Subjects, device and surgery
Three bilateral vestibulopathy patients who previously received a vestibular implant prototype , participated in this study. Details on the inclusion criteria, device and surgical procedures can be found in previous publications [5, 7, 24]. Briefly, the device consisted of a modified cochlear implant (MED-EL, Innsbruck, Austria) providing one to three extra-cochlear electrodes for vestibular stimulation (Table 1). These vestibular electrodes were implanted in the vicinity of the lateral, posterior and superior ampullary branches of the vestibular nerve (respectively, LAN, PAN and SAN) using an intralabyrinthine or extralabyrinthine surgical approach [20, 23, 38]. Note that the PAN electrodes in patients S1 and S2 were not tested during the experiments presented here (grayed out in Table 1). Stimulation with these electrodes did not evoke any vestibular responses even at the highest current levels available for safe stimulation. This is probably due to the traumatic etiology of these cases (temporal bone fracture going through the ampulla of the PAN).
All patients were recruited at the Division of Otorhinolaryngology and Head and Neck Surgery of the Geneva University Hospitals. Note that only one vestibular electrode was activated at a time for a given experimental trial. All cochlear electrodes were switched off during the experimental procedures.
The setup for the electrical stimulation was composed of a computer running custom software based on MATLAB R2014b (The Mathworks Inc., Natick, Massachusetts, USA). This software allowed customization of stimulation parameters (current intensity, pulse rate, phase width, electrode, current range, train pulse characterization, etc.). The computer communicated this information to the implanted stimulator via a special interface device (dRIB; MED-EL, Innsbruck, Austria) and the system’s antenna.
Each experimental trial consisted of 100 electrical stimuli presented at a repetition rate of 5 Hz. The electrical stimuli involved one or several cathodic-first, biphasic, charge balanced pulses delivered to the vestibular nerve with one of the implanted vestibular electrodes. First, we investigated the growth function of the responses of the three vestibular pathways (VOR, VCR, and VTC), using a single pulse with 200 µs phase duration (S1-SAN, S1-LAN, S2-SAN, S2-LAN and S3-PAN). Second, we compared the growth functions obtained with three different stimulation paradigms on one subject (S1-SAN and S1-LAN) who was available for this additional experiment. The three stimulation paradigms were (1) a single pulse with a phase of 200 µs (as used in the previous experiment and in our preceding studies ); (2) a single pulse with a short phase of 50 µs (similar to that commonly used in clinical cochlear implant fittings); and (3) a pulse train of four 50 µs/phase pulses presented at rate of 1600 pulses-per-second (pps) (total charge per stimulation trial equal to the single pulse with a 200 µs phase duration).
Characteristics and growth functions of the VOR, VSR, and VTC pathways
First, a measurement without any electrical stimulation (0 µA) was performed to record baseline response levels (e.g., noise). Then consecutive experimental trials were performed with increasing current amplitude (steps of 50 µA) to investigate the characteristics and the growth functions of each vestibular response, up to the upper comfortable level (UCL). The UCL is defined to be the current level immediately below the level where undesired effects are observed (i.e., facial nerve activation, uncomfortably loud sound) or at the maximum safe current level allowed by the device, similar to our previous studies . Note that this experimental design involving very short stimulation trials did not comprise special psychophysical paradigms to compensate for adaptation effects (e.g., ascending/descending) which would have resulted in increased experimental times. In our experimental conditions, the bias induced by increased experimental time would have been greater than that induced by any potential adaptation effects.
Figure 1 shows an example of one of the recordings for S1 obtained upon electrical stimulation of the vestibular electrode implanted in the proximity of the SAN. The markers in the figure illustrate the different time points that were considered in the analysis of latency and amplitude of the different responses, explained in detail below.
Electrically evoked VOR responses (eVOR) were recorded using a binocular, video-based eye tracking system at a high sampling rate (1000 Hz) to allow acquisition of short-latency eye-movements (EyeLink 1000 Plus; SR Research, Ottawa, Canada). Some patients suffered from strabismus which hindered accurate binocular fixation; therefore, only the dominant eye was recorded in all subjects. Each experimental trial started with a calibration procedure which consisted of nine sequential fixations of a dot moving randomly around the computer screen borders, followed by a similar nine-point validation procedure to ensure calibration accuracy (error < 0.1°). Horizontal and vertical eye velocity data from the EyeLink system were imported to MATLAB R2018b. Peak eye velocity (PEV) of the signal was calculated as the square root of the sums of the squares of horizontal and vertical eye movements. Trials including artefacts (saccades or blinks) were manually removed. Then the average of artefact-free PEV responses was calculated. The latency of the electrically evoked vestibulo-ocular reflex (eVOR) is determined by the beginning of the eye movement, calculated as the first inflection point of the total PEV signal (LATeVOR). The consecutive eVOR peaks were determined using the maximum or minimum, depending on signal polarity, of the second derivate of the total PEV signal (see Fig. 1a).
Electrical cervical vestibular evoked myogenic potentials (ecVEMPs) were recorded with the NeuroAudio system (Neurosoft, Ivanovo, Russian Federation), with the active recording electrodes positioned on the main belly of the sternocleidomastoid muscle (SCM), approximately equidistant from the mastoid process and the sternum. The ground electrode was placed on the superior part of the sternum, and the indifferent electrode on the forehead. Instead of being in the standard supine position and lifting the head, we had patients sit down with the head placed on a head support tower. The patient was requested to look straight ahead to a 24″ computer screen (XL2420-B; BenQ, Taipei, Taiwan) projecting a 12 mm-wide cross (eye-to-screen distance 63 cm). Sufficient SCM tension was obtained by having the patient turn the shoulders slightly. This non-standard patient configuration was necessary to allow simultaneous recording of eye movements and to limit patient fatigue resulting from repeated testing. ecVEMP results for each experimental trial were amplified, averaged, and imported into MATLAB R2018b (The Mathworks Inc., Natick, Massachusetts, USA). The signals were then low-pass filtered at 500 Hz using a ninth-order IIR filter with zero frequency shift. The latencies and amplitudes of the positive (P1) and negative (N1) peaks were determined using the Matlab function “findpeaks”. When multiple peaks were identified by this function for a given wave, the optimum peak was selected in consensus by four experienced clinical observers (authors AB, NG, SC, MR and APF). The latency of the ecVEMPs was calculated as the latency of the first peak of muscular contraction (P1), which corresponds to the initiation of the neck movement.
After each stimulation trial, the patient had to report the self-perceived intensity of the stimulus using the clinical 0–8 visual-analog scale (0—no perception, 8—too strong) used for fitting cochlear implant patients in our center. Patients were also asked to describe the percept. Only percepts that could be identified as vestibular in consensus between experimenters were considered (see also ). For example, percepts evoking motion or disorientation were included, while percepts evoking sound, pain, or tickling were not considered.
Data analysis and statistics
All analyses were carried out with SigmaPlot 14 (Systat Software, San Jose, CA, USA) and will be presented in “Results”.