Design and performance of a multichannel vestibular prosthesis that restores semicircular canal sensation in rhesus monkey

Abstract

In normal individuals, the vestibular labyrinths sense head movement and mediate reflexes that maintain stable gaze and posture. Bilateral loss of vestibular sensation causes chronic disequilibrium, oscillopsia, and postural instability. We describe a new multichannel vestibular prosthesis (MVP) intended to restore modulation of vestibular nerve activity with head rotation. The device comprises motion sensors to measure rotation and gravitoinertial acceleration, a microcontroller to calculate pulse timing, and stimulator units that deliver constant-current pulses to microelectrodes implanted in the labyrinth. This new MVP incorporates many improvements over previous prototypes, including a 50% decrease in implant size, a 50% decrease in power consumption, a new microelectrode array design meant to simplify implantation and reliably achieve selective nerve-electrode coupling, multiple current sources conferring ability to simultaneously stimulate on multiple electrodes, and circuitry for in vivo measurement of electrode impedances. We demonstrate the performance of this device through in vitro bench-top characterization and in vivo physiological experiments with a rhesus macaque monkey.

Fig. 1

(A) MVP2 Circuit Schematic. The microcontroller reads sensor inputs every 10ms, then pulse-frequency-modulates biphasic charge-balanced pulses one each of six channels (one for each component of 3D rotation and linear acceleration). Two power supplies, +3V and +12V, are generated from a 3.7V Li-ion battery. Four current sources (B) and four current sinks (C) are connected to electrodes via a crosspoint switch array. An Electrode Potential Amplifier (D) can connect to any pair of electrodes; its output is read by the microcontroller. Gray lines = digital. Black lines = analog.

Fig. 2

Comparison of the original (MVP1) and new (MVP2) multichannel vestibular prostheses. The height/thickness of MVP2 is less than half that of MVP1, mainly due to replacement of single-axis gyroscope daughter boards with a flush-mounted dual axis gyroscope.

Fig. 3

MVP2 electrode array designed using 3D reconstructions of CT images of a rhesus monkey labyrinth. Surgical implantation into the three semicircular canals (SCCs) is facilitated by manipulation of two silicone carriers, one that positions 6 electrodes (E4-9) within the anterior and horizontal ampullae and a second that positions E1-3 in the posterior SCC ampulla. Having multiple electrodes near each ampullary nerve allows post-surgical adjustment of nerve-electrode coupling under software control. The microelectrode array has two reference electrodes: a large-area distant reference (E10) and a smaller return electrode meant for insertion at the common crus of the vestibular labyrinth (E11).

Fig. 4

(A) Pulse-frequency modulation of electrode pulse rates by gyroscope inputs. In each case, the appropriate channels modulate. Pitch and roll elicit the appropriate antiphase and in-phase responses, respectively, on left-anterior/right-posterior (LARP) and right-anterior/left-posterior (RALP) channels. Z denotes yaw channel. (B) When prosthesis is mounted 20 cm off the rotator axis, electrode channels temporarily set to encode gravitational acceleration, centripetal acceleration, and tangential acceleration modulate appropriately.

Fig. 5

Mean head and eye angular velocities of a macaque during 2 Hz, 50°/s head rotations in darkness, about the horizontal (top), left-anterior/right-posterior (LARP- middle) and right-anterior/left-posterior (RALP - bottom) SCC axes. Column 1: Prior to lesion. Column 2: After bilateral intratympanic gentamicin, plugging all SCCs, and electrode implantation in left labyrinth. Prosthesis pulsing at baseline rate on all channels, but not modulating with head rotation. Column 3: With prosthesis modulating to encode gyro signals, after 3 days of prosthetic stimulation. Standard deviation of each trace at each time point is < 10°/s. Traces are inverted about the zero velocity axis as required to facilitate comparison; first half cycle represents excitation of left labyrinth in each case. N=20 cycles for each trace. Blanks indicate removal of nystagmus quick phases, which occur as needed to return the eye toward center position.

Fig, 6

Potential waveform and corresponding electrode impedances (inset) for each of 10 intralabyrinthine electrodes measured in series with a much larger distant reference (E10) using the EPA amplifier. Within each case, symmetric constant-current biphasic pulses were 150 µA peak and 200 µs per phase.

Fig. 7

Eye movements in response to tripolar stimulation with varying proportions of a 200 µA/phase, 200 µs/phase, biphasic cathodic-first stimulus current pulse (via electrode E3 near the left posterior SCC) returned by either a distant reference (E10, square), near/intralabyrinthine reference (E11, diamond), or a proportional distribution between the two (circles). α = fraction of current returned via E10. During tests, the monkey was stationary in darkness, and the MVP2 was set to modulate pulse rate on E3 as required to simulate a 1 Hz, ±300 °/s sinusoidal head rotation about the LP SCC axis. Peak eye movement response amplitude and misalignment (angle between desired and observed 3D axis of aVOR eye response) were computed as the mean (±SD) for 10 cycles at each α. Asterisk indicates α’s at which amplitude was significantly better (p<0.01) than the α=1 case while misalignment was significantly less (p<0.01) than the α=0 case. For comparison, the dashed horizontal line indicates the peak velocity and misalignment of the eye response to RALP rotation with the prosthesis off (Figure 5 middle column, bottom row) showing the residual eye movement after gentamicin treatment without prosthetic stimulation.