Can brain-computer interfaces restore movement in paralysis?

How do brain-computer interfaces actually restore movement?

Brain-computer interfaces work by recording neural signals from the brain, decoding the person's intended movement, and then using that decoded signal to control an external device or stimulate muscles. For example, the brain-spine interface (BSI) described in a 2023 Nature paper uses electrodes implanted in the motor cortex to detect neural activity related to walking, then wirelessly transmits those signals to a spinal cord stimulator that activates leg muscles in a coordinated pattern [2]. This digital bridge bypasses the damaged spinal cord, allowing a person with tetraplegia to stand, walk, and even climb stairs with natural control.

Another approach uses BCIs to control robotic limbs or exoskeletons. In a 2021 study, a person with tetraplegia used a bidirectional BCI that not only decoded motor commands but also provided tactile sensations via microstimulation of the somatosensory cortex. This reduced the time to complete a grasping task by half, from a median of 20.9 seconds to 10.2 seconds, because the user could feel when they had a grip on an object [5]. This shows that adding sensory feedback dramatically improves real-world performance.

For people with less severe injuries, BCIs can also trigger recovery through neurorehabilitation. A 2022 review notes that repeated use of BCI-controlled exoskeletons over several weeks can lead to motor recovery even in chronic paralysis, likely by retraining neural pathways [4]. The 2023 brain-spine interface study also found that neurorehabilitation supported by the BSI improved neurological recovery—the participant regained the ability to walk with crutches overground even when the BSI was switched off [2].

What does the evidence from human trials actually show?

The strongest evidence comes from recent human trials with implanted devices. The 2023 SWITCH study implanted an endovascular BCI (Stentrode) in four people with severe upper-limb paralysis due to ALS or primary lateral sclerosis. Over 12 months, there were no serious adverse events, no blood vessel occlusion, and no device migration. All four patients successfully controlled a computer by thought, decoding at least five attempted movement types [1]. This demonstrates that a minimally invasive BCI—delivered through a blood vessel rather than open-brain surgery—is safe and feasible.

The brain-spine interface trial, also published in 2023, involved a single participant with chronic tetraplegia. The BSI was calibrated within minutes and remained stable over one year, including during independent use at home. The participant could stand, walk, climb stairs, and traverse complex terrains. Importantly, after neurorehabilitation with the BSI, he regained the ability to walk with crutches even when the device was off, indicating lasting neurological recovery [2].

For upper-limb control, the BrainGate system has enabled people with tetraplegia to control a computer cursor, robotic arm, and even their own paralyzed limb via functional electrical stimulation. A 2023 review notes that participants in the BrainGate2 trial have achieved highly accurate point-and-click cursor control and intended handwriting [11]. A 2021 study showed that a wireless version of the BrainGate system allowed two participants to control a tablet computer at home over a 24-hour period, with communication bitrates equivalent to the wired version [9].

However, not all BCIs are equal. A 2019 study comparing Cartesian (endpoint) versus joint-based velocity commands found that both participants achieved significantly higher success rates using Cartesian control, and joint-controlled trajectories were more variable and curved [8]. This suggests that the optimal decoding strategy depends on the task and the user.

What are the limitations and what can't BCIs do yet?

Despite impressive results, BCIs are not a cure for paralysis. Most studies involve small numbers of participants—the SWITCH study had only four people [1], and the brain-spine interface trial was a single case [2]. Larger trials are needed to confirm safety and efficacy across diverse patient populations. Additionally, not all types of paralysis are equally treatable. A 2023 review on endovascular BCIs for post-stroke paralysis notes that limitations arise when the motor cortex itself is damaged, as in intracerebral hemorrhage, which can make decoding neural signals difficult [10].

The technology also requires surgical implantation, which carries risks. While the endovascular approach is less invasive than craniotomy, it still requires catheter delivery into a brain blood vessel. The SWITCH study reported no serious adverse events, but the sample was small [1]. Deep brain stimulation for walking recovery, tested in two participants with incomplete spinal cord injury, showed immediate improvements but requires further trials to establish safety regarding weight changes, hormonal profiles, and psychological effects [6].

For non-invasive BCIs using EEG, performance is lower. A 2024 study using deep learning to classify attempted arm and hand movements from EEG achieved a mean accuracy of 75.75% across four movement classes, which is promising but not yet reliable enough for critical daily tasks [3]. Another 2023 study found that a well-performing offline model does not necessarily translate to a well-performing online system, highlighting the gap between lab and real-world use [7].

Finally, BCIs currently restore specific functions—like walking or computer control—but not full, natural limb movement. The brain-spine interface allowed walking, but the participant still used crutches [2]. The bidirectional BCI improved grasping speed, but the robotic arm was still external [5]. Restoring fine motor control and proprioception remains a challenge.

Sources used in this answer

Related research questions