From severe-to-profound hearing loss to bionic hearing — candidacy criteria, electrode insertion, tonotopic mapping, and the first year of speech perception recovery
Cochlear implant candidacy is decided by combining pure-tone audiometry, aided speech-perception testing, and cochlear/auditory-nerve imaging. What began in the 1980s as a narrow rescue therapy for total deafness has, over four decades of outcome data, expanded into a well-defined pathway for anyone whose hearing aids no longer deliver usable speech understanding.
Candidacy begins with a standard audiogram across 250 Hz–8 kHz. Severe-to-profound bilateral sensorineural hearing loss is generally defined as pure-tone average (PTA) thresholds worse than 70–90 dB HL at the frequencies most important for speech (500 Hz, 1 kHz, 2 kHz, 4 kHz). Loss of this magnitude means even well-fitted, high-power hearing aids can amplify sound into the audible range without restoring usable clarity, because the cochlear hair cells needed to transduce that amplified sound are no longer functioning.
The decisive test, however, is not the audiogram but aided speech-perception score. Patients wear their best hearing aids, properly fitted and verified, and are scored on standardized open-set tests: the CNC (Consonant-Nucleus-Consonant) monosyllabic word test and the AzBio sentence test presented in quiet and in noise. Current U.S. Medicare and FDA-aligned criteria consider adults candidates when best-aided CNC word scores fall at or below roughly 40–60% correct, or AzBio sentence scores fall at or below about 60% in the ear to be implanted (≤80% in some expanded-criteria protocols for the contralateral ear in single-sided deafness).
This is a purely functional, not purely audiometric, definition: two patients with identical audiograms can score very differently on CNC/AzBio testing depending on auditory nerve integrity, central processing, and hearing aid experience — which is why speech-in-noise testing anchors the final decision.
The FDA approved the first multichannel cochlear implant, the Cochlear Nucleus 22, in 1985 for adults with bilateral profound deafness only. Candidacy has been expanded roughly a dozen times since — to children as young as 12 months (2000), to single-sided deafness and asymmetric hearing loss (2019), and to one-ear implantation in infants as young as 9 months (2020) — each expansion driven by outcome data showing implants outperforming hearing aids at progressively milder loss levels.
Pediatric criteria differ in two important ways. First, because prelingually deaf infants cannot complete open-set speech tests, candidacy relies more heavily on objective measures: auditory brainstem response (ABR), auditory steady-state response (ASSR), and otoacoustic emissions, combined with a documented trial of appropriately fit hearing aids and, critically, lack of progress in auditory-verbal habilitation over a defined observation period (often as short as 3 months in infants, given the urgency of the sensitive period for spoken-language development).
Second, pediatric thresholds have been progressively lowered: current FDA labeling allows implantation from 9–12 months of age for profound bilateral loss, and many centers implant earlier off-label when imaging and ABR clearly support it. This aggressive timeline reflects auditory cortex neuroplasticity: the central auditory pathways remain maximally plastic during the first 2–3 years of life, and delaying implantation past this window measurably reduces achievable spoken-language outcomes even when the cochlea and electrode performance are identical.
Before surgery, every candidate undergoes high-resolution temporal bone CT and/or MRI. CT (typically 0.5–0.6 mm slices) evaluates cochlear patency and anatomy: it identifies cochlear ossification or fibrosis (common after meningitis, which can obliterate the scala tympani within weeks), cochlear malformations (Mondini deformity, common cavity, cochlear aplasia), and the course of the facial nerve relative to the planned facial-recess approach.
MRI, usually a heavily T2-weighted 3D sequence (e.g., CISS/FIESTA), is used to confirm the fluid-filled membranous labyrinth is intact and — most importantly — to confirm the presence and caliber of the cochlear nerve within the internal auditory canal. Cochlear nerve deficiency/aplasia, seen in a subset of children with congenital profound deafness, is a relative or absolute contraindication for a standard cochlear implant and may redirect the workup toward an auditory brainstem implant (ABI) instead. In post-meningitic ossified cochleae, imaging findings determine whether a standard array, a shorter/split array, or a drill-out technique will be needed at surgery.
Cochlear implantation is a same-day, 1.5–3 hour otologic procedure performed under general anesthesia. The surgeon opens a corridor through the mastoid bone to the middle ear, then threads a flexible multi-contact electrode array into the fluid-filled scala tympani of the cochlea — all while working within millimeters of the facial nerve and preserving whatever residual low-frequency hearing the patient still has.
The standard approach begins with a postauricular incision and a cortical mastoidectomy, drilling through the mastoid air-cell system to expose the antrum and the horizontal semicircular canal — key landmarks for orientation. The surgeon then opens the facial recess: a narrow triangular corridor bounded by the facial nerve (posteriorly/medially), the chorda tympani nerve (laterally), and the incus buttress (superiorly). This "posterior tympanotomy" gives direct visual and instrument access to the middle ear and the round window niche without disturbing the ossicular chain, and it is precisely why facial nerve monitoring (continuous EMG) is used throughout the case.
Once the round window niche is visualized, the receiver-stimulator package is seated in a subperiosteal pocket drilled into the skull just behind and above the ear, and a channel is drilled to carry the electrode lead from that pocket to the facial recess.
There are two accepted entry points into the scala tympani. A round window insertion passes the array directly through the round window membrane, which best preserves the natural anatomy and is now the preferred technique at most centers. A cochleostomy drills a small (~1 mm) opening just anterior-inferior to the round window when the membrane is obscured or narrow, entering the scala tympani directly.
Modern "soft-surgery" (or "hearing preservation") technique aims to keep the cochlea's delicate hair-cell and fluid environment as undisturbed as possible: a small, round-window entry; slow, steady insertion (often 1 mm every few seconds) using an insertion tool rather than forceps; topical steroid (dexamethasone) application before and after insertion to blunt the inflammatory response; and avoiding suction or fluid loss from the cochlea (perilymph) during the case. These steps have raised low-frequency acoustic hearing preservation rates enough that many devices now support electro-acoustic stimulation (EAS) — combining a standard electrode array for high frequencies with amplified acoustic sound for preserved low-frequency residual hearing in the same ear.
Insertion depth is deliberately conservative. A full 35 mm cochlear duct is never targeted — the electrode is advanced only 20–31 mm depending on the array (e.g., Cochlear Contour Advance ~17–20 mm active length, MED-EL FLEX28/FLEXSOFT up to ~28–31.5 mm). Overinsertion risks tip fold-over or fracture through the thin apical scala tympani wall into the scala vestibuli, which degrades tonotopic fidelity and increases the risk of losing residual hearing.
Cochlear implantation is one of the safest procedures in otology, but it carries defined risks: facial nerve injury (<1% permanent; transient weakness slightly more common, usually from monopolar cautery current spread rather than direct trauma — mitigated by continuous intraoperative facial nerve monitoring), cerebrospinal fluid (CSF) gusher or oozing in ears with an underlying inner-ear malformation (managed with fascia/muscle packing around the electrode at the cochleostomy), postoperative infection or flap complications (~1–3%), device extrusion, and taste disturbance from chorda tympani manipulation.
Intraoperative telemetry — impedance testing and neural response telemetry (NRT/ART/AutoNRT depending on manufacturer) — is performed before closure to confirm every electrode contact is making a proper circuit and eliciting a measurable auditory nerve response, catching an electrode fold-over or open/short circuit before the patient ever leaves the operating room.
The healthy cochlea is a mechanical frequency analyzer: the stiff basilar membrane near the base resonates to high frequencies, while the floppier membrane near the apex resonates to low frequencies. A cochlear implant recreates this tonotopic organization electronically — each electrode contact along the array is assigned to a specific frequency band based on its physical position, following the same Greenwood function that describes the natural cochlea.
In 1990, Donald Greenwood formalized the relationship between position along the basilar membrane and characteristic frequency as f = A·(10^(a·x) − k), where x is fractional distance from the apex (0) to the base (1), and A, a, and k are species-specific constants (roughly A=165.4, a=2.1, k=0.88 for humans). This single equation predicts that a 35 mm human cochlea spans roughly 20 Hz at the extreme apex to over 20,000 Hz at the extreme base — a logarithmic map, not a linear one, which is why equal spacing of electrode contacts does not produce equal spacing of assigned frequencies.
When the electrode array is inserted, each contact lands at a measurable angular insertion depth (degrees of cochlear rotation, 0–630°+ for a full array) that converts, via the Greenwood function, to a "physical" characteristic frequency for that location. The processor's frequency allocation table then assigns each electrode a frequency band to analyze and stimulate with.
In practice, no commercial array reaches the true apex, so default frequency allocation tables compress the full audible range (roughly 188 Hz–8 kHz for Cochlear devices, similarly banded for Advanced Bionics and MED-EL) across whatever electrodes are actually inserted — typically 16–22 active contacts spanning 20–31 mm. This produces an intentional "frequency-to-place mismatch": a basal electrode sitting at a cochlear location whose natural characteristic frequency might be 6,000 Hz is instead assigned to process, say, 1,700–2,500 Hz of incoming sound, because that is where useful speech information lives.
Remarkably, the brain adapts to this mismatch over weeks to months of listening experience — a form of auditory perceptual plasticity — which is one reason speech perception scores continue to climb for months after activation rather than reaching ceiling immediately. Deeper insertion (more apical electrodes engaged) reduces the degree of basal mismatch and is associated, on average, with better low-frequency pitch perception and music appreciation, one motivation behind longer "hearing preservation" arrays.
A shallow 20 mm insertion typically activates around 14–16 of 22 electrode contacts and leaves the most apical, lowest-frequency channels unstimulated or compressed into the basal map. A deep 28–31 mm insertion can engage nearly the full array, extending usable place-coded frequency information down toward 250–300 Hz — closer to the natural apex — which is why insertion depth is one of the most closely audited outcomes on postoperative imaging.
The external sound processor is where acoustics becomes electricity. A microphone captures sound, a bank of bandpass filters splits it into parallel channels matched to the electrode array's tonotopic map, and each channel's amplitude envelope is converted into a train of charge-balanced current pulses delivered to its corresponding electrode — dozens to thousands of times per second.
All modern strategies share the same basic pipeline. The microphone signal is digitized, passed through an automatic gain control stage, and split by a bank of bandpass filters (typically 12–22 channels, one per active electrode). Within each channel, the envelope — the slow amplitude fluctuation of that frequency band over time — is extracted, compressed into the patient's electrical dynamic range (between their threshold, T-level, and comfort, C/M-level), and used to modulate the amplitude of a fixed-rate pulse train delivered to the matching electrode.
Because simultaneous stimulation of neighboring electrodes causes current to summate and blur place coding ("channel interaction"), essentially every clinically used strategy interleaves pulses in time — no two adjacent electrodes fire at exactly the same instant — which is the "interleaved" half of Continuous Interleaved Sampling (CIS), the foundational strategy from which most modern variants descend.
Continuous Interleaved Sampling (CIS): the original high-rate strategy (Wilson et al., 1991) — every active channel is stimulated on every cycle, non-simultaneously, at rates around 800–1,800 pulses per second per channel. It remains available on nearly every modern device as a robust fallback strategy.
Advanced Combination Encoder (ACE), Cochlear: an "n-of-m" strategy that, on each stimulation cycle, selects only the n channels (commonly 8) with the highest instantaneous amplitude out of all m available channels (commonly 22) and stimulates only those. By skipping low-energy channels, ACE can push effective per-channel stimulation rates higher within the same overall pulse budget, and it dominates commercial Cochlear Nucleus fittings today.
HiRes and HiRes with Fidelity 120, Advanced Bionics: HiRes runs at very high stimulation rates (up to roughly 5,156 pulses per second total) for finer temporal envelope resolution. Fidelity 120 adds "current steering" — simultaneously weighting current between two adjacent physical electrodes to create perceptually intermediate, virtual pitch percepts, effectively multiplying the 16 physical contacts into up to ~120 perceptible spectral bands.
Fine Structure Processing (FSP), MED-EL: rather than transmitting only the slow envelope of each channel, FSP additionally tracks and stimulates in sync with the zero-crossings of the temporal fine structure in the lowest 1–2 apical channels (below roughly 1 kHz) — the fast oscillations that carry musical pitch and, importantly, the pitch contours used in tonal languages like Mandarin. This is MED-EL's central strategy differentiator and one reason FSP-based mapping is emphasized in tonal-language-speaking populations.
A cochlear implant is useless without a program (a "MAP") that tells it how much current each electrode needs to be audible but not uncomfortable. Roughly 2–4 weeks after surgery, once postoperative swelling has resolved, the audiologist performs the first activation session — the moment most patients describe as switching from silence, or unintelligible buzzing, to their first sounds through the device.
For every active electrode, the audiologist measures two current levels that define the patient's electrical dynamic range. The threshold (T) level is the minimum current that produces a just-audible sensation on that electrode. The comfort level — called the C-level (Cochlear/MED-EL nomenclature) or M-level ("most comfortable," Advanced Bionics) — is the maximum current that is loud but still comfortable, never painful. Sound is then mapped so the incoming signal's dynamic range (roughly 30 dB in normal speech) is compressed to fit between each electrode's T and C/M levels, and the patient hears louder or softer sound as the delivered current is scaled between these two per-electrode bounds.
Early sessions, especially in infants and young children who cannot give verbal feedback, rely heavily on objective and behavioral proxies: electrically evoked stapedial reflex thresholds (ESRT), electrically evoked compound action potentials (ECAP/NRT/ART), and observation of behavioral responses (startle, orienting, play audiometry) to estimate T and C/M levels safely.
The first MAP is deliberately conservative — comfort levels are set slightly low to avoid uncomfortably loud or startling sound on day one. Over the following visits (commonly at 1, 2, 4, 8, 12, 24, and 52 weeks, more often in children), the audiologist raises C/M levels as the patient acclimates, adjusts channel-by-channel balance if certain electrodes sound too loud, too soft, or produce non-auditory sensations (facial nerve stimulation from current spread, most common near the round window electrodes), and fine-tunes overall loudness growth and pitch balance across the array.
Behavioral speech-perception testing (CNC, AzBio) is typically repeated at each major milestone (1, 3, 6, and 12 months) both to track progress and to guide further MAP adjustment — a plateau or regression in scores can signal an electrode problem, a MAP that is now too conservative, or the need for auditory rehabilitation therapy rather than a device change.
Because the peripheral auditory nerve and central auditory pathways continue to adapt to a fundamentally new, artificial pattern of stimulation, a "final" MAP does not really exist in the first year — most centers still make audible adjustments to T and C/M levels at every visit through 12 months, even in adults who are already understanding open-set speech well.
Speech perception with a cochlear implant is not instantaneous — it is learned. Most of the measurable gain happens in the first 3–6 months as the brain relearns to decode a fundamentally new, electrically coded version of sound, with scores generally plateauing by around 12 months. The single biggest predictor of long-term outcome in prelingually deaf children is age at implantation, reflecting a closing window of central auditory neuroplasticity.
A typical postlingually deafened adult enters surgery scoring near 0–10% on best-aided CNC word testing — hearing aids are amplifying sound the cochlea can no longer usefully transduce. Within the first month after activation, most patients already exceed their preoperative aided score. Gains continue steeply through months 2–6 as the auditory system adapts to electrical place coding and to whichever coding strategy (ACE, HiRes, FSP, CIS) their processor uses, then growth slows and scores typically plateau by 9–12 months, with many adults ultimately reaching 60–80% or higher on CNC words and even higher on AzBio sentences in quiet.
Outcomes vary with duration of deafness before implantation (shorter is better), residual low-frequency hearing (better with hearing-preservation surgery and EAS), auditory nerve health, and consistent device use — which is why postoperative counseling emphasizes daily wear time and structured listening practice, not just the surgery itself.
Children who are congenitally or prelingually deafened and implanted before roughly age 2–3 show dramatically better long-term spoken-language outcomes than children implanted later — many approach age-appropriate speech and language scores by school age, some statistically indistinguishable from normal-hearing peers on standardized language testing. This reflects a well-documented sensitive period in central auditory cortex development: the auditory pathways are maximally plastic in infancy, and each additional year of auditory deprivation before implantation measurably erodes the ceiling on achievable outcomes, even though the electrode array and coding strategy are identical to those used in adults.
This is the central argument behind lowering the FDA-approved implantation age over time (12 months in 2000, 9 months for one ear in 2020) and behind universal newborn hearing screening protocols that aim to identify profound congenital deafness within the first months of life, so implantation can occur well inside the critical window rather than after it has begun to close.
Graeme Clark implanted the first multichannel cochlear implant in Rod Saunders in Melbourne, Australia, in 1978 — the direct ancestor of today's Cochlear Nucleus device family. Fewer than 50 years later, more than 1 million people worldwide have received cochlear implants, and children implanted before their first birthday under current expanded criteria now routinely develop spoken language on a timeline close to their normal-hearing peers — an outcome unimaginable under the profound-only, adults-only criteria of 1985.