Stentrode

Stentrode is a brain–computer interface built into a self-expanding stent. Rather than opening the skull, a neurointerventional radiologist threads the device through the jugular vein and into the superior sagittal sinus, the large vein running along the midline of the brain, where it expands against the vessel wall directly above motor cortex. Electrodes on the stent record cortical activity through the vein wall; a lead runs under the skin of the neck to a telemetry unit implanted in the chest, which transmits wirelessly to an external computer.

The design trades signal quality for surgical accessibility. With sixteen electrodes recording population-level activity rather than individual neurons, the Stentrode captures far less information than a penetrating Utah array. It also requires no craniotomy, uses catheter techniques that thousands of clinicians already practise, and does not place foreign material in the brain parenchyma.

§How it works

The stent is a nitinol mesh of the type used for treating venous stenosis, with electrode contacts mounted on its struts. Delivered collapsed inside a catheter, it self-expands on release and presses the electrodes against the endothelium. Over the following weeks the vessel wall incorporates the stent — the same endothelialization that makes vascular stents stable, and which here fixes the electrodes in position without sutures or bone anchors.

Signals reach the electrodes through the vein wall, which acts as a spatial filter. The result resembles what surface arrays record in Electrocorticography interfaces but with fewer, larger, and more distant contacts: local field potentials and band power, not action potentials. Participants generate control signals by attempting movements — typically of a foot or ankle, since the leg representation of motor cortex sits along the midline nearest the sagittal sinus — and a decoder maps the resulting change in band power to a discrete output the company calls a digital motor output, essentially a click.

Because one reliable click is not enough to run a computer, the system is used with eye tracking or switch-scanning software: gaze selects the target, the neural signal confirms it. This division of labour is deliberate. The Neural decoding problem is much easier when the decoder only has to distinguish attempted movement from rest.

§Development history

The concept came from Thomas Oxley and Nicholas Opie's group at the University of Melbourne, who reasoned that the cerebral venous system already reaches within millimetres of cortex and that catheter delivery would avoid the complication profile of open neurosurgery. A 2016 report in sheep showed that an endovascular array could record cortical activity chronically, with signal quality that remained usable as the stent incorporated into the vessel wall.¹ Synchron was founded to commercialize the device.

The first-in-human study, SWITCH, implanted the device in four participants with severe paralysis, mostly from amyotrophic lateral sclerosis, at the Royal Melbourne Hospital beginning in 2019. The published results reported no device-related serious adverse events, no vessel occlusion, and participants able to use the system for texting, email, and online banking in combination with eye tracking.²³ A United States early feasibility study, COMMAND, began in 2022 with the first implant at Mount Sinai in New York, and the company has reported that its primary safety endpoint was met.

1. 2016
   Chronic recording in sheep
   An endovascular stent-electrode array records cortical activity for months, published in Nature Biotechnology.
2. 2019–2021
   SWITCH first-in-human study
   Four participants with severe paralysis receive implants in Australia; the trial reports no device-related serious adverse events.
3. 2022
   First US implant
   The COMMAND early feasibility study begins at Mount Sinai under an FDA investigational device exemption.
4. 2023–2025
   Integration with consumer platforms
   Synchron demonstrates control of commercial voice assistants and announces work with major computing platforms on native input support for neural devices.

§Clinical results and limits

What participants have demonstrated is real but narrow: reliable discrete selection sufficient for communication and device control, sustained over months to years, in people who otherwise depend on eye-gaze systems that fail when eye movement deteriorates. That is a meaningful clinical benefit for a population with few options.

What has not been demonstrated is continuous high-dimensional control. Nothing resembling the robotic-arm work or the fluent Speech neuroprosthesis results obtained with penetrating arrays⁴ has come from an endovascular device, and there is a physical reason to expect that gap to persist. Speech motor cortex lies on the lateral surface of the brain, far from the sagittal sinus; the venous anatomy determines which cortex is reachable, and it does not reach everywhere. Driving a wheelchair or a powered Powered exoskeletons would likewise require continuous multi-axis control that a click cannot supply.

Bandwidth versus access

Synchron's argument is that a device a vascular surgeon can implant in an ordinary catheter lab scales to a patient population that craniotomy never will, and that reliable simple control covers most of the clinical need. Critics respond that sixteen field-potential channels cap the device below the applications that justify an implant at all, and that the comparison should be against non-invasive eye trackers rather than against penetrating arrays.

Other limitations follow from the vascular setting. The device is difficult to remove once the vein wall has grown over it. Participants need antiplatelet therapy, at least initially, and venous thrombosis or stenosis remains the principal safety concern even though the published studies did not observe it. Long-term stability data extend to a few years across a small number of participants, which is not enough to characterize rare complications.

§Strategy and context

Synchron has positioned the Stentrode against both invasive and non-invasive alternatives. Compared with Neuralink and other penetrating systems, it accepts a much lower ceiling in exchange for a procedure that generalists can perform. Compared with Non-invasive neuromodulation and scalp-electrode devices, it offers a stable, always-available signal that does not require setup or tolerate hair and sweat artefacts.

The company has also pursued integration with mainstream computing platforms, so that a neural device appears to an operating system as an input source alongside touch and voice. If that standardization takes hold, it would matter beyond any one device: the persistent complaint from users of assistive Neuroprosthetics is that the hardware works and the software ecosystem does not.

§Outlook

The obvious next step is a pivotal trial in a defined indication — loss of communication in ALS being the clearest — that could support market authorization. No implanted BCI has cleared that bar anywhere, and the Stentrode's low channel count may actually help, because a simpler device with a simpler claim is easier to evaluate. The regulatory precedents that exist are for stimulating devices such as the Cochlear implant and Deep brain stimulation systems, where the therapeutic claim is a change in symptoms rather than a measured rate of communication.

The unresolved technical question is whether channel count can grow. Multiple stents, longer arrays extending further along the sinus, and improved electrode materials have all been proposed. Whether any of that lifts an endovascular interface out of the switch-like regime, or whether the vein wall imposes a hard ceiling on what can be resolved, has not been established. As with every implanted neural device, the governance questions — data ownership, Mental privacy, device support obligations, the Neurorights frameworks now emerging in several jurisdictions — arrive before the clinical ones are settled.

§See also

• Synchron
• Brain–computer interface
• Utah array
• Electrocorticography interfaces
• Neuralink
• Neural decoding
• Neuroprosthetics
• Speech neuroprosthesis

§References