EMG Instrumentation

Needle electrode physics: geometry defines the recording volume

A recording electrode is a transducer converting ionic current in tissue to electron current in wire, and its geometry sets the volume of muscle it samples — its recording radius — and therefore which fibres contribute to the waveform:

• Concentric needle — an insulated platinum or stainless core exposed at an oblique bevel, running inside a steel cannula that serves as the reference. The active recording surface is small (~150 µm × 600 µm bevel) and the core-to-cannula geometry yields a compact, somewhat directional pickup with an effective recording radius of roughly 0.5 mm for the spike-determining near field. Because the cannula reference picks up part of the same field, the concentric montage subtracts distant activity and produces relatively short- duration, sharply defined MUAPs.
• Monopolar needle — a Teflon-coated needle bared only at the tip, referenced to a distant surface or subcutaneous electrode. With no nearby reference to subtract the far field, its uptake area is larger and less directional, so MUAPs appear higher in amplitude and longer in duration than the same units recorded concentrically — which is why normative values are strictly electrode-specific.
• Single-fibre electrode — a 25 µm side-port surface that isolates one or two fibres for jitter and fibre-density studies.

A further electrical property is impedance: the smaller the recording surface, the higher the electrode impedance, which makes the electrode noisier and more susceptible to capacitive pickup unless the amplifier's input impedance vastly exceeds it. This is why electrodiagnostic amplifiers present input impedances on the order of hundreds of megohms or more — to avoid loading and attenuating the source.

Surface versus needle EMG: matched tool to question

The choice of electrode is dictated by the spatial scale of the question. Surface electrodes integrate over a large volume and are mandatory for nerve conduction recordings, where the whole muscle (for a CMAP) or whole nerve (for a SNAP) is the object of interest; they are non-invasive, reproducible, and ideal for gross timing and amplitude. What they cannot do is resolve individual motor units, distinguish fibrillation potentials from noise, or assess MUAP morphology — the deep, high-resolution information that requires placing a small recording surface inside the muscle. Needle EMG trades the broad view for single-unit resolution and access to deep and spontaneous activity, at the cost of invasiveness and a far smaller, position- dependent sample. The two are complementary: nerve conduction studies map the conducting pathways with surface recording, and needle EMG characterizes the muscle and motor unit directly.

Differential amplification and common-mode rejection

Electrodiagnostic amplifiers are differential: they record from an active input (G1) and a reference input (G2) and amplify only the difference between them. Any voltage common to both inputs — overwhelmingly the 50/60 Hz line noise and electrostatic fields that bathe the entire patient — is subtracted and rejected. The figure of merit is the common-mode rejection ratio (CMRR), the ratio of differential gain to common-mode gain, which should exceed 100 dB(a > 100,000:1 advantage for the wanted differential signal over the unwanted common one). The physical condition for high CMRR in practice is impedance balancing: a common voltage divides between the two electrode–tissue impedances and the amplifier inputs, and only cancels cleanly if those two impedances are nearly equal. When one electrode has a much higher impedance — dried gel, poor skin prep, a corroded contact — the common signal divides unequally, no longer subtracts, and reappears as 60 Hz hum. The technologist's discipline of cleaning the skin and balancing impedances is therefore not hygiene but a direct manipulation of the rejection physics.

Filtering and signal processing: the bandpass as a measurement decision

Every channel is band-limited by a high-pass (low-frequency) filter that removes slow baseline drift and a low-pass (high-frequency) filter that removes high-frequency noise above the signal of interest. The settings are not cosmetic; they are chosen to match the frequency content of each signal, and they bias the very measurements we report:

• Motor nerve conduction (CMAP) — typically ~2–10 Hz high-pass to ~10 kHz low-pass. The wide band preserves the relatively low-frequency, broad CMAP and its onset.
• Sensory nerve conduction (SNAP) — typically ~20 Hz to ~2 kHz. The narrower band, with averaging, suppresses noise that would otherwise drown the microvolt response.
• Needle EMG — typically ~10–20 Hz high-pass to ~10 kHz low-pass, capturing the fast spike content of MUAPs and spontaneous potentials.
• Single-fibre EMG — a deliberately high ~500 Hz high-pass to isolate the brief single-fibre spike and reject the slower potentials of neighbouring fibres, sharpening the trigger for jitter measurement.

The crucial expert insight is that filters distort latency and amplitudesystematically. Raising the high-pass cutoff removes the slow components that carry much of the waveform's area, reducing amplitude and shortening apparent duration; it can also shift the onset earlier. Lowering the low-pass cutoff rounds off the fast rising edge, reducing peak amplitude and adding latency by delaying the apparent peak. Two laboratories using different filter settings will therefore measure different latencies and amplitudes on the identical physiological signal — which is exactly why normative data are only valid when the recording bandpass matches the bandpass under which they were collected.

The 60 Hz notch trade-off and signal-to-noise optimization

The 60 Hz notch filter (50 Hz in much of the world) is a narrow band-reject filter that attenuates power-line interference. It is tempting to leave it permanently on, but it carries a real cost: much of the physiologic EMG and CMAP spectrum overlaps 50/60 Hz, so the notch removes signal along with noise, distorting waveform morphology and potentially blunting a genuine response. The correct strategy is to treat the notch as a last resort and to maximize the signal-to-noise ratio at the source first:

• Eliminate the noise rather than mask it — lower electrode impedance with skin preparation, balance G1 and G2 impedances to restore CMRR, move or unplug interfering equipment, and untangle and shorten electrode leads to reduce the loop area that couples 60 Hz.
• Averaging — for small, time-locked responses such as the SNAP, averaging many sweeps suppresses random noise in proportion to √n while the time-locked signal sums coherently, dramatically improving SNR without distorting the waveform.
• Reserve the notch for irreducible line noise, and even then recognize that any latency or amplitude measured through an active notch is suspect.

Artifact suppression: recognizing the machine's fingerprints

The final competency is distinguishing artifact from physiology, because the amplifier will faithfully reproduce non-biological signals that can masquerade as disease. The common offenders have characteristic signatures: 60 Hz interference appears as a regular, mains-frequency oscillation that vanishes when CMRR is restored; stimulus artifact is a large, sharp transient time-locked to the stimulus that can obscure short-latency onsets and is minimized by rotating the anode, cleaning the skin, and proper grounding; movement and cable artifact produce slow baseline excursions removed by the high-pass filter; and electrode "pop" and impedance artifact from an unstable contact create abrupt baseline shifts. The discipline is to fix the artifact at its physical source — impedance, grounding, lead geometry, stimulator placement — rather than to filter it away, because aggressive filtering trades one error (visible noise) for a subtler and more dangerous one (a quietly distorted measurement reported as truth).