A brain–spine interface restored control of walking after paralysis

STIMO-BSI brain–spine interface [1]

REW-MSLM, a decoder that combines a hidden Markov model with multilinear regressions. [1]

The decoder was calibrated for the participant. The paper names no software release. [1]

NeuroRestore (EPFL/CHUV/UNIL) with CEA-Clinatec and ONWARD Medical

The decoder estimated intended leg movements and adjusted spinal stimulation. [1]

WIMAGINE implants recorded brain activity. A portable computer controlled an ACTIVA RC pulse generator and a Specify 5-6-5 spinal lead. [1]

Cortical signals entered the decoder. Its predictions set the timing and strength of spinal stimulation. [1]

Stimulation followed movement intentions within preset amplitude limits. The participant could start activities and adjust the permitted range. [1]

Researchers calibrated the decoder, followed by 40 rehabilitation sessions and continued use at home. [1]

The participant and experimenter judged calibration before use. Physiotherapists supervised rehabilitation. [1]

The REW-MSLM decoder estimated intended leg movements from cortical signals and adjusted spinal stimulation. Researchers calibrated it for the participant and set the permitted stimulation ranges.

The REW-MSLM decoder estimated intended leg movements from cortical signals and adjusted spinal stimulation. Researchers calibrated it for the participant and set the permitted stimulation ranges. [1]

Not assessed

The participant regained control of standing and walking, including stairs and uneven ground. [1]

Improved Physical Health

Reported benefit. [1]

People.

Gert-Jan Oskam, a participant with an incomplete cervical spinal cord injury. [1] [3]

One participant. [1]

Access required surgery, personalized calibration, and rehabilitation within a clinical study. [1]

Control remained stable for over a year, including seven months of home use. [1]

Yes

An infection required removal of one cortical implant after 167 days. Rehabilitation resumed after treatment, and the implant was replaced in March 2023. [1]

STIMO-BSI, NCT04632290. [1]

Researchers followed one participant and compared walking under different stimulation conditions. Six independent physiotherapists assessed gait videos without knowing the condition. [1]

Researchers recorded movements and muscle activity, performed clinical walking tests, and collected the participant’s reports. [1]

One 38-year-old man at CHUV in Lausanne, ten years after an incomplete C5/C6 spinal cord injury. [1]

Walking distance and speed, balance, assistance required, muscle activity, and decoder accuracy. [1]

Walking with brain-controlled stimulation was compared with stimulation triggered by motion sensors and with no stimulation. [1]

After rehabilitation, the participant walked more than 10 meters with crutches and no stimulation. Brain-controlled stimulation also improved walking across uneven terrain. [1]

Brain-controlled stimulation produced more natural steps than stimulation triggered by motion sensors. The study did not compare different decoders. [1]

The study involved one participant. Surgery, stimulation, and rehabilitation accompanied the decoder, so their individual contributions to recovery were not isolated. [1]

One participant completed 40 rehabilitation sessions over 15 weeks, followed by seven months of home use. [1]

The authors deposited source data and supporting MATLAB scripts in Zenodo. [1] [6]

Not assessed

Not assessed

Not assessed

Not assessed

Not assessed

Not assessed

Not assessed

Not assessed

Not assessed

Not assessed

Not assessed

NeuroRestore and CEA-Clinatec developed the interface. CHUV conducted the clinical study, with Medtronic stimulation hardware. [1] [2]

Benefit: Lausanne, Switzerland. Developer: Switzerland. Operator: Switzerland.

Not assessed

The interface required cortical implants, a portable processor, wireless links, and an implanted spinal stimulator. [1]

The Swiss clinical team worked with CEA-Clinatec in France. Brain mapping took place in Grenoble, while surgery and rehabilitation took place in Lausanne. [1]

Not assessed

The approach requires viable spinal circuits below the injury, suitable brain signals, implanted devices, and personalized rehabilitation. [1]

ONWARD reported further ARC-BCI implantations in 2024 and 2026. Those announcements did not establish recovery in each participant. [4] [5]

By January 2026, ONWARD reported seven ARC-BCI participants. The later system paired WIMAGINE with ARC-IM. [4] [5]

ARC-BCI remained investigational and unavailable for commercial use in ONWARD’s January 2026 announcement. [5]

Not assessed

Not assessed

Not assessed

Not assessed

AIBD-0001-source-0001–0006

Numbered references below.

English. [1] [2] [3] [4] [5] [6]

Developer, Independent

The Nature study provides the clinical evidence. ONWARD’s announcements document later implantations, and Zenodo hosts the data. [1] [2] [3] [4] [5] [6]

The full article and later announcements were accessible. The repository lists source data and MATLAB scripts, which were not executed. [1] [2] [3] [4] [5] [6]

Not assessed

Not assessed

Not assessed

Not assessed

People.

People: 3.

The participant regained movement and used the interface at home over several months. [1]

Walking was observed in clinical tests and during home use. [1]

Not assessed

AIBD, assisted by Codex, on 25 October 2026.

Not assessed

Independently reported

One participant in a clinical feasibility study. [1]

Not assessed

Healthcare / neurorehabilitation. Spinal cord injury rehabilitation.

Researchers connected movement intentions to spinal stimulation to restore control of standing and walking.

Observed: 2021. Updated: 25 October 2026.

Clinical feasibility study. [1]

SHA-256: 8873303244c6…

STIMO-BSI is registered as NCT04632290. The earlier STIMO study is NCT02936453. [1]

Not assessed

Not assessed

7 revisions. Latest: 25 October 2026.