Frequency following responses and rate change complexes in cochlear implant users

Abstract

The upper limit of rate-based pitch perception and rate discrimination can differ substantially across cochlear implant (CI) users. One potential reason for this difference is the presence of a biological limitation on temporal encoding in the electrically-stimulated auditory pathway, which can be inherent to the electrical stimulation itself and/or to the degenerative processes associated with hearing loss. Electrophysiological measures, like the electrically-evoked frequency following response (eFFR) and auditory change complex (eACC), could potentially provide valuable insights in the temporal processing limitations at the level of the brainstem and cortex in the electrically-stimulated auditory pathway. Obtaining these neural responses, free from stimulation artifacts, is challenging, especially when the neural response is phase-locked to the stimulation rate, as is the case for the eFFR. In this study we investigated the feasibility of measuring eFFRs, free from stimulation artifacts, to stimulation rates ranging from 94 to 196 pulses per second (pps) and eACCs to pulse rate changes ranging from 36 to 108%, when stimulating in a monopolar configuration. A high-sampling rate EEG system was used to measure the electrophysiological responses in five CI users, and linear interpolation was applied to remove the stimulation artifacts from the EEG. With this approach, we were able to measure eFFRs for pulse rates up to 162 pps and eACCs to the different rate changes. Our results show that it is feasible to measure electrophysiological responses, free from stimulation artifacts, that could potentially be used as neural correlates for rate and pitch processing in CI users.

Keywords: Artifact removal; Auditory change complex; Cochlear implants; Frequency following response; Pitch; Rate; Stimulation artifacts; Temporal processing.

Fig. 1

The EEG of two trials of the 94 – 196 pps condition. The first 2.048 s were stimulated at 94 pps and the second 2.048 were stimulated at 196 pps. The red dashed line indicates the deviant-to-base change, the orange dashed line the base-to-deviant change, and the blue dashed line indicates the timepoint where the reduction in amplitude due the compliance issue starts . (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 2

Example of a stimulation artifact recorded ipsilateral to the CI of S1. A) illustrates the blanking procedure, where the red section is removed by linear interpolation from the EEG recording after each stimulation pulse. B) the first pulse of A showing that the hyper-rate EEG system was able to sample the artifact well. It shows the RF-power pulse, implemented for CIC4 implants followed by the stimulation pulse. The signal shown here is averaged across trials (n = 387), unreferenced (i.e. relative to CMS) and filtered to compensate for the drift in the recording over time. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 3

Artifact removal assessment showing A) the response amplitude at the stimulation rate (here we used the 164 pps pulse rate) as a function of the blanking length for each individual subject. Colors show the different non-inverting channels and the different symbols and line types show the different reference channels (Ref. Chan.). The solid black line and the shaded area shows the average, and range of noise levels across all recording-electrode configurations as a function the blanking length, respectively. Response amplitudes that were lower than the noise level for a specific combination (i.e. recording-electrode configuration and blanking length) were set to the noise level of that specific combination for illustrative purposes. The dashed vertical line shows the blanking length used in the analysis of the eFFR. B) The latency (group delay) of the response as a function of blanking length. The latency was based on the significant responses to the 128- and 164- pulse-rate stimuli. Latencies could only be estimated when both responses were significant. The lack of latencies in e.g. S5 for blanking lengths > 2 ms is a result of this and indicate that the response was not significantly different from the noise level, which is also clear from Panel A. The ear of stimulation is shown for each subject in the lower-right corner of subject-specific graph and the EEG-electrode on the scalp is shown on the right-hand side of Panel B.

Fig. 4

Example of the effect of the blanking length on the response amplitude and the neural background noise amplitude for the different eFFRs to the different pulse rates. Data shown are from a vertical configuration (i.e. left mastoid – Cz) of S1. Vertical dashed lines are shown for illustrative purposes and show the point at which the blanking length exceeds the 50% point of the inter-pulse interval for each specific pulse rate. For illustrative purposes the blanking length was set to a maximum of 5 ms, which corresponds to a removal of 98% of the waveform for the 196-pps pulse rate.

Fig. 5

A) The number of significant eFFRs as a function of pulse rate each (n = 5). B) The response amplitude (solid lines) and neural background amplitude (dashed lines) as a function of pulse rate. . The response amplitude line is set to the average of the three measurements at 94 pps, and only significant responses are shown (p ≤ < 0.05). C) The phase as a function of pulse rate, only significant responses are shown (p  <0.05) and the lines represent the phase lag between the 128- and 162-pps eFFRs that are used to calculate the latencies. D) The latencies derived from the phase lag of the ipsi and contralateral positioned recording electrodes. Color coding is the same in Panel B, C, and D.

Fig. 6

Individual ACC waveforms as a function of rate change. The waveforms have an offset per rate change to aid visibility (horizontal black dashed lines). Subjects S1 and S5 do not have an ACC to the base-to-deviant condition due to compliance issues (see text for details).

Fig. 7

A) The number of significant eACCs as a function of deviant pulse rate, note that only 3 subjects were included for the base-to-deviant change. B) The response N1 latency as a function of the deviant pulse rate. Only responses that are significantly different from the background noise are shown. C) The absolute response amplitude of the N1 peak and neural background noise (mean across electrodes) as a function of the deviant pulse rate. Color coding is the same in A, B, and C.