Experiments were conducted in barbiturate-anesthetized cats. Each experiment involved the following steps: (1) A multisite recording probe was implanted in the ICC. (2) ICC responses to acoustical tones in normal-hearing conditions were recorded. Based on the responses to tones, the position of the recording probe was adjusted to sample a range of characteristic frequencies (CFs) from <1 to >32 kHz, and then the probe was fixed in place. (3) The animal was deafened, and a conventional multichannel scala tympani stimulating electrode array was implanted. Responses of the ICC were recorded in response to single electrical pulses delivered through each of the electrodes of the scala tympani array. (4) The scala tympani array was removed and a multisite silicon-substrate stimulating array was inserted into the modiolar trunk of the auditory nerve. Responses of the ICC were recorded in response to single electrical pulses delivered through each of the sites along the intraneural stimulating array. Successful acute experiments were conducted in 12 cats, 10 of which yielded responses to acoustic, scala tympani, and intraneural stimulation; two of the cats yielded only acoustic and scala tympani responses.
Stimulus presentation and data acquisition used System 3 equipment from Tucker-Davis Technologies (TDT; Alachua, FL, USA) and custom software running in MATLAB (The MathWorks, Natick, MA, USA). The electrical pulses were generated by custom optically isolated 8- or 16-channel current sources controlled by 16-bit digital-to-analog converters (TDT RX8). The 8-channel current source used for the scala-tympani stimulation had a maximum output of 3 mA. The 16-channel current source used for the intraneural stimulation had a maximum output of 1 mA.
Recording probes (NeuroNexus Technologies, Ann Arbor, MI, USA) consisted of single silicon-substrate shanks, 15 μm thick and a maximum of 240 μm wide, tapering toward the tip. Each probe had 32 iridium-plated recording sites, 413 μm2
in area, arrayed in a line at 100-μm intervals. The 32-channel neural waveforms were digitized simultaneously, displayed online, and stored on computer disk. Offline, the waveforms were processed to remove interchannel correlated noise, and then single- or multiunit spikes were detected based on spike wave shapes (Middlebrooks 2004
). Well-isolated single units, characterized by discrete clusters in histograms of peak-to-trough spike amplitudes, were recorded at 11.6% of recording sites. All other recordings that were used for quantitative analysis consisted of spikes from unresolved clusters of two or more units. In response to stimulation with single electrical pulses at saturating levels, single units typically responded with mean spike rates of one to five spikes per trial; saturated multiunit responses typically averaged 5–10 spikes per trial. Spike counts were obtained in a time window of 2 to 15 ms relative to the onset of a single electrical pulse or 5 to 40 ms relative to the onset of a 40-ms tone burst. Those time windows were selected based on the inspection of poststimulus-time histograms in pilot experiments.
The right ICC was visualized by aspiration of overlying occipital cortex. Stimuli were presented to the left ear; the right ear was deafened by the removal of the ossicular chain. One recording probe was inserted in each animal, oriented in the coronal plane and 40° from the sagittal plane, passing dorsolateral to ventromedial through the ICC. That orientation corresponds approximately to the tonotopic axis of low-to-high CF in the ICC (Rose et al. 1963
). Acoustical stimuli were presented through a calibrated hollow ear bar. Acoustic tones were 40 ms in duration with 5-ms rise/fall times; the levels of tone stimuli were calibrated from 0.5 to 40 kHz. Frequency response areas were compiled from responses to 5 to 10 repetitions of tones in 1/6-octave intervals and sound levels in 5-dB intervals. The CF at each recording site was the frequency at which neurons responded with the lowest threshold. The recording probe was placed at the desired location relative to the frequency representation based on responses to tones, and then the probe was fixed in place. Upon completion of the presentation of acoustic stimuli, the left cochlea was deafened by scala tympani infusion of an ototoxic agent, neomycin sulfate (10% w/v) (Nuttall et al. 1977
). In our pilot tests, this procedure eliminated all auditory responses measurable by compound-action-potential or ICC-unit recording within ~10 min.
Scala tympani stimulating arrays were an animal version of the clinical Nucleus22 banded electrode array (Cochlear, Ltd., Lane Cove, NSW, Australia), differing from the human version only in that the array was truncated to eight electrodes. The electrodes each were 400-μm-diameter bands, arrayed on a silicone-elastomer carrier, centered at 750-μm intervals. The electrode arrays were inserted through the round window as far as possible into the scala tympani; i.e., until the most apical electrode wedged into the scala. Because the scala tympani arrays were inserted until they would go no further, the apical electrodes probably achieved a tighter fit in the scala tympani, and closer proximity to the modiolus, than typically is achieved in human patients with banded electrode arrays. We estimate that the scala tympani arrays in the present experiments occupied all of the first and the basal half of the second turns of the scala tympani.
The intraneural stimulating arrays consisted of NeuroNexus probes similar to the recording probes. The stimulating arrays had 16 iridium-plated sites, 703-μm2 in area, arrayed at 100-μm intervals spanning a distance of 1.5 mm along a single, 15-μm-thick silicon-substrate shank. The intraneural array was implanted by enlarging the round window margin, then making a hole in the osseous spiral lamina with the bevel of a 30-gauge hypodermic needle and advancing the stimulating array with a micropositioner. The array was oriented approximately in the coronal plane, ~45° from the horizontal plane, passing from ventrolateral to dorsomedial. That orientation was approximately perpendicular to the midmodiolar axis of the cochlea. Several placements of the intraneural array near that general orientation were tested in each animal. In most of the successful array placements, all 16 stimulation sites lay within excitable neural tissue, indicating a >1.5-mm traverse of excitable tissue. Summary data (e.g., spread of activation in Fig. ) represent all the tested stimulating sites from a single array placement in each cat. The representative array placement selected from each cat was the one that gave the broadest range of activation of the ICC frequency representation; in some cases, all 16 stimulation sites were tested, whereas in other cases, only eight sites were tested, at 200-μm intervals.
FIG. 7 Distributions of spread of ICC activation. These box-and-whisker plots show the distribution among animals and stimulating electrodes (or tone frequencies) of spread of ICC activation. The datum for each case is the percent of ICC recording sites at which (more ...)
Scala tympani electrodes were stimulated in two configurations. In the monopolar configuration, the active electrode consisted of one scala tympani band and the return electrode was a wire in a neck muscle. In the bipolar configuration, the active electrode was one of the scala tympani bands and the return electrode was the adjacent more-apical scala tympani band. In most conditions, electrical stimuli for scala tympani and intraneural stimulation were single biphasic pulses, 41 μs per phase, initially cathodic. One exception was the situation in which thresholds for scala tympani stimuli were so high that it was necessary to lengthen the phase duration to 205 μs to achieve adequate stimulus levels. In that condition, the reported currents are multiplied by 205/41 to permit an equivalent-charge comparison with stimuli that used 41-μs phase durations. That procedure was used for monopolar stimulation in one animal and bipolar stimulation in three animals; the example illustrated in Figure used the standard phase duration of 41 μs per phase. The only other exceptions were the tests of two-pulse temporal integration presented in Figures and and accompanying text. In that condition, stimuli were pairs of monophasic cathodic pulses, 20 μs per phase. All intraneural stimulation employed a monopolar configuration, with single electrode sites on the multielectrode arrays serving as the active electrodes and a wire in a neck muscle serving as the return electrode.
FIG. 2 Spatial tuning curves elicited by individual and simultaneous monopolar stimulation of two scala tympani electrodes. The left ordinate axis plots recording depth in the ICC relative to the most superficial recording site; corresponding CFs are plotted (more ...)
FIG. 11 Two-pulse temporal integration. (A) plots normalized thresholds for two-pulse stimulation as a function of interpulse-interval. Data are from a single ICC recording site tested with intraneural (X) and scala-tympani (O) stimulation. Thresholds were normalized (more ...)
FIG. 12 Distribution of time constants for two-pulse temporal integration. Time constants were computed according to Eq. 1 for individual ICC recording sites in four animals. (A) Monopolar scala tympani stimulation (mean and standard deviation: 260±85 μs). (more ...)
The sensitivity of neural spike counts to changes in stimulus level was quantified by a procedure derived from signal detection theory (Green and Swets 1966
; Macmillan and Creelman 2005
). The stimulus-driven growth of response at each of the 32 ICC recording sites was represented by the discrimination index (d
′) for the discrimination of stimulus levels differing by 5 dB (for sounds) or 1–2 dB (for electrical pulses); a no-stimulus condition was also included in each stimulus set. Based on responses on 20 trials of a given stimulus level and 20 trials of the next higher level, we formed an empirical receiver operating characteristic (ROC) curve based on the trial-by-trial distribution of spike counts among lower- and higher-level trials. The area under the ROC curve gave the probability of correct discrimination, which was expressed as a standard deviate (z
-score) then multiplied by √2 to obtain d
′. This procedure was repeated for each successive pair of stimulus levels. The contours in the figures show d
′ cumulated across increasing stimulus levels. Threshold was taken as the interpolated stimulus level at which cumulative d
1. The threshold obtained from cumulating d
′ over successively increasing stimulus levels gave essentially the same value as that obtained from the comparison of various stimulus levels to a no-stimulus condition; across 2,468 cases of recording site and stimulating electrode, 98.5% showed a difference between the two procedures of no greater than 1 dB (mean
0.079 dB). The advantage of using cumulative d
′ in this way is that it gave a measure of the ability of neurons to signal increasing stimulus levels well beyond the stimulus level at which stimuli could be detected almost perfectly (i.e., beyond d
3.3). This signal-detection procedure was favored over, for instance, a measure of spike rate divided by standard deviation of the rate because the signal-detection procedure explicitly incorporated the trial-by-trial variation in spike rates for both lower- and higher-level stimuli. Also, values expressed as d
′ lend themselves to comparison with psychophysical results.