A new approach for the determination of ECAP thresholds.

BACKGROUND

Electrically evoked compound action potentials (ECAPs) of the auditory nerve are routinely recorded for testing the cochlear implant integrity and its functional connection to the auditory system. The response thresholds derived from ECAP recordings are widely used as a helpful guide in the fitting of the dynamic range of electric stimulation, although they may not always predict the behavioral thresholds of individuals well. Conventionally, this threshold is based on the identification of a minimum N peak and maximum P peak and linear extrapolation of the resulting amplitude growth function (AGF). As an alternative, a new procedure involving numeric signal processing and requiring less user intervention is presented here. Data acquisition: In 12 adults implanted with MED-EL FLEX28 electrodes, two series of ECAPs were recorded immediately after implantation: (i) a full profile involving all 12 channels across the whole stimulus range in steps of 200 current units and (ii) a high resolution section (20 records in the immediate neighborhood of the threshold) of the AGF in one selected channel. Data treatment: It was observed that N and P wave latencies do not depend on stimulus intensity. Fixed time windows were hence defined for stimulus plus noise and noise alone regions. In these windows, the variance of the compound signal representing response and noise is extracted, whereas the noise variance is extracted from the tail of the curve following this time window. The base line is corrected by fitting an exponential function to reduce stimulus or amplifier artifacts. The response threshold is then derived from the response to noise ratio which should exceed the limit of 6 dB.

RESULTS

The ECAP thresholds obtained from the new procedure coincide well with those determined by the conventional linear extrapolation of the AGF and they correlate to a greater degree with psychometric thresholds than the existing approach.

CONCLUSIONS

The new ECAP algorithm looks promising and may reduce the need for user intervention in determining thresholds.