A Model-Based Approach for Separating the Cochlear Microphonic from the Auditory Nerve Neurophonic in the Ongoing...

Area of Science:

• Auditory Neuroscience
• Biomedical Engineering
• Otoacoustic Emissions

Background:

• Electrocochleography (ECochG) shows promise for predicting speech perception in cochlear implant (CI) recipients.
• Current ECochG analysis struggles to distinguish hair cell (cochlear microphonic, CM) and neural (auditory nerve neurophonic, ANN) contributions in low-frequency tone responses.
• Separating CM and ANN signals is crucial for understanding residual auditory physiology in CI users.

Purpose of the Study:

• To develop and validate a model capable of separating mixed CM and ANN signals from ECochG recordings.
• To quantify hair cell and neural contributions in ECochG responses, particularly for low-frequency tones.
• To assess the model's performance using human CI data, animal models, and simulated signals.

Main Methods:

• A source property-based model was developed to analyze ECochG waveform shapes.
• The model incorporates independent parameters for CM (sinusoidal, saturation) and ANN (unit potential convolution, spread of excitation).
• Adaptive fitting was used to identify CM and ANN parameters that best reproduced recorded and simulated waveforms.

Main Results:

• The model accurately fit ECochG waveforms from 284 CI recipients with an average R-squared of 0.97.
• In gerbils, neurotoxins significantly reduced the ANN, particularly at frequencies ≤1000 Hz, while CM remained largely unaffected.
• ANN contributions in human CI subjects were highly variable, ranging from negligible to exceeding CM contributions.

Conclusions:

• The developed model effectively isolates hair cell and neural activity within mixed ECochG responses.
• This tool enables characterization of residual auditory physiology in CI recipients and other clinical applications.
• Quantifying CM and ANN components can improve our understanding of auditory function and potentially predict CI outcomes.