9.2 Processing strategies: CIS, ACE, FSP

The CI processor takes the microphone’s broadband acoustic signal and converts it to a per-electrode pulse train. The conversion algorithm — the processing strategy — is the most clinically variable parameter of a CI fitting. Different strategies emphasise different acoustic features: spectral resolution vs temporal continuity vs fine-structure preservation. The choice of strategy has measurable effects on speech-in-noise perception, music appreciation, and pitch coding fidelity.

Three families dominate clinical use in 2026:

CIS (Continuous Interleaved Sampling): the foundational multichannel strategy, still the default in MED-EL devices and a benchmark in research.
ACE / N-of-M: the dominant strategy in Cochlear (Nucleus) devices since 2000.
FSP / FS4 (Fine Structure Processing): MED-EL’s strategy that attempts to encode temporal fine structure in addition to the envelope.

strategy:

per-channel rate = 900 pps

An incoming speech signal is split into 22 frequency channels (matching the 22 electrodes), the envelope is extracted from each channel, and the envelope amplitude is mapped to a per-electrode stimulation level. In CIS (Continuous Interleaved Sampling, Wilson 1991), ALL channels are stimulated once per cycle, in sequence. In ACE (Advanced Combination Encoder, the dominant strategy in modern Cochlear-brand devices), only the N loudest channels per cycle are stimulated — reducing the total stimulation rate while focusing electrical current on the most informative channels. N-of-M strategies trade temporal continuity (CIS is uniform) for spectral focus (ACE skips quieter channels). Modern devices typically use 12-of-22 with 900 pps per channel = 10,800 pps total stimulation rate.

CIS — Continuous Interleaved Sampling

CIS, developed by Blake Wilson and colleagues at Research Triangle Institute and described in their seminal 1991 Nature paper, established the modern CI processing paradigm. The five steps:

Filterbank. Split the broadband input into N parallel frequency channels (N = 16 to 22, matching the number of intracochlear electrodes). Each channel covers a logarithmic frequency band corresponding to the cochlear place of its assigned electrode — apical electrodes get low-frequency channels (200–500 Hz), basal electrodes get high-frequency channels (4–8 kHz).
Envelope extraction. Rectify and low-pass filter each channel to extract the slowly-varying envelope (~50–400 Hz). The fine-structure (carrier) of each channel is discarded.
Compression mapping. Apply a logarithmic compression to each envelope: $\text{stim}(t) = \alpha \log(1 + \beta \text{env}(t))$ — mapping the wide acoustic dynamic range into the narrow electric dynamic range (typically 6–20 dB between T and C levels in the patient’s measured thresholds).
Interleaved pulse trains. Generate a biphasic pulse train per electrode at a fixed per-channel rate (typically 900–1200 pps in modern devices). The pulses across electrodes are time-interleaved — no two electrodes fire simultaneously. This prevents inter-electrode current summation that would scramble the place coding.
Per-electrode mapping. Each pulse’s amplitude is set by the current envelope value (after compression) of its assigned channel.

CIS is all-channels-all-the-time: every electrode fires every cycle. The total stimulation rate is rate × N = 900 × 22 = 19,800 pps with 22 electrodes at 900 pps each. This dense activity preserves the temporal envelope information across the full spectrum but provides no special prioritisation of important channels.

ACE / N-of-M

The ACE (Advanced Combination Encoder) strategy, developed at the University of Melbourne in the 1990s and commercialised in Cochlear’s Nucleus devices, replaces CIS’s all-channels rule with an N-of-M rule: in each cycle, only the N loudest channels (out of M = 22 total) are stimulated. Typical N = 8 or 12. The strategy:

Filterbank and envelope extraction, identical to CIS.
Per-cycle channel selection. At each stimulation cycle, sort the channel envelopes by amplitude and select the top N. Channels below the top-N threshold are not stimulated this cycle.
Compression, mapping, and pulse generation identical to CIS, but only for the selected N channels.

The rationale: in any short window, most acoustic information is concentrated in a few frequency bands (the spectral peaks). Stimulating only the loudest channels focuses the available electrical “bandwidth” on the most informative channels while leaving the quieter channels silent. The total stimulation rate is rate × N = 900 × 12 = 10,800 pps — about half the rate of full CIS.

ACE shows small but consistent advantages over CIS in:

Power consumption. Half the stimulation rate, half the battery drain. Modern CI users routinely get 16+ hours per battery on ACE; CIS-only operation halves that.
Frequency selectivity in moderate noise. The peak-picking selectively passes spectral peaks (formant frequencies, the most informative parts of speech) while suppressing low-energy channels (often dominated by noise).
Music perception in some patients: peak-picking preserves the pitch-relevant spectral peaks of musical tones better than continuous envelope coding.

Advantages of CIS over ACE:

Steady-state signals. CIS preserves continuous envelope coding for all channels; ACE may drop a channel that’s a quiet but persistent target (e.g., a flute note among other instruments).
Theoretical clarity. Research applications often prefer CIS because the per-channel stimulation is independent of the input statistics.

By 2026, ACE is the dominant strategy in clinical use, with most CI users worldwide running ACE-N12 at 900 pps as their default mapping.

FSP / FS4 — Fine Structure Processing

CIS and ACE both discard the temporal fine structure of each channel — the rapid oscillations within each frequency band that carry phase-locked information. Normal hearing uses temporal fine structure for pitch perception (especially fundamental-frequency pitch of complex tones), speech-in-noise, and binaural localisation cues (see Hearing 5.3 — Phase locking refresher →).

MED-EL’s FSP strategy (introduced 2007) and its successor FS4 attempt to encode temporal fine structure on the most apical 1–4 electrodes (the low-frequency channels where natural fine-structure coding lives). The mechanism: instead of fixed-rate pulse trains at 900 pps, the apical electrodes fire trigger-locked to zero-crossings of the band-passed input signal. A 300 Hz input produces 300 pps stimulation on the apical electrode, time-locked to the input’s instantaneous phase.

The strategy is in principle attractive — preserving fine-structure information should improve pitch and music perception. The clinical reality: outcomes are modest and patient-dependent. Some users report better music appreciation and speech-in-noise; others show no measurable advantage over ACE or CIS. The variability has limited FSP’s adoption to the MED-EL device family, where it is the default.

Research strategies (HiResolution, OPTIMA, various academic implementations) push further in the fine-structure direction with sub-millisecond temporal precision and rate-modulated pulse trains, but commercial uptake has been limited.

Strategy fitting parameters

A CI map (the patient’s individualised programming) sets dozens of per-electrode parameters:

T levels (threshold) and C levels (comfort) for each electrode, set behaviourally during programming sessions.
Frequency allocation table: which acoustic frequencies map to which electrodes. Default is logarithmic from ~200 Hz at the apical electrode to ~8 kHz at the basal, but can be customised for partial insertion or specific perceptual needs.
Stimulation rate per channel (typically 900 pps; can be 250–3500 pps).
N value in ACE (typically 8 or 12).
Pulse width (typically 25–50 µs per phase of the biphasic pulse).
Strategy variants (FSP/FS4 trigger frequencies, ACE peak-picking thresholds).

Initial mapping after CI activation takes 60–90 minutes; the map is refined over the first 6–12 months as the user adapts. Most CI users have 3–4 distinct programs in their processor (different N values, different sensitivity settings, etc.) that they switch between for different listening environments.

Strategy choice in clinical practice

The dominant determinant of strategy choice in clinical practice is manufacturer rather than patient-specific tuning:

Cochlear (Nucleus) devices: ACE by default; CIS available as an alternative.
MED-EL devices: FSP / FS4 by default; CIS-equivalent available.
Advanced Bionics devices: HiResolution (a CIS variant with parallel-pair stimulation); other strategies optionally available.

Within a manufacturer, strategy choices are typically left at the default unless a patient is having specific clinical issues (music perception complaints, speech-in-noise deficits) that suggest an alternative might help. Comparing strategies across manufacturers requires switching devices entirely, which is not clinically practical post-implantation.

The implication: the choice of implant matters more than the choice of strategy, because each implant comes with a default strategy. For most adult patients in 2026, the three major manufacturers’ default strategies produce similar speech-in-noise outcomes, with only modest differences in music and pitch perception. The clinical choice of manufacturer is often driven by accessory ecosystem (Bluetooth integration, smart-phone apps, environmental classifier sophistication) rather than core processing-strategy considerations.

Next lesson: what these processing strategies actually produce perceptually — the strange, useful, and ultimately learnable experience of electric hearing.