People with severe paralysis regained control of a computer at home

Stentrode, with a custom decoder and Tobii Dynavox software. [1]

A support-vector machine with a radial-basis kernel for participant 1, and a manually tuned threshold classifier for participant 2. [1]

No software release or trained-weight identifier is reported. NCT03834857 identifies the trial, not the model. [1]

Synchron supplied Stentrode. The research team developed the decoding method, and Tobii Dynavox supplied eye tracking and computer-access software. [1]

The decoder generated clicks from attempted movements. Eye tracking moved the cursor. [1]

The first participant used a trained SVM. The second used a threshold that changed during typing. Both produced short and held clicks. [1]

Electrical signals recorded from a cerebral vein were converted into no click, a short click, or a held click. [1]

Participants controlled selections through attempted movements. The system did not choose messages or purchases for them. [1]

A caregiver attached the receiver and opened the software. The participant then used eye movements and brain-generated clicks to operate a computer. [1]

Clinicians selected and implanted participants, monitored safety, and managed antiplatelet treatment. Researchers tuned decoders, and caregivers helped with setup at home. [1]

Participant 1 used a support-vector machine trained on movement attempts. Participant 2 used a manually tuned threshold classifier. Both converted brain signals into clicks, while an eye tracker handled cursor movement. The measured benefit belonged to the combined system.

The learned decoder supplied the click channel for one of the first two participants. The other used a threshold rule. Eye tracking and commercial software supplied the rest of the interface. [1]

AIBD-0052-outcome-01 [1]

Participants regained the ability to select items and perform everyday computer tasks at home. [1]

Improved Accessibility

Demonstrated in a small, uncontrolled clinical study. [1]

People.

People with severe upper-limb paralysis who retained suitable brain activity and venous anatomy. [1]

Four unique implanted participants are described across both reports. The initial two are included in the later four. [1] [2]

The four implanted participants were men of European ancestry. The small, selected group does not establish broad suitability. [1]

The original report covered 12 months for participant 1 and three months for participant 2. The later paper followed four implanted participants for 12 months each. [1]

Yes

Implantation required anesthesia and antiplatelet treatment. One participant fainted briefly after surgery. The later four-person report recorded eight mild device effects. [1]

NCT03834857 [1]

A prospective, open-label feasibility study without a control group. The later report includes the first two participants. [1]

Researchers measured typing and selection accuracy, observed everyday computer tasks, recorded adverse events, and checked vessel patency and device position. [1]

Two implanted participants appeared in the first report. The later study enrolled five people, implanted four, and excluded one because of venous anatomy. [1]

Correct characters per minute, selection accuracy, completion of everyday tasks, serious device-related events, vessel occlusion, and migration. [1]

There was no concurrent control. The comparison with the Utrecht interface involved different participants and protocols. [1]

The first two participants typed 13.81 and 20.10 correct characters per minute, with selection accuracy above 92%. [1] [2]

The study demonstrates use of the combined implant, decoder, eye tracker, and software. It does not isolate the benefit of machine learning. [1]

The cohort was small and selected, and decoders differed between participants. Without a control group, superiority over other assistive systems cannot be established. [1]

The first participants began home use without research staff on days 86 and 71 after implantation. Caregiver assistance with setup remained necessary. [1]

The 2020 supplement gives participant-specific filters, training splits, and click rules. It does not publish exact trained weights or a software release. [1]

The reviewed sources do not report a monetary valuation or financial return. [1] [2]

The reviewed sources do not report a monetary valuation or financial return. [1] [2]

The reviewed sources do not report a monetary valuation or financial return. [1] [2]

The reviewed sources do not report a monetary valuation or financial return. [1] [2]

The reviewed sources do not report a monetary valuation or financial return. [1] [2]

The reviewed sources do not report a monetary valuation or financial return. [1] [2]

The reviewed sources do not report a monetary valuation or financial return. [1] [2]

Surgery, follow-up, equipment, and caregiver time were required, but the reports do not price them or value the regained abilities. [1] [2]

The reviewed sources do not report a monetary valuation or financial return. [1] [2]

The reviewed sources do not report a monetary valuation or financial return. [1] [2]

The reviewed sources do not report a monetary valuation or financial return. [1] [2]

Synchron and University of Melbourne researchers developed the interface. The Royal Melbourne Hospital implanted and followed participants, who used it at home. [1]

The first clinical study and home use took place in Melbourne, Australia. Synchron was based in the United States and worked with Australian investigators. [1]

The reviewed clinical sources identify the organizations involved, but do not establish their complete ownership or control arrangements. [1]

The system required implantation, a chest receiver, a computer, eye tracking, individual training, caregiver support, and clinical follow-up. [1]

The reviewed sources do not document international transfers of patient data or quantify cross-border financial flows. [1]

Not assessed

Suitable anatomy, preserved motor signals, training, caregiver setup, and continuing clinical care were required. [1]

The later four-person report includes the original two participants. [2]

The four-person follow-up found no serious device-related events over 12 months. Eight mild device effects occurred, and one participant later demonstrated texting without eye tracking. [2]

The reports concern an investigational device. They do not establish routine commercial access at the time of the study. [1]

Australian investigators tested the safety and feasibility of an endovascular brain–computer interface. [1] [2]

The study received hospital ethics approvals and authorization under Australia’s Clinical Trial Notification scheme. [1] [2]

Investigators selected participants, implanted the device, and monitored home use and safety. [1] [2]

The primary safety outcome was a device-related serious event causing death or permanent disability within 12 months. [1] [2]

2 retained sources.

Original citations and the passages that support each finding. [1] [2]

English. No translation was used. [1]

Synchron funded and participated in the later study. An independent safety monitor reviewed safety data, but this does not make the whole efficacy study independent. [2]

The original studies and their supporting materials are linked to the findings that they support. [1] [2]

Access and version details are listed for each reviewed source. [1] [2]

Stentrode supported independent computer tasks after caregiver setup, with eye tracking for cursor navigation. [1]

The first paper documents the two decoders and home tasks. The later paper adds two participants and longer safety follow-up. [1] [2]

Original peer-reviewed studies by the developers and clinical team. The four-person follow-up is an extension, not an independent replication. [1]

The studies do not establish a financial return, a broadly representative effect, or one fixed decoder used by every participant. [1]

The benefit described here, for the beneficiary groups shown.

Depth 3 for a restored ability that participants used at home.

Depth 3 for a restored ability that participants used at home.

The depth describes the reported result. Evidence strength is assessed separately.

AIBD’s five-level depth rubric, reviewed on 25 October 2026.

AIBD, 25 October 2026.

Depth 2 understates the restored practical ability. A separate societal grade would require a demonstrated societal benefit.

Peer-reviewed studies by the developers and clinical team. The outline is retained.

An early feasibility study of an implanted brain–computer interface. [1]

AIBD-0265 describes the Utrecht interface cited as a historical comparison. The studies were not head-to-head.

Neurotechnology / assistive communication. People with ALS and severe upper-limb paralysis.

Restore digital device control for people with severe paralysis without open-brain surgery

The first implants occurred in mid-2019 and early 2020. The initial report appeared on 28 October 2020. [1] [2]

Investigational implantation and supervised training, followed by use at home. [1]

SHA-256: 1bedfa433679…

NCT03834857 [1]

Not assessed

Not assessed

6 revisions. Latest: 25 October 2026.