AI Knee Torque Sensor for Real-Time Joint Health

Knee injuries are among the most common—and most disruptive—orthopaedic problems, affecting everyone from athletes to the elderly. Despite advances in surgical techniques and rehabilitation therapies, monitoring stress on the knee joint remains a major challenge, especially outside of clinical settings. Without real-time insight into joint forces, patients often unknowingly overexert themselves, hindering recovery and risking re-injury.

Now, a team of researchers from the University of Oxford, UCL, and Xi’an Jiaotong University may have found a solution. They’ve developed a soft, self-powered wearable sensor that uses AI to continuously monitor torque on the knee joint in real time—a critical metric for preventing and managing injury.

But what makes knee injuries so difficult to monitor, and how does this new AI-powered device overcome those barriers? More importantly, could this kind of wearable tech reshape the future of joint health?

The Challenges With Knee Injuries

Knee injuries are anything but fun, and that's putting it mildly. Whether you're a weekend athlete or someone who just took an unlucky fall, the aftermath of a knee injury can range from annoying to life-changing. The knee is a marvel of biomechanics, supporting body weight, absorbing shocks, and enabling movement, all while being incredibly vulnerable to strain, impact, and degeneration.

Knee injuries can arise from trauma, such as a car accident or a sports mishap, or from underlying genetic and developmental conditions like patellar tracking disorder or early-onset arthritis. Regardless of the source, the result is often the same: pain, restricted mobility, and a significant impact on quality of life.

Modern surgical techniques have come a long way; ACL reconstructions, meniscus repairs, and joint replacements can restore function and reduce pain. But let's be honest, surgery is not a magic wand. It comes with recovery time, risk of complications, and, in some cases, only partial relief. In many situations, surgery helps stabilise the problem, but doesn't eliminate the need for long-term care or stress management on the joint.

And that's the key point; minimising stress on the knee is critical for both prevention and recovery. Unfortunately, this is easier said than done, as human knees weren't exactly designed with built-in stress meters, and most people don't have an intuitive sense of how much force they're applying during daily activities. Thus, by the time pain appears, the damage may already be done.

This brings us to a major technical hurdle: how do we measure knee stress in a reliable, continuous, and user-friendly way? The knee is a complex joint with a wide range of motion, frequent changes in angle, and plenty of surrounding tissue that complicates sensor placement. Trying to attach sensors that remain stable, accurate, and comfortable on a moving joint is, to put it bluntly, a real pain in the...knee.

Not only is sensor placement a challenge, but compliance is too. No one wants to strap on a clunky, obtrusive device just to walk around the house. And if the system isn't comfortable or easy to use, people won't wear it, no matter how good the data might be.

In short, knee injuries present a dual challenge: they're both physically debilitating and technically difficult to monitor. Solving this problem requires more than just good intentions as it demands clever engineering, patient-centered design, and a deep understanding of human movement.

Researchers Create Wearable Knee Torque Sensor

A team of international researchers has developed a groundbreaking AI-powered wearable that can monitor the torque exerted on the knee joint in real time: a key metric in diagnosing and managing joint-related conditions such as osteoarthritis, rheumatoid arthritis, and sports injuries. The device, which relies on a piezoelectric sensor and machine learning, aims to make continuous, accurate monitoring accessible even in resource-limited settings.

The collaborative effort, bringing together scientists from the University of Oxford, University College London, and Xi'an Jiaotong University, addresses a major gap in joint health monitoring. As their paper, AI-Enabled Piezoelectric Wearable for Joint Torque Monitoring, published in the May 2025 issue of Nano-Micro Letters, explains, current technologies fail to deliver a low-cost, portable, and high-precision solution for dynamic joint torque measurement.

To meet that need, the team designed a flexible, soft, and lightweight sensor powered by piezoelectric materials, specifically, boron nitride nanotubes embedded in a polydimethylsiloxane (PDMS) substrate. The wearable generates its own electricity through mechanical stress, using the wearer's knee movements as a power source.

On the computational side, the sensor system includes an Arduino Nano 33 BLE and an STM32 NUCLEO F401RE development board. These embedded platforms host a lightweight artificial neural network (ANN) that processes complex dynamic motion signals from the knee, extracting accurate torque estimates in real time. The combination of embedded AI with energy-harvesting materials enables continuous monitoring without dependence on bulky batteries or constant connectivity.

"This technology offers a sustainable solution for long-term joint health monitoring. It's particularly suited for resource-constrained environments where traditional healthcare infrastructure and computational resources may be limited."

In keeping with the spirit of open science, all of the project's code and design materials will be made available on GitHub, and the researchers hope this will encourage others to build on their work and deploy similar systems for health monitoring in a wide range of scenarios, from post-surgical rehabilitation to injury prevention in athletics.

Could Such Devices Be Critical For Future Healthcare?

Could Such Devices Be Critical For Future Healthcare? The short answer? Yes, and probably sooner than we think.

Wearable devices like the AI-powered knee sensor may not get flashy headlines or biotech startup valuations, but they're quietly becoming one of the most important tools in modern healthcare. As chronic conditions rise and healthcare systems strain under cost and capacity, these unobtrusive, data-rich devices might be the unsung heroes of the next medical era.

Where the the real strength lies in their design. Flexible, skin-friendly materials, like the boron nitride nanotube-PDMS combo used in the knee torque sensor, allow for sensors that move with the body, rather than fight against it. That comfort and adaptability isn't just nice to have, it's essential for long-term monitoring. If a wearable is annoying or clunky, it ends up in a drawer, but something soft, light, and self-powered? That's wearable in the truest sense of the word.

And then there's the invaluable medical data. For physicians, such data provides a clearer view of patient recovery and a better ability to spot deviations from the norm. For patients, it means seeing a problem coming before it becomes symptomatic, and this flips the healthcare model from reactive to proactive, from "let's treat the injury" to "let's avoid the injury altogether."

Of course, we're not quite there yet, as this particular device is still a prototype, albeit a very promising one. It needs refinement, broader testing, and real-world trials before it lands in clinics or pharmacies. But the direction is clear: the convergence of soft materials, embedded AI, and low-power electronics is creating a class of medical wearables that are smarter, smaller, and more intuitive than anything we've had before.