Assessment of Safety of a Fully Implanted Endovascular Brain-Computer Interface for Severe Paralysis in 4 Patients: The Stentrode With Thought-Controlled Digital Switch (SWITCH) Study

Abstract

Importance: Brain-computer interface (BCI) implants have previously required craniotomy to deliver penetrating or surface electrodes to the brain. Whether a minimally invasive endovascular technique to deliver recording electrodes through the jugular vein to superior sagittal sinus is safe and feasible is unknown.

Objective: To assess the safety of an endovascular BCI and feasibility of using the system to control a computer by thought.

Design, setting, and participants: The Stentrode With Thought-Controlled Digital Switch (SWITCH) study, a single-center, prospective, first in-human study, evaluated 5 patients with severe bilateral upper-limb paralysis, with a follow-up of 12 months. From a referred sample, 4 patients with amyotrophic lateral sclerosis and 1 with primary lateral sclerosis met inclusion criteria and were enrolled in the study. Surgical procedures and follow-up visits were performed at the Royal Melbourne Hospital, Parkville, Australia. Training sessions were performed at patients' homes and at a university clinic. The study start date was May 27, 2019, and final follow-up was completed January 9, 2022.

Interventions: Recording devices were delivered via catheter and connected to subcutaneous electronic units. Devices communicated wirelessly to an external device for personal computer control.

Main outcomes and measures: The primary safety end point was device-related serious adverse events resulting in death or permanent increased disability. Secondary end points were blood vessel occlusion and device migration. Exploratory end points were signal fidelity and stability over 12 months, number of distinct commands created by neuronal activity, and use of system for digital device control.

Results: Of 4 patients included in analyses, all were male, and the mean (SD) age was 61 (17) years. Patients with preserved motor cortex activity and suitable venous anatomy were implanted. Each completed 12-month follow-up with no serious adverse events and no vessel occlusion or device migration. Mean (SD) signal bandwidth was 233 (16) Hz and was stable throughout study in all 4 patients (SD range across all sessions, 7-32 Hz). At least 5 attempted movement types were decoded offline, and each patient successfully controlled a computer with the BCI.

Conclusions and relevance: Endovascular access to the sensorimotor cortex is an alternative to placing BCI electrodes in or on the dura by open-brain surgery. These final safety and feasibility data from the first in-human SWITCH study indicate that it is possible to record neural signals from a blood vessel. The favorable safety profile could promote wider and more rapid translation of BCI to people with paralysis.

Trial registration: ClinicalTrials.gov Identifier: NCT03834857.

Figure 1.. Motor Neuroprosthesis System Overview

A, Schematic of the fully implanted brain-computer interface (BCI). A device with electrodes is implanted in the superior sagittal sinus blood vessel (inset) and connected to an implantable receiver transmitter unit (IRTU) in the subcutaneous pocket. IRTU communicates with an external receiver telemetry unit (ERTU), which relays signals to a signal control unit for controlling a laptop computer or tablet. B, BCI with an eye tracker for computer control. Eye tracking is used to move the cursor, and BCI is used to click. C, BCI without eye tracking for computer control. An item scanner highlights items in sequence, and the BCI is used to click an item when it is highlighted.

Figure 2.. Endovascular Electrode Array Placement and Migration Across a 12-Month Period

A, Location of the recording device in the superior sagittal sinus adjacent to motor cortex. The precentral gyrus can be identified immediately anterior to the central sulcus (cyan line). The left image shows a 3-dimensional rendering of the preimplantation magnetic resonance imaging, with regions of significant activation during attempted movement and the implanted electrodes highlighted (green dots). The right image shows a 3-dimensional rendering of computed tomography (CT) taken 3 months after implantation. B-D, Stable device position at 3 and 12 months after implantation. B, Bars represent distance of the anterior-most electrode from a fiducial point measured from CT at 3 months (dark blue) and 12 months (light blue) after implantation, depicting submillimeter differences between time points for all patients (mean [SD] distance, 0.45 [0.28] mm). C, Magnified CT taken before as well as 3 and 12 months after the implantation. D, Whole-head CT at 12 months after implantation.

Figure 3.. Feasibility Evaluations of an Endovascular Brain-Computer Interface

A, Signal bandwidth for all 4 patients (P1, P2, P3, and P4) measured across a 12-month period. Shading depicts SD across channels. B, Mean and single-trial spectrography of neural activity during movement attempt for P4. Feature selection for classification was patient specific, with β band (13-30 Hz) power chosen in this example. The gray time-series trace depicts the 13-30 Hz bandpass-filtered signal, and the black traces depict β power normalized to the premovement rest period. Time 0 indicates cued attempted movement onset. Shaded areas indicate standard error of the mean across trials. C, Offline movement classification balanced accuracy for all 4 patients. Colored areas show the accuracy calculated with cross-validation, and gray areas show chance-level accuracy calculated by shuffling the labels. D, Performance of online BCI control for all 4 patients. The left panel depicts correct characters per minute (CCPM); middle panel, character accuracy; and right panel, response accuracy from typing test. Errors bars indicate SDs across sessions.