3ECAP / Neural Response Telemetry

FWhat the ECAP is

The electrically-evoked compound action potential (ECAP) is the summed, synchronous response of the auditory nerve fibres near a stimulating electrode to a single electrical pulse. It is the electrical analogue of the acoustic compound action potential recorded in electrocochleography, and the same physiological event as wave I of the auditory brainstem response — but recorded right at its source, inside the cochlea, through the implant itself.[1990, 2017]

Each manufacturer brands the feature differently — NRT (Neural Response Telemetry, Cochlear), ART (Auditory nerve Response Telemetry, MED-EL), NRI (Neural Response Imaging, Advanced Bionics), and, beyond the big three, NRM (Nurotron) and NRD (Listent) — but all record the same biological signal. Module 11 maps the terminology across the whole field.[1999]

TMorphology & generators

A clean ECAP is a biphasic wave with two named peaks:

• N1 — a negative peak, typically around 0.2–0.4 ms after the stimulus, generated by the synchronous firing of the distal auditory nerve fibres.
• P2 — the following positive peak, around 0.6–0.8 ms.

The clinical readout is the N1–P2 amplitude, measured in microvolts: roughly 25–30 µV near threshold, rising to as much as ~1500 µV at high levels — far larger than the sub-microvolt acoustic CAP, because the electrode sits directly on the generator. (A small N2 trough may sit between N1 and P2 at suprathreshold levels.) The dominant generator is believed to be the spiral ganglion cell bodies of the auditory nerve. Latency is short and relatively stable; amplitude is the parameter that carries most of the clinical information, and the next module builds its measures on amplitude.[2002, 2020]

TCThe artifact problem

Here is the central difficulty. To evoke the ECAP you inject a current pulse; that pulse creates an electrical stimulus artifact at the recording electrode that is orders of magnitude larger than the few-microvolt neural response, and it arrives at essentially the same time. Worse, the amplifier itself can saturate on the artifact and take time to recover, masking the early neural response under a decaying baseline.[2000]

Every clinical ECAP system therefore has to separate the neural response from the artifact. Two methods dominate, and they can be combined.

CForward-masking subtraction

The forward-masking paradigm exploits neural refractoriness — the brief period after a fibre fires during which it cannot fire again. The recording uses a clever subtraction of several traces:

1. Record the response to the probe alone (= neural response + artifact).
2. Deliver a maskerpulse first, then the probe at a short interval. The masker has just driven the nerve into refractoriness, so the probe now evokes little or no neural response — what remains is essentially the probe's artifact(plus the masker's own response, which is handled by recording the masker alone and a baseline frame).
3. Subtract the masked frame from the probe-alone frame. The artifact, common to both, cancels; the neural ECAP survives.

Forward-masking subtraction is the reference method and the one underlying Cochlear's AutoNRT. It is robust, but it needs several stimulus frames per measurement and assumes the masker fully saturates the response.[2000, 2007]

CAlternating polarity

The alternating-polarity method exploits a symmetry: the electrical artifact reverses sign when you reverse the stimulus polarity (anodic vs cathodic), whereas the neural response is, to a first approximation, polarity-independent. Average the responses to anodic and cathodic pulses and the artifact tends to cancel while the neural response reinforces.

It is simpler and faster than forward masking, but the approximation is imperfect — real nerves do show polarity sensitivity, and that very asymmetry is informative about neural health (a theme picked up in Module 4). In practice systems may use alternating polarity, forward masking, or a scaled-template subtraction (model the artifact from a sub-threshold trace and scale it up), and sometimes a combination.

The defaults split by manufacturer: forward-masking subtraction is the default on Cochlear (NRT) and is generally the most effective — it yields the largest amplitude, lowest threshold, and most robust morphology — while alternating polarity is the default on Advanced Bionics (NRI) and MED-EL (ART). Array geometry matters too: perimodiolar arrays, sitting closer to the modiolus for both stimulating and recording, tend to give lower ECAP thresholds than lateral-wall arrays.[2020]

TRecording in practice

A few practical points the trainee meets at the chair:

• Recording electrode choice. The ECAP is recorded on an electrode near, but not the same as, the stimulating one (commonly two contacts away) to limit direct artifact pickup.
• Run impedance first. An ECAP cannot be evoked or recorded on an open contact — clear the interface with impedance telemetry before interpreting an absent response.
• Averaging. Even after artifact rejection, several sweeps are averaged to lift the µV response out of noise; a noisy or moving patient degrades the recording.
• Typical acquisition settings. The ECAP is evoked at a slow stimulation rate (~30–80 pps, far below the coding-strategy rate) with a pulse width around 25–40 µs; the amplifier runs at 40–70 dB gain with a recording delay of ~50–125 µs (too short and the artifact saturates, too long and the response is missed); and the system noise floor is about 5–25 µV, so a waveform below ~20–25 µV is treated as within the noise. Only 32–50 sweeps are usually needed.[2020]
• Absent ≠ no nerve. A flat ECAP can mean poor neural survival, but also an open electrode, residual artifact, or a recording fault. Interpret it alongside impedance and, where the question is nerve integrity, the eABR.

Two numbers explain why the ECAP, not the eABR or a cortical response, is the everyday objective measure. Because it is recorded right on top of its generator it is large — hundreds of µV, up to ~1 mV — so it needs only about 50–100 sweeps to average out of the noise (versus a thousand or more for the far-field brainstem and cortical responses), and it is measurable in ~95% of recipients. No sedation, no still patient, seconds per electrode.[2013]