Feeding himself, scratching his face and petting his dog may appear to be ordinary actions. For Keith Thomas, however, they represent abilities that had been lost after a severe spinal cord injury left him paralyzed from the chest down.
Researchers at Northwell Health’s Feinstein Institutes for Medical Research have now reported that Thomas regained meaningful arm and hand function through an experimental system called a double neural bypass. The technology interprets his intention to move, electrically activates the relevant muscles and spinal pathways, and sends touch information back into his brain.
The results were published in the peer-reviewed journal Nature Medicine on July 16, 2026. The study describes both immediate computer-assisted movement and improvements that continued when parts of the system were switched off, suggesting the therapy may have encouraged lasting changes within Thomas’s nervous system.
A Diving Accident Interrupted the Brain’s Messages
Thomas suffered a serious neck injury after diving into a swimming pool in July 2020. The accident damaged his spinal cord, preventing signals from traveling normally between his brain and much of his body.
He was diagnosed with chronic C4 sensory and C5 motor complete tetraplegia. Although the term “complete” does not necessarily mean that every nerve fiber has been physically severed, it indicates that standard clinical examinations found no preserved sensory or motor function in key areas below the injury.
When Thomas joined the clinical trial in 2021, he could not lift his arms from his wheelchair, bring his hands to his face or feel touch in his hands and wrists. He depended on other people for many tasks that require reaching and grasping.
In March 2023, neurosurgeon Ashesh Mehta and a multidisciplinary Northwell team performed a 15-hour brain operation. Five small microelectrode arrays were implanted in carefully mapped regions associated with movement and sensation in Thomas’s right hand and fingers.
What a Double Neural Bypass Actually Does
A spinal cord injury can be compared with damage to a communication cable. The brain may still produce a clear command to open the hand, but that message cannot reliably travel through the injured area to reach the muscles.
The double neural bypass creates an electronic route around that interruption. It is described as “double” because information moves in both directions. One pathway carries movement commands from the brain toward the body, while another delivers touch-related information from the hand back to the brain.
When Thomas imagines moving, the implanted arrays record patterns of electrical activity in his motor cortex. Artificial-intelligence algorithms interpret those signals and determine the movement he is attempting to perform.
The computer then sends instructions to electrodes placed on his skin. These electrodes stimulate selected muscles in his forearm, causing his own hand to open, close or adjust its grip. The process occurs rapidly enough for the movement to respond to his thoughts in real time.
The system also delivers patterned stimulation to the spinal cord below the damaged area. Rather than merely operating the muscles like an external switch, this stimulation is intended to activate surviving neural circuits and encourage the brain and spinal cord to communicate more effectively.
The Second Bypass Restores a Form of Touch
Movement without sensation has serious limitations. A person may be able to close a hand but remain unable to judge whether an object is slipping or being squeezed too tightly.
Thomas’s system addresses this problem through sensors incorporated into a custom brace around his hand. The sensors measure contact and grasping pressure when he touches an object.
That information is processed by the computer and translated into stimulation delivered through electrodes in the sensory region of his brain. Thomas does not experience the sensation exactly as he did before his injury, but the artificial feedback allows him to detect contact and better control his grip.
During testing, he learned to handle fragile objects, including hollow eggshells. Researchers reported that he could lift the delicate shells without breaking them in almost 90% of attempts. The movement-decoding system recognized his intended hand actions with close to 85% accuracy and remained effective for more than five months without needing continuous retraining.
The restored sensation also allowed Thomas to experience emotionally meaningful contact. He could feel his sister holding his hand and sense the fur of his dog while petting her—experiences that had been absent since the accident.
He Can Now Perform Tasks That Once Required Assistance
After months of training, Thomas became able to bring his hands toward his face, scratch his nose and wipe his eyes independently. With the system operating, he could grasp food, feed himself and lift a cup to drink.
These abilities required more than a single successful demonstration. The research involved repeated sessions in which the algorithms learned to recognize movement intentions while Thomas practiced controlling his arms and hands.
According to the published results, 35 weeks of therapy produced an 86% improvement in right-arm strength and a 62% improvement in the left arm. Spinal stimulation improved elbow flexion sufficiently for Thomas to bring both hands toward his face, even though the implanted brain interface primarily targeted movements associated with his right hand.
The ability to perform two tasks at once was another notable development. Some brain-computer interfaces demand intense mental concentration, making control less reliable when the user is speaking or distracted. Thomas was able to complete delicate grasping tasks while holding a conversation, suggesting that the system was becoming more intuitive.
Cortical Mirroring May Have Helped Rebuild Sensation
The team later introduced a technique called cortical mirroring to strengthen the sensory portion of the therapy.
Researchers recorded Thomas’s brain activity while he imagined being touched on an area where he could still feel sensation. They then reproduced related stimulation patterns in the sensory cortex while simultaneously stimulating his skin and spinal cord.
The purpose was to provide the nervous system with matching signals from several directions, encouraging the brain to associate activity in the affected wrist with the experience of touch.
After approximately 25 weeks of targeted therapy, Thomas regained sensation in part of his right wrist that had remained numb since the accident. Researchers reported that the improvement persisted for more than two months after the stimulation phase ended. Later follow-up suggested that broader motor and sensory gains remained detectable for more than two years.
Lasting Improvement Is More Important Than Temporary Control
Brain-computer interfaces have previously allowed people with paralysis to operate computer cursors, robotic arms or electrically stimulated muscles. Such systems can restore useful control while the equipment is active, but the benefit often ends when the person disconnects from the technology.
The double neural bypass is different because it combines an assistive device with a potential rehabilitation therapy. It helps Thomas move during a session while repeatedly stimulating the brain, spinal cord and muscles in coordinated patterns.
Researchers believe this repeated activity may promote neuroplasticity, the nervous system’s ability to reorganize and strengthen functional connections. The system is not regenerating the damaged spinal cord or erasing the original injury. Instead, it may be training surviving pathways and dormant neural circuits to contribute more effectively.
That interpretation remains a scientific hypothesis requiring further testing, but the persistent improvements provide stronger evidence than a device that works only while connected. The Nature Medicine study describes the approach as a hybrid neuroprosthesis intended to provide both immediate assistance and durable sensorimotor recovery.
This Is Not Yet a Cure for Paralysis
The results are encouraging, but the research has major limitations. The published study centers on one participant, making it impossible to know whether people with different injuries would experience similar improvements.
Thomas also underwent invasive open-brain surgery, extensive neurological mapping and months of closely supervised training. The present system uses implanted electrodes, external computers, skin-mounted stimulators and specialized researchers. It is not a treatment that hospitals can currently provide as routine spinal cord rehabilitation.
Brain surgery carries risks that must be considered against the possible benefit. Long-term implants may also face challenges involving infection, scar tissue, signal stability, hardware maintenance and future replacement.
Independent experts interviewed by STAT welcomed the recovery of touch and lasting function but emphasized that larger trials are necessary before the results can be generalized.
The Research Could Lead to More Practical Treatments
The Feinstein Institutes team plans to test the approach in additional participants with different spinal cord injuries and neurological conditions. Future versions may use smaller computers, wireless connections and more wearable stimulation technology.
The researchers are also investigating less-invasive systems that stimulate the nervous system through the skin. Those versions may not provide the precision of implanted brain electrodes, but they could eventually reach far more people, including some stroke survivors and patients who are not suitable candidates for brain surgery.
For Thomas, the importance of the technology is already visible in everyday life. Feeding himself restores a degree of independence. Wiping his face removes the need to request assistance for a minor discomfort. Feeling his dog’s fur restores a form of personal connection that paralysis had taken away.
The double neural bypass remains experimental, expensive and highly complex. It has not reversed spinal cord injury generally, and one successful case cannot predict how widely it will work. Yet it demonstrates that the damaged nervous system may be more responsive to carefully coordinated brain, spinal and muscle stimulation than previously believed.
