
Nearly six years after a diving accident left Keith Thomas, then 42, with complete paralysis below the chest, no movement, no sensation, he can now feed himself, drink from a cup, and feel the pressure of a handshake. The change comes from a “double neural bypass” (DNB), a hybrid brain-computer interface (BCI) neuroprosthesis that simultaneously restores both movement and sensation.
The results, published July 16 in Nature Medicine by a team led by Chad Bouton at the Feinstein Institutes for Medical Research in New York, represent the first time a BCI has produced durable neurological recovery that persisted months after stimulation was turned off.
“The field has been waiting for proof that a bidirectional system could work in a human being,” said Bouton, professor and vice president of advanced engineering at the Feinstein Institutes. “We now have that proof.”
How the double bypass works
The DNB system has two parallel pathways. On the movement side, five microelectrode arrays (224 electrodes total) implanted in Thomas’s brain, two in primary motor cortex (M1) for reading movement intentions, three in primary somatosensory cortex (S1) for delivering tactile feedback, connect to a laptop running a long short-term memory (LSTM) recurrent neural network. The network decodes neural signals into four hand states (rest, open, close, reaching) with 84.6% accuracy, sustained for more than five months without retraining using only 10 of 128 M1 channels.
A deep reinforcement learning agent regulates grasp force in real time, enabling delicate handling. The decoded commands drive transcutaneous spinal cord stimulation patches over the cervical spine, neuromuscular electrical stimulation on the forearm, and a custom 3D-printed active orthosis.
On the sensation side, force sensors at the fingertips and palm detect pressure, and intracortical microstimulation delivers patterned electrical pulses to S1, creating tactile percepts. A novel intervention the team calls “cortical mirroring” replays neural activity patterns recorded from S1 during imagined or physical touch back to the same area, driving neuroplasticity and durable sensory recovery.
Measured outcomes
Over 35 weeks of the trial, Thomas’s right arm strength improved by 86% and left arm strength by 62%. He achieved an 87% success rate grasping hollow, delicate objects like eggshells while holding a conversation. Restored tactile sensitivity in the wrist dropped from a detection threshold of 100 grams to 10 grams on monofilament testing. Sensory gains persisted for more than two months after stimulation stopped.
At the two-year follow-up, gains in arm strength and sensation remained, suggesting long-term neuroplastic changes rather than temporary assistive effects.
N of one, with caveats
The study is a single-participant, open-label feasibility trial (N=1), and experts have cautioned that the results may not generalize to other individuals with different injury levels or causes. Thomas’s particular injury, a complete C4 sensory/C5 motor tetraplegia from a diving accident, represents a specific lesion profile.
“The procedure requires a 15-hour craniotomy with the patient awake for portions to confirm electrode placement,” said Santosh Chandrasekaran, lead author at the Feinstein Institutes. “It is not a simple intervention.”
Thomas was awakened mid-surgery on March 8, 2023, at North Shore University Hospital in Manhasset, New York, to provide real-time feedback confirming that electrical stimulation produced sensations in the correct locations. The surgical team, led by neurosurgeon Ashesh Mehta, used the feedback to adjust electrode positioning before closing.
Independent experts interviewed by STAT News raised questions about practical scalability. The DNB requires months of pre-surgical MRI mapping, a large surgical team, and ongoing AI recalibration. Some of the most impressive functional outcomes, the fine grasping, the steady self-feeding, required connection to the computer system in the laboratory. Gains that persisted outside the lab were more modest.
“These are extraordinary results for one person,” said an expert who spoke to STAT on condition of anonymity. “The question is whether you can do this for 100 people, or 1,000.”
Path forward
The Feinstein team is now recruiting for an expanded trial of up to three participants with tetraplegia. The DNB technology was recognized in TIME Magazine’s Best Inventions Hall of Fame in 2024. The team is also exploring applications in stroke rehabilitation, multiple sclerosis, and traumatic brain injury.
“The brain is plastic,” Bouton said. “We are showing that if you give it the right input and output, it can rewire itself in ways we did not think were possible.”
Sources
Chandrasekaran S, Wandelt SK, Jangam A, et al. “Double neural bypass restores volitional movement and tactile sensation in a person with tetraplegia.” Nature Medicine 32, 2591-2601 (2026). DOI: 10.1038/s41591-026-04498-0
Broderick OR. “Brain implant restores the sensation of touch in a person with quadriplegia.” STAT News, July 16, 2026. https://www.statnews.com/2026/07/16/brain-computer-interface-implant-restores-sensation-double-neural-bypass/

