Vestibular implant modulation boosts mid and high frequency gaze reflex

In a striking demonstration of how engineering can stand in for biology, researchers in the Netherlands and Switzerland have shown that a vestibulocochlear implant can substantially restore the eye-stabilizing reflex that patients with bilateral vestibulopathy have lost. The study, published in the Journal of Neurology, reports that motion-modulated electrical stimulation of the inner ear’s balance organs significantly improved the vestibulo-ocular reflex (VOR) at both mid and high head-movement frequencies in patients who had previously lived with the disorienting symptom of oscillopsia, the sensation that the world blurs and bounces with every movement of the head.

The VOR is one of the body’s most elegant reflexes. Every time the head turns, the semicircular canals of the inner ear detect the motion and, within milliseconds, drive the eyes to rotate in the opposite direction, keeping the visual world fixed on the retina. In healthy people, eye velocity closely matches head velocity across a broad frequency range, yielding a gain close to one. In bilateral vestibulopathy, a severe chronic loss of vestibular function in both ears, this reflex is largely abolished. The result is a life in which reading a street sign while walking or recognizing a face while turning is a daily struggle. Until now, no curative treatment has existed, and patients have been limited to rehabilitation exercises that help them compensate but do not repair the underlying deficit.

The device at the center of the new findings is a vestibulocochlear implant (VCI), developed by the medical technology company MED-EL and investigated by the Geneva-Maastricht collaboration, the European Institute for Vestibular Implantation. The device combines a cochlear implant, which restores hearing, with a vestibular implant designed to reanimate the balance system. Its latest prototype features three intracanalicular electrodes, one for each semicircular canal, positioned to directly stimulate the ampullary nerves, the branches of the vestibular nerve that relay canal signals to the brain. Because the implanted electrodes mechanically plug the canals and prevent normal endolymph flow, any residual function on the implanted side is essentially eliminated, meaning the artificial signals must do all the work on that side.

The principle behind the implant mimics natural vestibular physiology. A continuous baseline of electrical pulses, delivered at rates between 322 and 388 pulses per second as charge-balanced, cathodic-first, biphasic rectangular pulses with a phase duration of 200 microseconds, stands in for the resting firing rate of healthy vestibular nerve fibers. When the implant’s onboard motion sensors detect head rotation, the pulse amplitude is modulated around that baseline according to a head velocity-to-pulse amplitude sigmoidal mapping. Turning the head toward the implanted ear increases the stimulation above baseline, mimicking excitation; turning away decreases it, mimicking inhibition.

Nine patients with bilateral vestibulopathy and severe hearing loss, all enrolled in the VertiGO! clinical trial, received a unilateral intralabyrinthine VCI. The researchers then assessed the electrically evoked VOR in two frequency domains using complementary laboratory tests. Mid-frequency responses between 0.5 and 2.0 hertz were measured with rotatory chair testing, in which participants were rotated in complete darkness at sinusoidal velocities with a 30 degrees-per-second peak amplitude at 0.5, 1.0 and 1.5 hertz. High-frequency responses above 2 hertz were measured with the video head impulse test (vHIT), in which brief, unpredictable head thrusts exceeding 120 degrees per second are applied while a camera tracks the pupil at 250 hertz.

Crucially, the team compared five stimulation conditions. The Reference condition had the vestibular implant switched off entirely. The Baseline only condition delivered constant, unmodulated pulses. The Default condition modulated the pulse amplitude around a baseline set at 50 percent of each patient’s dynamic range, the span between the threshold of activation and the upper comfortable level. The Lowered baseline condition moved the baseline down to threshold level, freeing more of the dynamic range for excitatory modulation. Finally, the Boosted condition, applied only during vHIT to prevent discomfort during longer chair protocols, used a steeper sigmoidal mapping so that stimulation current rose more rapidly with head velocity.

Before the main experiments, each patient’s peak eye velocity (PEV) was measured during brief two-second block stimuli delivered to each ampullary nerve individually. Six of the nine patients produced peak eye velocities of at least 30 degrees per second during stimulation of the lateral or superior ampullary nerves and were classified as subgroup A, or high responders. The remaining three, all but one of whom had vestibular loss caused by the genetic condition DFNA9, produced minimal responses and were classified as subgroup B, or low responders.

The results were dramatic for the high responders and sobering for the low responders, a split that carries important implications for patient selection and personalized fitting. In rotatory chair testing, subgroup A showed significant increases in excitatory VOR gain under modulated stimulation at all three tested frequencies. Median gains rose from as low as 0.09 in the Reference condition to 0.82 under Lowered baseline stimulation at 1.5 hertz. By contrast, Baseline only stimulation produced no meaningful improvement over the implant-off condition, demonstrating that constant stimulation alone is essentially useless; it is the motion-modulated component that does the therapeutic work. No statistically significant improvements were seen during inhibitory rotations at the group level.

The vHIT results told a similar story at clinically relevant high frequencies. Across all three semicircular canal planes, modulated stimulation significantly improved excitatory median gains. In the lateral plane, median gain climbed from 0.12 with the implant off to 0.69 under Boosted stimulation. In the anterior plane, gains reached 0.74. Posterior canal gains improved as well, though they remained below 0.60, likely reflecting the lower peak eye velocities elicited by acute stimulation of the posterior ampullary nerve. Notably, in one patient with residual anterior canal function, modulated stimulation actually drove gain above 1.0, producing a hyperfunctional, overcompensated eye response, an overfitting problem the authors suggest can be corrected by flattening the stimulation mapping at the relevant head velocities.

Why did the inhibitory direction lag behind? The asymmetry appears rooted in both physiology and hardware. In a healthy ear, vestibular afferent fibers fire at high spontaneous rates, leaving room to decrease firing during inhibition. Electrical stimulation, however, cannot easily encode the saturating behavior of natural inhibition, and lowering the baseline toward threshold appears to erode whatever small inhibitory modulation the Default condition produced. The authors note that this limitation is consistent with previous work showing a reduced contribution of inhibitory inputs to self-motion perception in implant recipients.

For subgroup B, none of the stimulation conditions produced meaningful gain improvements, with gains staying at or below 0.40 across every frequency and plane. The reduced responsiveness may relate to the etiology of vestibular loss; all subgroup B patients had DFNA9, a progressive genetic disorder, and the duration and nature of neural degeneration may influence how effectively the remaining nerve fibers can be driven. Still, the researchers emphasize that even patients without a measurable VOR response might benefit from the implant through other channels of vestibular information that support balance and spatial orientation.

The clinical significance of the gains achieved is substantial. Using established grading criteria, a horizontal canal VOR gain of 0.40 to 0.69 corresponds to moderate vestibular loss, and 0.70 to 0.79 to mild loss, meaning several recipients moved from near-total absence of the reflex into the mild-loss range. Three patients achieved gains approaching physiological levels during lateral and anterior plane testing. Since everyday activities such as walking generate head velocities of roughly 19 to 29 degrees per second, the peak velocities tested in this study fall squarely within the functionally relevant range, suggesting the restored reflex could translate into real reductions in oscillopsia during daily life. The raw traces also showed that when VOR gain improved, corrective catch-up saccades occurred less frequently, with shorter latencies and a more organized pattern, further evidence that the artificial reflex was genuinely doing the visual stabilization work.

The researchers acknowledge the limitations of their exploratory case series, including the small sample size, the loss of usable rotatory chair data from two subjects, and the known test-retest reliability issues of vertical plane vHIT. They also caution that four days of continuous stimulation, the longest period tested, may not capture the neuroplastic changes that could emerge over months or years of daily use, as animal studies have shown peak eye velocities evolve over time and require ongoing refitting.

Looking forward, the team argues that vestibular implant fitting must be fundamentally personalized, with individualized sigmoidal mappings, potentially multimodal inputs such as head acceleration rather than velocity alone, and multi-frequency modulation strategies that optimize performance across the entire spectrum of natural head movement. The next milestones will be home-use trials relating VOR gains directly to dynamic visual acuity and patient-reported oscillopsia. For a condition long considered untreatable, the message of this study is unequivocal: with the right stimulation strategy, the broken reflex can be rebuilt, pulse by pulse.