and (B) voltage. (C) A sine wave with an amplitude of 1, frequency of 10 Hz and phase of zero, showing the variation in the signal over time. The instantaneous value of the sine wave, y(t), shown in (C) can be found at each point in time, t, using the equation in (D). See Example (i) in Tutorial Code.

Figure 4

(A) A sine wave with an amplitude of 1, frequency of 10 Hz and phase of zero, showing the variation in the signal over time. A second sine wave with the same amplitude and frequency but different phase (–π/2 = −90°) is indicated with the green dashed lines. As the value of a sinusoidal signal at any point in time is based on circular motion, the phase of a signal is expressed as an angle in radians or degrees (start of period = 0°, end of period = 360° or 2π radians). (B) A sine wave with an amplitude of 0.5, frequency of 60 Hz and phase of zero. See Example (ii) in Tutorial Code.

Figure 5

(A) A 10 Hz sine wave with an amplitude of 1, shown in the time domain, y(t), and in the frequency domain, Y(f), after applying a Fourier Transform. All the power within the signal is contained at a single frequency (i.e., fundamental frequency or first harmonic−10 Hz). (B) A triangle wave with a repetition rate of 10 Hz shown in the time and frequency domain, y(t) and Y(f), respectively. As a non-sinusoidal wave, it contains frequency components at multiples of the first harmonic (note: triangle waves contain only odd harmonics). See Example (iii) in Tutorial Code. (C) The firing of a single motor unit over 2 s, shown in the time and frequency domain. The motor unit fires at an average frequency of 12 Hz (fundamental frequency of the spike train), but spectral peaks at multiples of 12 Hz can be observed in the frequency domain. (D) A 0.25 s EMG signal in the time and frequency domain. The length of the signal determines the frequency resolution (1/Tr = 4 Hz) and the lowest frequency that can be detected in the frequency domain (4 Hz). (E) A 0.1 s EMG signal is too short to observe frequencies lower than 10 Hz and can only detect frequency components that are multiples of 10 Hz.

Figure 6

(A) Schematic of the cross-section of the forearm, with the approximate locations of different muscles: flexor carpi radialis (FCR), palmaris longus (PL), flexor digitorum superficialis (FDS), flexor digitorum profundus (FDS), and flexor carpi ulnaris (FCU) (62). Note that electrode contacts should be placed approximately parallel to the muscle fiber direction. Electrode sensor 1 is placed over PL, but it will detect muscle activity (or crosstalk) from the adjacent/deep muscles, FCR and FDS. Electrode sensor 2 has a smaller inter-electrode distance (IED) than Electrode sensor 1 and will thus have a smaller pick-up volume. (B,C) Two EMG signals recorded using different electrodes with different IEDs, shown in both the time and frequency domains. EMG signals recorded using smaller IEDs can capture more high frequency components when compared with larger IEDs. Note that this diagram is for illustrative purposes and that the power spectra of the EMG signals cannot be directly compared between (B) and (C), as they were recorded in different muscles, under different conditions, using different electrodes. See Example (xi) in Tutorial Code.

Figure 7

(A) A schematic of an ideal differential amplifier with two inputs V− and V+, an output Vout and power supply connections (+10 and −10 V). (B) A schematic of the internal input impedance, Zi. present at both amplifier inputs V− and V+. Zi consists of a resistive component, Ri, and a capacitive component, Ci. (C) A schematic illustration the function of a differential amplifier, which receives two input signals at V− and V+, calculates the difference between these signals, Vd (Vd = V+ - V−), and multiplies (amplifies) Vd by the gain of the amplifier, Ad (in this example the signal is increased by a factor of 2). See Example (xii) in Tutorial Code. (D) If the gain of the amplifier is increased to a level where the expected Vout exceeds the level of the power supply voltage (e.g., ± 10 V in the amplifier shown in A), the actual Vout will be “clipped” or “limited” at the power supply voltage. (E) An example of an offset voltage being present in the amplifier output, Vout.

Figure 8

(A) A 15 Hz sine wave sampled at a rate of 40 samples/s will produce a corresponding peak at 15 Hz in the frequency domain (40 Hz is above the Nyquist frequency of 30 Hz). (B) The same sine wave sampled at 25 Hz (below the Nyquist frequency for the 15 Hz sine wave) will have a distorted amplitude spectrum, and the fundamental frequency of the signal is mis-identified as 10 Hz. See Example (iv) in Tutorial Code.

Figure 9

(A) An EMG signal sampled at 2,000 samples/s in the time domain and (B) the power spectrum of the signal in the frequency domain. The signal spectrum contains several spurious peaks. (C) Welch's method breaks the total signal (5 s long, shown in D) into shorter segments (0.5 s) and multiplies (convolves) each segment by a window function (some examples are the Hann, Hamming, and Nuttall windows) before averaging all the modified segments. See Example (viii) in Tutorial Code. (D) In Welch's averaging method, the EMG signal is divided into a number of segments (K). K depends on the length of the segment (L) and the degree of overlap between successive segments, Equation 3 in Supplementary Material. Each successive segment starts D samples after the previous segments. (E) By obtaining an average power spectral density across K segments, the spurious peaks in (B) are reduced. (F) The smoothness of the power spectral density function can be increased by increasing K (i.e., increasing the number of averages, NAVG), which can be achieved by decreasing the length of L or increasing the overlap between segments. See Example (vii) in Tutorial Code.

Figure 10

(A) Different types of filters that can be used to shape the EMG spectrum, to keep, remove, or attenuate certain frequency components of the EMG signal + noise. In the example shown, the low-pass filter has a cut-off frequency (fc) of 170 Hz and the high-pass filter has a cut-off frequency of 140 Hz. The band-pass and notch filters have lower cut-off frequencies (fc1) of 70 Hz and upper cut-off frequencies (fc2) of 108 Hz. (B) Schematic of the typical stages in recording surface EMG signals, with filtering at several points along the process (filters that operate on the analog signal, i.e., hardware filters). With active electrodes, pre-amplification is performed within the electrode itself (rather than the amplification being performed in an external circuit), see section Choice of Amplifier. EMG signals are low-pass filtered before sampling to suppress high-frequency components and prevent the distortion of the spectral content, see Figure 8. See Examples (ix) and (x) in Tutorial Code. (C) An example of a noisy sEMG signal contaminated with 50 Hz interference, the frequency response of a 50 Hz notch filter and a 20–500 Hz band-pass filter (the figure indicates how much the sEMG signal is attenuated by the filter, in dB, at each frequency), and filtered sEMG spectrum (after notch and band-pass filtering the raw sEMG signal). Note that notch filters are typically used when only an approximate estimate of EMG amplitude is required.

Figure 11

(A) A raw surface EMG signal recorded from the soleus muscle during walking, with vertical lines to indicate where heel strike and toe off occur during the gait cycle. (B) The absolute value or rectified surface EMG signal. (C) The outline or “shape” of the rectified EMG signal obtained by low-pass filtering the EMG signal at 50 Hz (after applying the Teager-Kaiser Energy Operator to the signal) with muscle onset times shown in red. (D) A 0.2-s moving average of the rectified surface EMG signal (with 25% overlap) and an average obtained using a 5-Hz low-pass filter (with the filter applied twice so that there is no time delay in the filter output, see Example (xiv) in Tutorial Code).

Figure 12

A 5-s segment of surface EMG signal from at (A) the start and (B) the end of a fatiguing isometric contraction in the first dorsal interosseous muscle. With fatigue the duration of the motor unit action potential lengthens, and there is a shift in the surface EMG signal to lower frequencies. (C) The shape of the power spectral density of the surface EMG segments shown in (A) and (B). A decrease in the mean frequency (vertical line) of the surface EMG signal is observed, accompanied by an increase in the surface EMG amplitude, see increase in RMS and ARV of the surface EMG amplitude in (B). See Example (xvii) in Tutorial Code.