Electromyography

Electromyography (EMG) is the diagnostic and experimental electrophysiological technique used to sense, amplify, and decompose the extracellular bioelectric potentials generated by skeletal muscle fibers during voluntary contraction, reflex activation, or electrical nerve stimulation. The fundamental functional element recorded in an electromyogram is the motor unit (first defined by Charles Scott Sherrington in 1925), which comprises a single alpha motor neuron (α-MN) whose cell body resides in the ventral horn of the spinal cord (or brainstem motor nuclei), its myelinated motor axon, and the ensemble of 5 to 2,000 skeletal muscle fibers it innervates according to its innervation ratio. When a central motor command depolarizes the α-MN past threshold, a neural action potential propagates down the motor axon to the neuromuscular junction (NMJ)—the motor endplate—where voltage-gated Ca²⁺ channels trigger the exocytosis of acetylcholine (ACh) vesicles across the synaptic cleft.

Binding of acetylcholine to nicotinic acetylcholine receptors (nAChRs) on the postjunctional muscle membrane opens cation channels that raise the endplate potential from the resting equilibrium voltage (governed by the Nernst equation and Goldman–Hodgkin–Katz equation at V_rest ≈ −85 to −90 mV) past −55 mV. This initiates a bidirectional muscle fiber action potential (MFAP) that propagates longitudinally along the sarcolemma at conduction velocities of v_c = 3.0–5.5 m/s and plunges radially into the transverse-tubule (T-tubule) system to trigger dihydropyridine-ryanodine receptor (DHPR–RyR1) calcium release and cross-bridge sarcomere contraction. The spatiotemporal summation of MFAPs across all fibers belonging to a single motor unit produces the extracellular Motor Unit Action Potential (MUAP), while the superposition of asynchronous MUAP trains from dozens of active motor units forms the raw interference EMG signal (ranging from 50 μV to 30 mV in amplitude across a 10–500 Hz spectral bandwidth).

Extracellular Potential and Second Spatial Derivative Modeling

Within the framework of electrophysiological biophysics, an excitable muscle fiber embedded in interstitial fluid, connective tissue, adipose tissue, and skin acts as a linear core-conductor current source inside an anisotropic three-dimensional volume conductor. Along a cylindrical muscle fiber of radius a and intracellular axial conductivity σ_i oriented along the z-axis, core-conductor cable theory establishes that the transmembrane current per unit length i_m(z, t) exiting the sarcolemma into the extracellular medium is proportional to the second spatial derivative of the intracellular transmembrane action potential V_m(z, t):

i_m(z, t) = π a² σ_i (∂²V_m(z, t) / ∂z²)

Because the rising depolarization front and falling repolarization tail of a propagating MFAP produce a positive-negative-positive sequence of curvature in ∂²V_m/∂z², each traveling action potential is modeled electrostatically as a current tripole (source–sink–source). By quasi-static Maxwell volume conduction in a medium of radial conductivity σ_r and axial conductivity σ_z, the extracellular potential φ(r, z, t) measured by a recording electrode at radial distance r from the fiber is given by the convolution integral:

φ(r, z, t) = ( a² σ_i / (4 σ_r) ) ∫_−∞^+∞ [ (∂²V_m(z′, t) / ∂z′²) / √( (σ_z / σ_r)r² + (z − z′)² ) ] dz′

Consequently, a propagating MFAP sensed by a monopolar or differential electrode located away from the endplate and tendon junctions produces a characteristic triphasic waveform. Furthermore, the biological volume conductor behaves as a spatial low-pass filter: high spatial and temporal frequencies attenuate rapidly with electrode-to-fiber radial distance r, explaining why intramuscular needle electrodes placed within <0.5 mm of active fibers record sharp, millisecond-duration spikes up to 5–10 kHz, whereas surface electrodes separated by 2–15 mm of subcutaneous tissue record broadened, smoothed potentials band-limited below 400–500 Hz.

Henneman's Size Principle and Rate Coding

The central nervous system grades isometric and dynamic muscle force through two synergistic mechanisms monitored by EMG: spatial recruitment of additional motor units and rate coding (modulating the firing frequency of already active motor units from ~5–8 Hz at recruitment threshold up to 35–60 Hz during maximal voluntary contraction). As demonstrated by Elwood Henneman in 1957 (Henneman's Size Principle), motor units within a spinal motor pool are recruited in a fixed order of increasing soma surface area. Because a smaller α-MN soma possesses fewer parallel leak ion channels, it exhibits a higher input resistance R_in; by Ohm's Law (ΔV = I_syn R_in), a given common synaptic current I_syn produces a larger depolarization in small α-MNs, bringing Type S (slow-twitch, oxidative, fatigue-resistant) motor units to threshold first, followed progressively by Type FR (fast-twitch, fatigue-resistant) and high-force Type FF (fast-twitch, glycolytic, fatigable) motor units, which produce the largest-amplitude MUAPs.

Differential Amplification and Common-Mode Rejection

Because raw surface EMG potentials are microvolt-scale signals superimposed on 50/60 Hz power-line interference (which can induce 10–100 mV of common-mode voltage across the human body via capacitive coupling) and DC half-cell electrode polarization offsets (~220 mV for Ag/AgCl electrodes), EMG front-ends utilize bipolar differential instrumentation amplifiers. Recording the voltage difference V_out = G_diff (V₁ − V₂) + G_cm ((V₁ + V₂)/2) between two closely spaced electrodes positioned parallel to the muscle fiber orientation cancels both common-mode power-line hum and far-field crosstalk when the amplifier provides a Common-Mode Rejection Ratio (CMRR) of:

CMRR_dB = 20 log₁₀( G_diff / G_cm ) > 100–110 dB

alongside an ultra-high input impedance (Z_in > 1–10 GΩ) to prevent voltage-divider attenuation from uneven skin-electrode contact impedances (5–50 kΩ).

Time-Domain Envelopes, Spectral Fatigue Shifts, and Pathology

In quantitative movement science and rehabilitation analyzed by a clinical neuroscientist or biomechanist, the stochastic raw sEMG interference pattern x(t) is quantified in both the time and frequency domains:

• Root-Mean-Square (RMS) Amplitude: Computed over a sliding window of duration T (typically 50–250 ms) as EMG_RMS = √[ (1/T) ∫_t^t+T x²(τ) dτ ], which reflects the square root of signal power and scales monotonically with neural drive and isometric muscle torque.
• Spectral Compression During Muscle Fatigue: During sustained submaximal or maximal contractions, extracellular potassium (K⁺) accumulation in the T-tubules and intracellular lactic acid / H⁺ buildup slow down sarcolemmal Na⁺/K⁺ pump kinetics, reducing muscle fiber action potential conduction velocity v_c from ~4.5 m/s toward 2.8 m/s. Because the power spectral density S(f) of the surface EMG scales linearly in frequency with conduction velocity, the median frequency (f_med) (which divides the power spectrum into two equal-area halves, typically 80–120 Hz in unfatigued biceps brachii) shifts progressively downward toward lower frequencies (40–65 Hz) well before mechanical force output collapses.

In clinical intramuscular needle EMG, examining individual MUAP morphology distinguishes neuropathic from myopathic disorders:

1. Neurogenic Disorders (e.g., Amyotrophic Lateral Sclerosis, Radiculopathy, Peripheral Neuropathy): Denervated muscle fibers spontaneously generate microscopic fibrillation potentials and positive sharp waves at rest. Surviving motor axons undergo collateral sprouting to reinnervate orphaned muscle fibers, producing giant, long-duration (>15 ms), polyphasic MUAPs (>2–10 mV) with reduced recruitment numbers.
2. Myopathic Disorders (e.g., Muscular Dystrophy, Polymyositis): Random degeneration of individual muscle fibers within intact motor units reduces the number of fibers contributing to each firing event, yielding low-amplitude (<300 μV), short-duration (<5 ms), polyphasic MUAPs paired with early (rapid) recruitment of many units to generate even weak forces.

High-Density Blind Source Separation and Subvocal Decoding

Advances in High-Density Surface EMG (HD-sEMG)—employing flexible 64- to 256-channel electrode grids combined with convolutive blind source separation (BSS) algorithms (such as FastICA and Convolution Kernel Compensation)—enable non-invasive mathematical decomposition of the surface interference signal back into the exact millisecond discharge timings of 20 to 60 individual spinal alpha motor neurons in vivo. In upper-limb amputees who have undergone Targeted Muscle Reinnervation (TMR) (where severed residual peripheral nerves are surgically transferred to denervated chest or residual-limb muscle segments), decoded motor unit spike trains provide simultaneous, proportional multi-degree-of-freedom control of robotic myoelectric prosthetic hands.

Furthermore, ultra-thin epidermal sEMG electrode arrays placed over the orofacial, submental, and laryngeal musculature (recording from the orbicularis oris, zygomaticus, genioglossus, mylohyoid, and cricothyroid muscles) capture the subtle neuromuscular motor unit commands generated during silent (subvocal) speech—when a user articulates words internally without expelling acoustic breath or vibrating the vocal folds. Deep temporal convolutional and recurrent neural networks translate these multi-channel subvocal sEMG spectrograms into continuous text or synthesized speech at >90% word accuracy, establishing non-invasive peripheral neural interfaces for individuals with laryngectomy, dysarthria, or motor neuron disease.

Why does a surface EMG signal look like random noise during a muscle contraction?

A strong voluntary contraction recruits dozens to hundreds of independent motor units whose alpha motor neurons fire asynchronously at 10–40 Hz without phase-locking to one another. Because each motor unit generates a positive- and negative-going triphasic MUAP that sums linearly with overlapping potentials in the volume conductor (algebraic phase cancellation), their superposition forms a zero-mean stochastic interference pattern whose amplitude envelope (RMS) tracks overall neural drive.

What causes the EMG median frequency to drop when a muscle fatigues?

During sustained contraction, metabolic accumulation of extracellular potassium ions (K⁺) inside the narrow transverse tubules and intracellular acidosis (H⁺ and inorganic phosphate buildup) reduce the excitability of voltage-gated Na⁺ channels. This slows the conduction velocity (v_c) of action potentials propagating along the muscle fibers, stretching the MUAP waveforms in time and shifting the power spectrum's median frequency (f_med) toward lower frequencies.

What is the difference between needle EMG and surface EMG?

Intramuscular needle EMG inserts a fine concentric wire electrode directly through the skin into the muscle belly, recording within <1 mm of 8–20 individual muscle fibers across a wide 10–10,000 Hz bandwidth to diagnose denervation, ALS, or myopathy. Surface EMG (sEMG) uses non-invasive adhesive electrodes on the skin (20–450 Hz bandwidth) to measure the global activity of whole muscles or decompose motor unit pools via high-density arrays.

1. Henneman, E. (1957). "Relation Between Size of Neurons and Their Susceptibility to Discharge." Science, 126(3287), 1345–1347. https://doi.org/10.1126/science.126.3287.1345
2. De Luca, C. J. (1997). "The Use of Surface Electromyography in Biomechanics." Journal of Applied Biomechanics, 13(2), 135–163. https://doi.org/10.1123/jab.13.2.135
3. Merletti, R., & Farina, D. (Eds.). (2016). Surface Electromyography: Physiology, Engineering, and Applications. IEEE Press / John Wiley & Sons. https://doi.org/10.1002/9781119082934
4. Farina, D., Merletti, R., & Enoka, R. M. (2004). "The Extraction of Neural Strategies from the Surface EMG." Journal of Applied Physiology, 96(4), 1486–1495. https://doi.org/10.1152/japplphysiol.01070.2003