Over the last couple of decades, many people have regained hearing functionality with the most successful neurotech device to date: the cochlear implant. But for those whose cochlear nerve is too damaged for a standard cochlear implant, a promising alternative is an auditory brainstem implant (ABI). Unfortunately, current ABIs are rigid implants that do not allow for good tissue contact. As a result, doctors commonly switch off a majority of the electrodes due to unwanted side effects such as dizziness or facial twitching—leading most ABI users to perceive only vague sounds, with little speech intelligibility.

Designing a soft implant that truly conforms to the brainstem environment is a critical milestone in restoring hearing for patients who can’t use cochlear implants.

Now, a team at EPFL’s Laboratory for Soft Bioelectronic Interfaces has developed a soft, thin-film ABI. The device uses micrometer-scale platinum electrodes embedded in silicone, forming a pliable array just a fraction of a millimeter thick. This novel approach, published in Nature Biomedical Engineering, enables better tissue contact, potentially preventing off-target nerve activation and reducing side effects.

“Designing a soft implant that truly conforms to the brainstem environment is a critical milestone in restoring hearing for patients who can’t use cochlear implants. Our success in macaques shows real promise for translating this technology to the clinic and delivering richer, more precise hearing,” says Stéphanie P. Lacour, head of Head of the Laboratory for Soft Bioelectronic (LSBI) Interfaces at EPFL.

Probing “prosthetic hearing” with a complex behavioral task
Rather than simply relying on surgical tests, the researchers ran extensive behavioral experiments in macaques with normal hearing. This allowed them to measure how well the animals could distinguish electrical stimulation patterns as they would with natural acoustic hearing.

“Half the challenge is coming up with a viable implant, the other half is teaching an animal to show us, behaviorally, what it actually hears,” says Emilie Revol, co-first author on the project and a former PhD student at EPFL. She meticulously trained the animals to perform an auditory discrimination task: the monkeys learned to press and release a lever to indicate whether consecutive tones were the “same” or “different.”

“We then introduced stimulation from the soft ABI step by step, blending it with normal tones at first so the monkey could bridge the gap between acoustic and prosthetic hearing,” says Revol. “Ultimately, the goal was then to see if the animal could detect small shifts from one electrode pair to another when only stimulating the soft ABI. Our results suggest that the animal treated these pulses almost the same way it treated real sounds.”

Why a soft array?
“Our main idea was to leverage soft, bioelectronic interfaces to improve electrode-tissue match,” explains Alix Trouillet, a former postdoctoral researcher at EPFL and co-first author of the study. “If the array naturally follows the brainstem’s curved anatomy, we can lower stimulation thresholds and maintain more active electrodes for high-resolution hearing.”

Conventional ABIs rest on the dorsal surface of the cochlear nucleus, which has a 3 mm radius and a complex shape. Rigid electrodes leave air gaps, leading to excessive current spread and undesired nerve stimulation. By contrast, the EPFL team’s ultra-thin silicone design easily bends around the tissue.

Beyond conformability, the soft array’s flexible microfabrication means it can be reconfigured for different anatomies. “The design freedom of microlithography is enormous,” says Trouillet. “We can envision higher electrode counts or new layouts that further refine frequency-specific tuning. Our current version houses 11 electrodes—future iterations may substantially increase this number.”

Improved comfort and fewer side effects
A crucial outcome of the macaque study was the absence of noticeable off-target effects. The researchers report that, within the tested range of electrical currents, the animal showed no signs of discomfort or muscle twitches around the face—common complaints from human ABI users. “The monkey pressed the lever to trigger stimulation itself, time and again,” explains Revol. “If the prosthetic input had been unpleasant, it probably would have stopped.”

Path to clinical translation
Although these findings are promising, the path to a commercially available soft ABI will require additional research and regulatory steps. “One immediate possibility is to test the device intraoperatively in human ABI surgeries,” says Lacour, noting that the team’s clinical partners in Boston regularly perform ABI procedures for patients with severe cochlear nerve damage. “They could briefly insert our soft array before the standard implant to measure if we truly reduce stray nerve activation.”

In addition, every material in an implant destined for human use must be fully medical grade and show robust, long-term reliability. Yet the researchers are confident, thanks to the demanding tests the device has already withstood: “Our implant remained in place in the animal for several months, with no measurable electrode migration,” notes Trouillet. “That’s a critical step forward given how standard ABIs often migrate over time.”

Spinal cord injuries are life-altering, often leaving individuals with severe mobility impairments. While rehabilitation robotics—devices that guide movement during therapy—have improved training for those with spinal cord injuries, their effectiveness remains limited. Without active muscle engagement, robotic-assisted movement alone does not sufficiently retrain the nervous system.

A team at .NeuroRestore, led by Grégoire Courtine and Jocelyne Bloch, has now developed a system that seemlessly integrates an implanted spinal cord neuroprosthesis with rehabilitation robotics. The researchers’ device delivers well-timed electrical pulses to stimulate muscles in harmony with robotic movements, resulting in natural and coordinated muscle activity during therapy. The neuroprosthetics innovation leveraged the the robotic expertise of Professor Auke Ijspeert’s lab at EPFL. This advancement not only enhances immediate mobility but also fosters long-term recovery.

“The seamless integration of spinal cord stimulation with rehabilitation or recreational robotics will accelerate the deployment of this therapy into the standard of care and the community of people with spinal cord injury,” says Courtine. This adaptability ensures that rehabilitation professionals can incorporate this technology into existing rehabilitation protocols worldwide. Combining therapies also presents significant challenges, as each requires precise synchronization. Spinal cord stimulation strategies must be modulated in both space and time to match the patient’s movement, and integrating them with widely used robotic rehabilitation systems requires a flexible and adaptable framework.

The technology relies on a fully implanted spinal cord stimulator that delivers biomimetic electrical epidural stimulation (electrical epidural stimulation). Unlike traditional functional electrical stimulation, this method activates motor neurons more efficiently by mimicking natural nerve signals.

The researchers integrated electrical epidural stimulation with various robotic rehabilitation devices—including treadmills, exoskeletons, and stationary bikes—ensuring that stimulation is precisely timed with each phase of movement. The system uses wireless sensors to detect limb motion and automatically adjust stimulation in real time, allowing for a seamless user experience.

In a proof-of-concept study involving five individuals with spinal cord injuries, the combination of robotics and electrical epidural stimulation resulted in immediate and sustained muscle activation. Not only did participants regain the ability to engage muscles during robotic-assisted therapy, but some also improved their voluntary movements even after the stimulation was turned off.

The researchers also worked closely with rehabilitation centers to test how well the stimulation system integrated with widely used robotic devices. “We visited multiple rehabilitation centers to test our stimulation technology with the robotic systems they routinely use, and it was incredibly rewarding to witness their enthusiasm,” say .NeuroRestore researcher Nicolas Hankov and BioRob researcher Miroslav Caban, the study’s first authors. “Seeing firsthand how seamlessly our approach integrates with existing rehabilitation protocols reinforces its potential to transform care for people with spinal cord injury by providing a technological framework that is easy to adopt and deploy across multiple rehabilitation environments.”

The study also showed the potential of this approach beyond clinical settings, as participants used the system to walk with a rollator and cycle outdoors, validating its real-world impact.

This innovative technology offers new hope for individuals with spinal cord injuries, presenting a more effective rehabilitation approach than robotics alone. By making rehabilitation more dynamic and engaging, it has the potential to significantly enhance recovery outcomes. Future clinical trials will be needed to establish long-term benefits, but the initial results suggest that integrating neuroprosthetics with rehabilitation robotics could redefine mobility restoration after paralysis.

List of contributors

• EPFL Neuro X
• Lausanne University Hospital (CHUV) and University of Lausanne (UNIL)
• Defitech Center for Interventional Neurotherapies (.NeuroRestore)
• EPFL Biorobotics Laboratory
• ONWARD Medical
• Bern University of Applied Sciences
• VAMED Management and Service Switzerland AG
• ETH Zurich Sensory-Motor Systems Lab
• University of Zurich Spinal Cord Injury Center
• Hocoma AG
• Medtronic
• Oxford University
• GBY (Go-by-Yourself) SA
• Université de Bordeaux Institut des Maladies Neurodégénératives
• Zurich University of Applied Sciences (ZHAW)
• MyoSwiss AG

Press kit

Electrical stimulation of the spinal cord is a promising strategy for reestablishing walking after spinal cord injury, recent studies show. But for patients suffering from muscle spasms, the stimulation protocols have a limited effect due to the unpredictable behaviour of involuntary muscle stiffness related to spasticity. Muscle spasticity affects almost 70% of spinal cord injured patients

Now, scientists at EPFL, Università San Raffaele and Scuola Sant’Anna have found a promising way to address and reduce muscle spasticity in patients with incomplete spinal cord injury. It involves zapping the spinal cord with high-frequency electrical stimulation that blocks the abnormal muscular contractions. This high-frequency treatment gives patients suffering from spasticity access to rehabilitation protocols that were previously inaccessible to them with a very good clinical outcome. The results are published today in Science Translational Medicine.

“We’ve found that high frequency electrical stimulation of the spinal cord, coupled with the usual continuous, low-frequency spinal stimulation, is effective during rehabilitation after spinal cord injury, overcoming muscular stiffness and spasms in paralyzed patients and effectively assisting the patients during locomotion,” explains Silvestro Micera, professor at EPFL’s Neuro X Institute and Scuola Sant’Anna.

“This is a safe and effective surgical procedure that offers a new perspective in the treatment of patients with severe damage to the spinal cord. We are planning to extend the indications to different clinical conditions we will define in the next month. We are deeply grateful to the patients who trusted us,” says Pietro Mortini, Head of the Neurosurgery and Stereotactic Radiosurgery Unit at IRCCS Ospedale San Raffaele (Milan) and full professor of Neurosurgery at the University Vita-Salute San Raffaele.

Electrical stimulation of the spinal cord is an indirect way to reach the motor neurons that make muscles move. That’s because the backside of the spinal cord contains sensory neurons which in turn communicate with the motor neurons. In muscle spasticity, it is known that the spinal sensory-motor circuits are overreactive. In fact, the spinal cord is naturally overreactive to stimuli, which is good since it leads to fast reflexes. Normally, that over-reactivity is balanced out by the brain that inhibits the motor circuits. In spinal cord injury, the patient loses messaging from the brain and these inhibitory mechanisms. By indirectly stimulating the motor circuits, the research team has found that high-frequency stimulation of the spinal cord is an artificial and safe way to inhibit that over-reactivity without producing discomfort in patients.

During the clinical trial at San Raffaele Hospital, coordinated by Mortini and Micera, Simone Romeni, first author of the study and researcher at EPFL and Università San Raffaele, proposed to implement high-frequency stimulation taking inspiration from previous work on high-frequency kilohertz blocks of motor circuits by stimulating peripheral nerves.

“At this stage, we can only speculate that high-frequency stimulation acts as a kilohertz block that prevents muscle spasticity," says Micera.

"The clinical data with the two patients point to the benefits of implementing high-frequency stimulation for reducing muscle stiffness and spasms in paralysis. More experiments will be necessary to confirm the potentials of this approach,” concludes Mortini.

Researchers at EPFL and Lausanne University Hospital (CHUV), led by professors Grégoire Courtine and Jocelyne Bloch, have achieved a major milestone in the treatment of spinal cord injuries (SCI). By applying deep brain stimulation (DBS) to an unexpected region in the brain—the lateral hypothalamus (LH)—the team has improved the recovery of lower limb movements in two individuals with partial SCI, greatly improving their autonomy and well-being.

Last year on vacation, it was no problem to walk a couple of steps down and back to the sea using the stimulation.

Wolfgang Jäger, a 54-year-old from Kappel, Austria, has been in a wheelchair since 2006 after a ski accident left him with a spinal cord injury. Participating in the clinical trial, he experienced firsthand how deep brain stimulation could restore his mobility and independence. “Last year on vacation, it was no problem to walk a couple of steps down and back to the sea using the stimulation,” Jäger shared, describing the newfound freedom DBS has given him. Beyond walking, the therapy has improved everyday tasks. “I can also reach things in my cupboards in the kitchen,” he added.

DBS is a well-established neurosurgical technique that involves implanting electrodes into specific brain regions to modulate neural activity. Traditionally, DBS has been used to treat movement disorders like Parkinson’s disease and essential tremor by targeting areas of the brain responsible for motor control. However, applying DBS to the lateral hypothalamus to treat partial paralysis is a novel approach. By focusing on the LH, the researchers at .Neurorestore tapped into an unexpected neural pathway that had not been considered before for motor recovery.

In the study published in Nature Medicine, not only did the DBS show immediate results to augment walking during rehabilitation, but patients also showed long-term improvement that persisted even when the stimulation was turned off. These findings suggest that the treatment promoted a reorganization of residual nerve fibers that contribute to sustained neurological improvements.

This research demonstrates that the brain is needed to recover from paralysis.

“This research demonstrates that the brain is needed to recover from paralysis. Surprisingly, the brain is not able to take full advantage of the neuronal projections that survive after a spinal cord injury. Here, we found how to tap into a small region of the brain that was not known to be involved in the production of walking in order to engage these residual connections and augment neurological recovery in people with spinal cord injury,” says Courtine, professor of neuroscience at EPFL, Lausanne University Hospital (CHUV) and UNIL and co-director of the .NeuroRestore center..

Fundamental neuroscience combined with neurosurgical precision

The success of this DBS therapy hinged on two complementary approaches: discoveries enabled by novel methodologies in animal studies and the translation of these discoveries into precise surgical techniques in humans. For the surgery, the researchers used detailed brain scans to guide the precise locations of the small electrodes into the brain, performed by Bloch at CHUV, while the patient was fully awake.

At this moment, I knew that we were witnessing an important discovery for the anatomical organization of brain functions.

“Once the electrode was in place and we performed the stimulation, the first patient immediately said, ‘I feel my legs.’ When we increased the stimulation, she said, ‘I feel the urge to walk!’ This real-time feedback confirmed we had targeted the correct region, even if this region had never been associated with the control of the legs in humans. At this moment, I knew that we were witnessing an important discovery for the anatomical organization of brain functions,” says Bloch, neurosurgeon and professor at the Lausanne University Hospital (CHUV), UNIL and EPFL, and co-director of the .NeuroRestore centre. .

The lateral hypothalamus’ role in walking recovery

The identification of the LH as a key player in motor recovery after paralysis is in itself an important scientific discovery, given that this region has traditionally only been associated with functions like arousal and feeding. This breakthrough emerged from the development of a novel multi-step methodology that began with whole-brain anatomical and functional mapping to establish the role of this region in walking, followed by experiments in preclinical models to establish the precise circuits involved in the recovery. Ultimately, these results led to clinical trials in human participants.

Without this foundational work, we would not have uncovered the unexpected role this region plays in walking recovery.

“It was fundamental research, through the creation of detailed brain-wide maps, that allowed us to identify the lateral hypothalamus in the recovery of walking. Without this foundational work, we would not have uncovered the unexpected role this region plays in walking recovery," says Jordan Squair, a lead author of the study.

The advanced imaging platform at the Wyss Center played a critical role in this research by providing high-resolution imaging capabilities that enabled the team to map the anatomical and functional activity of neurons across the brain, enabling the identification of the lateral hypothalamus.

Combining DBS with spinal implants for enhanced recovery

These remarkable results pave the way for new therapeutic applications to augment recovery from SCI. Future research will explore integrating DBS with other technologies, such as spinal implants that have already shown their potential in restoring movement after SCI. “Integrating our two approaches—brain and spinal stimulation—will offer a more comprehensive recovery strategy for patients with spinal cord injuries,” says Courtine.

As we age, our cognitive and motor functions deteriorate, which in turn affects our independence and overall quality of life. Research efforts to ameliorate or even completely abolish this have given rise to technologies that show a lot of promise.

Among these is non-invasive brain stimulation: a term encompassing a set of techniques that can affect brain functions externally and noninvasively, without the need for surgery or implants. One such promising technique, in particular, is anodal transcranial direct current stimulation (atDCS), which uses a constant, low electrical current delivered via electrodes on the scalp to modulate neuronal activity.

However, studies exploring atDCS have produced inconsistent results, which has prompted researchers to explore why some people benefit from atDCS while others don’t. The problem seems to lie in our understanding of factors that may influence responsiveness to brain stimulation, leading to responders and non-responders; among these, age has been suggested as one important factor.

Some studies suggest further factors such as baseline behavioral abilities and previous training might be important considerations, but an interplay of these factors with behavior has not been determined in detail, pointing to the need of refined predictive models of the effects of atDCS.

Now, scientists led by Friedhelm Hummel at EPFL have identified an important factor affecting an individual’s responsiveness to atDCS. The team looked at how native learning abilities determine the effect of brain stimulation applied while learning a motor task. Their findings suggests that individuals with less efficient learning mechanisms benefit more from stimulation, while those with optimal learning strategies might experience negative effects.

The researchers recruited 40 participants: 20 middle-aged adults (50-65 years old) and 20 older adults (over 65). Each group was further divided into those receiving active atDCS and those receiving placebo stimulation.

Over ten days, participants practiced a finger-tapping task designed to study motor sequence learning at home while receiving atDCS. The task involved replicating a numerical sequence using a keypad, trying to be as fast and as accurate as possible.

The team then used a machine-learning model trained on a public dataset to classify participants as either “optimal” or “suboptimal” learners, based on their initial performance. This model aimed to predict who would benefit from atDCS, based on their ability to integrate information about the task efficiently early during training

The study found that suboptimal learners, who were seemingly less efficient at internalizing the task at the early stages of learning, experienced an accelerated accuracy improvement while performing the task when receiving atDCS. This effect was not limited to people of a certain age (e.g., older adults), with suboptimal learners being found among younger individuals as well.

In contrast, participants with optimal learning strategies, regardless of age, even showed a negative trend in performance when receiving atDCS. This difference suggests that brain stimulation is more beneficial for individuals who initially struggle with motor tasks. As such, atDCS seems to possess a restorative rather than an enhancing quality, with important implications for rehabilitation.

“By leveraging different methods in Machine learning, we were able to untangle the influence of different factors on the individual effects of brain stimulation,” says Pablo Maceira, the study’s first author. “This will pave the way to maximize the effects of brain stimulation in individual subjects and patients.”

The study implies that, in the long run, personalized brain stimulation protocols will be developed to maximize benefits based on an individual’s specific needs, rather than a common trait such as age. This approach could lead to more effective brain stimulation-based interventions, targeting specific mechanisms supporting learning, especially in the view of neurorehabilitation, for which the main basis is the re-learning of lost skills due to a brain lesion (e.g., after a stroke or a traumatic brain injury).

“In the future, clinicians could apply a more advanced version of our algorithm to determine whether a patient will benefit from a brain stimulation-based therapy, to enhance the effects of neurorehabilitation and personalize treatment,” says Hummel.

Other contributors

• University of Geneva
• University of Applied Sciences Western Switzerland (HES-SO)
• NIH National Institute of Neurological Disorders and Stroke

As we age, it becomes more difficult to remember where things are—whether it’s recalling where we left the keys or where we parked the car. This spatial memory deteriorates further with the onset of dementia, a condition that someone in the world develops every three seconds, according to Alzheimer’s Disease International.

We are addressing a serious concern for those affected by dementia.

Researchers at two EPFL labs have joined forces to give a boost to spatial memory by creating a unique experimental setup that combines non-invasive deep-brain stimulation, virtual reality training, and fMRI imaging—all housed within Campus Biotech in Geneva. Published in Science Advances, the study demonstrates that targeted, painless electric impusles to the hippocampus and adjacent structures, a deep brain region implied in memory and spatial navigation, can improve the brain’s ability to recall locations and navigate more effectively.

“By finding ways to improve spatial memory without surgery or medication, we are addressing a serious concern for a large and growing population: the elderly, as well as brain trauma patients and those affected by dementia,” says Friedhelm Hummel, head of the Hummel Lab.

The study is the result of a collaboration between the Hummel Lab and Olaf Blanke’s Laboratory of Cognitive Neuroscience (LCNO), both at EPFL’s Neuro X institute. By combining Hummel’s expertise in non-invasive brain stimulation with Blanke’s cognitive research of spatial navigation in virtual reality environments, the researchers developed a unique neuro-technological setup.

A one-of-a-kind combination of neuro-technologies
The experiment begins with researchers placing four harmless electrodes on the heads of healthy individuals to stimulate the hippocampus and adjacent structures. This non-invasive technique, called transcranial temporal interference electric stimulation (tTIS), sends targeted pulses without causing any discomfort for the participant.

This leads us to believe that by stimulating the hippocampus, we temporarily increased brain plasticity.

Next, volunteers are immersed in a virtual world using VR goggles. Building on previous research by co-first-author Hyuk-June Moon, the scientists task the participants with navigating through a series of locations and remembering key landmarks. This immersive virtual setting allows researchers to precisely measure how well participants can recall and navigate spatial information whilst receiving tTIS.

“When stimulation was applied, we observed a clear improvement of the participants’ recall time—the time it took to start moving toward where they remembered the object to be,” says Elena Beanato, the other first author of the study. “This leads us to believe that by stimulating the hippocampus, we temporarily increased brain plasticity, which, when combined with training in a virtual environment, leads to better spatial navigation.”

The entire experiment was conducted within an fMRI scanner. This provided researchers with real-time images of brain activity, allowing them to monitor how the hippocampus and surrounding regions responded to tTIS during the spatial navigation tasks. The fMRI data revealed changes in neural activity associated with the observed behavioral changes, specifically in the regions responsible for memory and navigation, giving the researchers deeper insight into how non-invasive stimulation modulates brain function.

This integration of advanced technologies at EPFL's Neuro X Institute, makes Campus Biotech one of the few places where all three experimental techniques can be combined in a single study.

“In the long term, we envision using this approach to develop targeted therapies for patients suffering from cognitive impairments.

“The alliance of tTIS, virtual reality, and fMRI offers a highly controlled and innovative approach to studying the brain’s response to stimulation and its impact on cognitive functions,” adds Olaf Blanke. “In the long term, we envision using this approach to develop targeted therapies for patients suffering from cognitive impairments, offering a non-invasive way to enhance memory and spatial abilities.”

Neural spheroids — 3D clusters of brain cells — are emerging as essential tools for understanding neural networks and studying neurological diseases in the lab. EPFL’s e-Flower, a flower-shaped 3D microelectrode array (MEA), allows researchers to monitor the electrical activity of these spheroids in a way that was previously impossible. This breakthrough, published in Science Advances, lays the groundwork for more sophisticated research on brain organoids, which are complex, miniaturized models of brain tissues.

Our flexible technology makes it possible to get accurate recordings without damaging the 3D neural models.

“The e-Flower lets us record neural activity from much more of the surface of neural spheroids in real-time — something that wasn’t possible with earlier tools. Our flexible technology makes it possible to get accurate recordings without damaging the 3D neural models, giving us a better understanding of how their complex circuits work,” says Stéphanie P. Lacour, lead author of the paper and head of the Laboratory for Soft Bioelectronic Interfaces (LSBI) at the Neuro X Institute.

Why neural spheroids?

“We focused on neural spheroids for this study because they provide a straightforward and accessible model,” says Eleonora Martinelli, one of the lead researchers on the project.

Neural spheroids are three-dimensional clusters of neurons that replicate some of the key functions of brain tissue. They are simpler than organoids, which contain multiple cell types and more closely mimic the brain. The LSBI team at Campus Biotech worked in collaboration with Luc Stoppini and Adrien Roux at the Tissue Engineering Laboratory (HEPIA-HESGE), researchers with long-standing experience with neural spheroids electrophysiology.

“Spheroids are relatively easy to produce and manipulate in the lab, which makes them ideal for early-stage testing,” Martinelli continues. “However, our goal is to eventually apply the e-Flower to brain organoids, which more accurately model brain development and disorders.”

“Organoids represent an exciting interface both for neuroscience research and next-generation neurotechnology,” says Stéphanie Lacour. “They bridge the gap between simplified in vitro models and the complexities of the human brain. Our work with the e-Flower is a critical step toward being able to explore these 3D models.”

The serendipity behind the innovation

What was originally a problem for one project became the solution for another.

Interestingly, the e-Flower was born out of an unexpected discovery. Outman Akouissi, a collaborator on the project, encountered a challenge while working on soft implants for peripheral nerves: the hydrogels he used caused his devices to curl unpredictably when exposed to water. What started as a frustration turned into a breakthrough when Akouissi and Martinelli realized this curling mechanism could be harnessed for a completely different application — wrapping around neural spheroids.

“This was a perfect example of how serendipity can lead to innovation,” says Martinelli. “What was originally a problem for one project became the solution for another.”

A new approach to neural electrophysiology

The device consists of four flexible petals equipped with platinum electrodes, which curl around the spheroid when exposed to the liquid that supports the cell structure. This actuation is driven by the swelling of a soft hydrogel, making the device both gentle on the tissue and easy to use.

Designed to be compatible with existing electrophysiological systems, the e-Flower offers a plug-and-play solution for researchers, avoiding the need for complex external actuators or harmful solvents.

Once the technology is applied to organoids, the ability to record electrical activity from all sides will provide a much more comprehensive understanding of brain processes. Researchers hope this will lead to new insights into neurodevelopment, brain injury recovery, and neurological diseases.

In neuroscience and biomedical engineering, accurately modeling the complex movements of the human hand has long been a significant challenge. Current models often struggle to capture the intricate interplay between the brain's motor commands and the physical actions of muscles and tendons. This gap not only hinders scientific progress but also limits the development of effective neuroprosthetics aimed at restoring hand function for those with limb loss or paralysis.

EPFL professor Alexander Mathis and his team have developed an AI-driven approach that significantly advances our understanding of these complex motor functions. The team used a creative machine learning strategy that combined curriculum-based reinforcement learning with detailed biomechanical simulations.

We’re diving deep into the core principles of human motor control.

Mathis's research presents a detailed, dynamic, and anatomically accurate model of hand movement that takes direct inspiration from the way humans learn intricate motor skills. This research not only won the MyoChallenge at the NeurIPS conference in 2022, but the results have also been published in the journal Neuron.

Virtually controlling Baoding balls
“What excites me most about this research is that we’re diving deep into the core principles of human motor control—something that’s been a mystery for so long. We’re not just building models; we’re uncovering the fundamental mechanics of how the brain and muscles work together” says Mathis.

The NeurIPS challenge by Meta motivated the EPFL team to find a new approach to a technique in AI known as reinforcement learning. The task was to build an AI that precisely manipulate two Baoding balls—each controlled by 39 muscles in a highly coordinated manner. This seemingly simple task is extraordinarily difficult to replicate virtually, given the complex dynamics of hand movements, including muscle synchronization and balance maintenance.

In this highly competitive environment, three graduate students—Alberto Chiappa from Alexander Mathis’ group, Pablo Tano and Nisheet Patel from Alexandre Pouget’s group at the University of Geneva—outperformed their rivals by a significant margin. Their AI model achieved a 100% success rate in the first phase of the competition, surpassing the closest competitor. Even in the more challenging second phase, their model showed its strength in ever more difficult situations and maintained a commanding lead to win the competition.

Breaking the tasks down in smaller parts – and repeat them
“To win, we took inspiration from how humans learn sophisticated skills in a process known as part-to-whole training in sports science,” says Mathis. This part-to-whole approach inspired the curriculum learning method used in the AI model, where the complex task of controlling hand movements was broken down into smaller, manageable parts.

“To overcome the limitations of current machine learning models, we applied a method called curriculum learning. After 32 stages and nearly 400 hours of training, we successfully trained a neural network to accurately control a realistic model of the human hand,” says Alberto Chiappa.

A key reason for the model’s success is its ability to recognize and use basic, repeatable movement patterns, known as motor primitives. In an exciting scientific twist, this approach to learning behavior could inform neuroscience about the brain’s role is in determining how motor primitives are learned to master new tasks. This intricate interplay between the brain and muscle manipulation points to how challenging it can be to build machines and prosthetics that truly mimic human movement.

“You need a large degree of movement and a model that resembles a human brain to accomplish a variety of everyday tasks. Even if each task can be broken down into smaller parts, each task needs a different set of these motor primitives to be done well,” says Mathis.

Harness AI in the exploration and understanding of biological systems

This research gives us a solid scientific foundation that reinforces our strategy.

Silvestro Micera, a leading researcher in neuroprosthetics at EPFL’s Neuro X Institute and collaborator with Mathis, highlights the critical importance of this research for understanding the future potential and the current limits of even the most advanced prosthetics. “What we really miss right now is a deeper understanding of how finger movement and grasping motor control are achieved. This work goes exactly in this very important direction,” Micera notes. “We know how important it is to connect the prosthesis to the nervous system, and this research gives us a solid scientific foundation that reinforces our strategy.”

The brain and spinal cord are the central pillars of the human central nervous system (CNS), orchestrating everything from movement to sensation. Despite significant advances in neuroscience, our understanding of how these two crucial components of the CNS interact remains limited. A comprehensive view of the entire CNS will bolster the development of therapies for neurological disorders, particularly those involving both the brain and spinal cord, such as spinal cord injuries and neurodegenerative diseases.

This research allows us to see the CNS as a whole, which is crucial if we want to develop effective therapies that target complex neurological conditions.

Bridging this critical knowledge gap, researchers at EPFL’s Neuro X Institute, UNIGE’s Department of Radiology and Medical Informatics, and McGill University’s Montreal Neurological Institute (The Neuro) have developed a tool to significantly advance our understanding of the CNS organization. Their new study, published in Imaging Neuroscience, successfully maps the functional connectivity between the brain and spinal cord in humans, providing a more detailed view of how these two critical components interact.

"This research allows us to see the CNS as a whole, which is crucial if we want to develop effective therapies that target complex neurological conditions. The imaging tool—a set of technologies and protocols—greatly improves our understanding of somatotopic organization. This means that we can now see how different parts of the body are mapped onto specific areas of the brain and spinal cord,” says Dimitri Van De Ville, head of EPFL’s Medical Image Processing Laboratory (MIP:lab) and one of the lead authors of the study.

The CNS network down to the smallest detail
This breakthrough, achieved through a detailed analysis of simultaneous brain and cervical spinal cord functional magnetic resonance imaging (fMRI) data, provides unprecedented insights into how the brain and spinal cord work together to represent the body as a whole. What emerges is a body map that extends across the different levels of the CNS.

Building on prior research in functional imaging techniques developed at the MIP:Lab, researchers used algorithmic models to isolate and identify specific regions of interest in both the brain and spinal cord. For the spinal cord, this approach allowed to accurately delineate the spinal segments responsible for innervating different body parts, enhancing our understanding of the organization within the spinal cord itself.

To explore how the brain and spinal cord interact, the researchers employed functional connectivity analyses. By examining synchronized activity between the brain and spinal cord, they were able to uncover how specific areas in the brain and spinal cord correspond to different parts of the body. This approach provided a detailed map of the networks that connect the brain and spinal cord, offering new insights into how the central nervous system operates as a unified whole.

Our findings show that even at rest, the brain and spinal cord maintain a precise map of the body, this knowledge could contribute to the development of more accurate treatments.

A methodological framework for future progres
“Our findings show that even at rest, the brain and spinal cord maintain a precise map of the body,” says Nawal Kinany, co-first author of the paper. “The methodological framework introduced in this study can also be deployed in clinical populations, for instance to investigate CNS reorganization after lesions or limb amputation. In the future, this knowledge could contribute to the development of more accurate treatments and interventions, ultimately improving patient outcomes.”

Another key innovation lies in the researchers' ability to overcome the technical challenges of imaging two anatomically distinct parts of the central nervous system at the same time, which traditionally require separate scans. The data for the study were collected at the The Neuro in Montreal, under the supervision of Julien Doyon, a leading researcher in the field and a key collaborator on the project. His expertise in neuroimaging and the lab’s state-of-the-art facilities were crucial in capturing the high-quality fMRI data that made this detailed mapping of brain-spinal cord connectivity possible.

By integrating cutting-edge imaging techniques and collaborative research, the findings provide a foundational framework for future explorations into CNS function and dysfunction.

Thanks to advances in medical imaging, doctors can localize a bone fracture, detect a tumor and observe a baby inside the uterus, all in a completely noninvasive manner. There’s no telling just how far we’ll be able to see inside the human body one day. The technology is developing at a rapid pace, generating images with ever-higher resolution that can be used to spot ever-smaller anomalies.

In the area of magnetic resonance imaging (MRI), Prof. Dimitri Van De Ville, the head of EPFL’s Medical Image Processing Lab, has identified two opposing trends.

“The first trend is an increase in the strength of machines’ magnetic fields, enabling them to reveal tiny irregularities such as microscopic injuries and very early-stage cancer cells,” say Van de Ville.

Most MRI machines in hospitals today have a magnetic field of 1.5 or 3 teslas s. Engineers at the Alternative Energies & Atomic Energy Commission near Paris have invented a machine with a magnetic field of 11.7 teslas – the most powerful in the world. According to Prof. Jean-Philippe Thiran, the head of EPFL’s Signal Processing Laboratory, “the stronger the magnetic field, the better we can pick up weak signals that are otherwise hard to catch, giving us more granular information.”

At EPFL, engineers have developed a 7-tesla machine – powerful enough to map human brains by neural layer in vivo. Prof. Friedhelm Hummel, the holder of the Defitech Chair of Clinical Neuroengineering, explains: “This will give us a better understanding of human brain structures, because for now the exact role of each structure isn’t really clear.”

The second trend that Van De Ville identified goes in the opposite direction: the development of machines that have a magnetic field of well below 1.5 teslas, yet that can still generate images of good enough quality to make sound diagnoses. The goal is to create low-cost devices that are easy to transport and install, which can be especially useful in developing countries. “This will be possible thanks to breakthroughs in imaging sensors, devices and data processing – some of which are being made right here at EPFL,” says Van De Ville.

The stronger the magnetic field, the better we can pick up weak signals that are otherwise hard to catch, giving us more granular information.

Ultrasound making a comeback

Another imaging technology – ultrasound – has changed very little since it was first invented. “Ultrasound is used to observe a patient’s heartbeat or a baby moving inside the womb, for example,” says Thiran, who specializes in this technology.

Scientists have been taking a fresh look at ultrasound’s potential in recent years, as it can be coupled with systems for performing real-time calculations. “The latest machines are equipped with extremely powerful calculators that can process huge amounts of data in real time,” says Thiran. “For instance, we can now measure a tissue’s physical properties such as its elasticity. That will be useful for detecting cirrhosis and other liver diseases.”

The powerful calculators will also enable ultrasound machines to run a lot faster. Today they can generate 30 to 40 images per second, but in the not-too-distant future their output will rise to 1,000 to 2,000 images per second. “That will let doctors observe dynamic processes such as blood flow, including in the brain,” says Thiran.

The artificial intelligence revolution

Artificial intelligence, including machine learning, data processing and algorithms, will be a key component of tomorrow’s medical imaging systems. “AI is revolutionizing the field of medical imaging because it lets doctors compile information from different types of patient examinations,” says Van De Ville. “Soon they’ll be able to combine the results of an MRI with those of an X-ray or even a patient’s medical records in order to obtain a comprehensive view of a disease or the functioning of a specific organ.”

Van De Ville, whose research involves human brain modeling, reckons that one day doctors will be able to establish forecasts by asking questions to an interactive program. “AI can already be used to classify images and spot anomalies, but the technology will go further and become more powerful,” he says. Thiran agrees: “You’ll soon hear people talking about computational or calculative medical imaging. The goal of all these advancements is to gain a better understanding of human organs and identify diseases more effectively.”

However, Thiran notes that AI-enhanced imaging does have its limitations. “We need to use high-quality models to train AI programs so that the images and forecasts they produce are accurate. Otherwise, they’ll give you hallucinations.” The programs will therefore need to be fed vast amounts of data and be driven by robust algorithms.

Hummel, for his part, points to the ethical questions surrounding AI. “Suppose this kind of medical imaging predicts that someone has a fair probability of developing Alzheimer’s, for example, years before the disease presents clinically. Should that person be told? And if so, how? And what if doctors aren’t 100% sure of the prediction, and if there’s still no treatment for the disease at that time?”

The latest advances in this area, like all forms of technological progress, should be accompanied by a consideration of the associated ethical issues – especially since medical imaging provides insight into the most intimate aspects of who we are.