Next-gen neuroprosthetics

In 1958, engineer Rune Elmqvist, and surgeon, Ake Senning, implanted the first cardiac pacemaker into the chest of Arne Larsson to keep his failing heart beating at a steady rhythm. At first, it wasn’t a great deal. The first device stopped working within hours. A replacement failed within days. However, over his lifetime, Larsson received eleven pacemakers in total and went on to live until the age of 86, outliving both Elmqvist and Senning.

That winding yet successful beginning marked the start of a medical transformation. Today, active implantable devices are capable of much more than controlling the cardiac rhythm and are entering a whole new frontier where implantable devices interface directly with the brain and the nervous system to restore speech, movement, and vision to people who have lost them. The field is called neuroprosthetics and after decades of progress largely confined to the laboratory, it is beginning to deliver on its funding promise.

One recent example comes from San Francisco where a team at the University of California worked with a man unable to produce intelligible speech following a brainstem stroke. His motor cortex had not forgotten how to produce or listen to speech. Every time he attempted to form a word, the corresponding neural signals were intact. They simply had nowhere to go.

An array of electrodes implanted over his motor cortex region responsible for speech, captured the neuronal signals of the brain. Then, a machine learning system trained on the neural patterns emerging from the attempt to form fifty common words, decoded the neural signals in real time. A language model then assembled the decoded words into plausible sentences, estimating what words were likely to follow the one just identified. The result was speech decoded directly from brain activity at up to 15 words per minute, three times faster than his previous assistive technology.

A second participant in the same trial fitted with a denser 253-channel implant reached 78 words per minute with output rendered through a digital avatar which used orofacial movements to convey non-speech communicative gestures.

Another successful example comes from Lausanne, Switzerland, where a team from EPFL and Lausanne University Hospital addressed a different problem: restoring movement following a spinal cord injury.

The spinal circuits that connect to muscles and coordinate walking are, in most injuries, physically intact. They are just cut off from the brain signals that activate them, owing to the damaged spinal cord.

The neuroprosthetic here was a wireless digital bridge, a 64-electrode array implanted over the motor cortex. It records brain activity continuously. In real-time, a processing unit worn in a small backpack decodes the participant’s intention to move and transmits the necessary stimulation commands to a second implant positioned along the spinal cord, beyond the injury zone. Whenever the brain commands a step from the motor cortex, the spinal implant fires the appropriate motor circuits for the leg to move.

These developments have attracted well-deserved attention, but they share an engineering constraint that received far less coverage: the feedthrough problem.

Every active implant must not only fit into tight spaces within the human body, but must also protect its electronics from warm, salty, electrically conductive and corrosive environments. The standard solution is hermetic encapsulation: electronics sealed inside a leak-tight titanium and ceramic casing, with each electrical signal passing through the wall via a conductive feedthrough. And here’s the issue. The number of feedthroughs must equal the number of signals, without exceptions.

For a pacemaker, one feedthrough may suffice. A cochlear implant needs a few dozen. A visual prosthesis or a brain-machine interface need hundreds or even thousands of feedthroughs packed into a casing small enough to implant without causing damage or discomfort. At that density, fabricating reliable implants becomes the critical bottleneck.

Until recently, the whole assembly has been largely handcrafted: connections established one at a time, by skilled technicians working at the limits of human dexterity. This is slow, expensive and difficult to scale.

A paper published in 2026 in the Journal of Biomedical Materials Research Part B, led by Gregg Suaning and colleagues, including the neuroprosthetic engineer and VPH member Anne Vanhoestenberghe (King’s College London) describes the fabrication strategy designed to push past these limits.

Two innovations stand out. The first is a split architecture: instead of a single hermetic unit, the device is divided into two implanted components. A small module near the retina houses the electrode array and custom electronics. A larger unit behind the ear manages wireless power and data transfer. This keeps the most constrained component as small as possible.

The second innovation tackles feedthrough density. Rather than connecting each electrode to its feedthrough individually, all connections are made simultaneously in a single process step built like a printed circuit board, but using alumina ceramic and platinum instead of fiberglass and copper. The result is a feedthrough density of approximately 250 contacts per square centimetre.

The approach was tested in vitro in a 99-channel visual prosthesis called Phoenix99, which was subjected to two weeks of continuous electrical stimulation in saline solution. Electrodes operating under conditions representative of normal clinical use showed no failure or degradation. Critically, the use of materials such as alumina ceramic and platinum allows to sidestep lengthy certification processes as they have a long established regulatory history in implantable devices.

What connects Larsson’s eleven pacemakers to a brain interface that restores vision, speech, or lets a paralysed person walk, is not simply an increase in the number of active channels. It’s a gradual renegotiation of the boundary between electronics and biology, one that has proved more permeable and more consequential than most people anticipated, at least in such a short time span.

The challenges ahead are not trivial. Implants must survive for decades in a hostile environment. Decoders must stay accurate as neural signals may shift over time. Manufacturing must scale from experimental cohorts of a handful of patients to clinical populations in the thousands, or millions.

The transition from scientific investigation to a prescribed treatment is not anymore a question of if, but of when that moment arrives. Since that first failing pacemaker in a Stockholm operating theatre, active implantable technology has come to an extraordinary distance. The next step is making sure it reaches everyone who needs it.

Next-gen neuroprosthetics

Further Readings:

Collaborate

Join us!

Keep in touch

Further Readings:

Related news

2026 CompuCell3D Virtual Tissue Modeling Workshop and Hackaton

Observatory for digital health technologies in Europe

In the literature: May 2026 highlights

Collaborate

Join us!

Keep in touch