OPTIMIZING SPIKE-RATE ADAPTATION MECHANISMS IN BIOPHYSICAL MODELS OF HUMAN AUDITORY NERVE FIBERS VIA DIFFERENTIAL EVOLUTION

Temporal coding in cochlear implants (CIs) is influenced by spike-rate adaptation under high-rate pulsatile stimulation, with implications for CI speech perception. While standard Hodgkin–Huxley models capture refractoriness, they fail to replicate adaptation beyond the first milliseconds. We aimed to obtain biophysically grounded model parameters matching experimental cat single-fiber adaptation data via optimization.
Stochastic multicompartment auditory nerve fiber models with human morphologies were extended with adaptation mechanisms: slow sodium channel inactivation and M-type potassium channels. Models included either mechanism alone or both jointly and were fitted separately. Parameters were optimized using differential evolution to match cat single-fiber responses to constant-amplitude pulse trains (250, 1000, 5000 pulses/s) and sinusoidally amplitude-modulated (SAM) pulse trains (5000 pulses/s). Fibers were embedded in a 3D finite-element cochlear model derived from μCT scans of human temporal bones and stimulated via a virtual electrode array.
Optimized models reproduced onset spiking followed by rate- and stimulus-dependent adaptation across pulse rates. The model incorporating both slow sodium inactivation and M-type potassium current provided the best fit, particularly for SAM conditions, indicating that a single mechanism is insufficient to properly capture temporal response patterns. Only the two-mechanism model robustly expressed two adaptation time constants (~5 ms and ~100 ms), consistent with cat single-fiber recordings.
Our models successfully replicated spike-rate adaptation in auditory nerve fibers, highlighting the necessity of multiple adaptation mechanisms. By fitting these mechanisms to cat single-fiber data, we enable studying effects observed in human psychophysical experiments, such as multi pulse integration and stimulation with modulated signals and speech.