Roughly 80% of people will experience low back pain at some point in their lives, and one of the oldest problems in treating it hasn’t changed in decades: physical therapists only see how a patient moves for the hour they’re in the clinic, not during the other 167 hours of the week. A research team at the University of California San Diego is trying to close that gap with a wearable that looks less like a smartwatch and more like a piece of medical tape — and a set of machine learning models trained to make sense of the signal it produces.
A Fabric Sensor, Not Another Wristband
The project, called MS-ADAPT (Multi-Sensor Adaptive Data Analytics for Physical Therapy), is led by Emilia Farcas, a research scientist at UC San Diego’s Qualcomm Institute, alongside co-principal investigator Ken Loh, a structural engineering professor in the Jacobs School of Engineering, with additional collaborators including physical therapy researcher Sara Gombatto at San Diego State University and machine learning specialists Rose Yu and Arun Kumar. The team’s core hardware is “Motion Tape,” a low-cost, fabric-based strain sensor built with carbon nanotube-infused material that adheres to the skin over the lower back and flexes with it, tracking posture and spinal movement direction throughout ordinary daily activity — paired with a wrist accelerometer and a smartphone app that also monitors whether patients are keeping up with prescribed exercises. The project has run on National Science Foundation funding exceeding $1 million since 2022.
Why Off-the-Shelf Motion Capture Doesn’t Work Here
Clinical-grade motion capture systems — the camera-and-marker rigs used in biomechanics labs — can measure spinal movement with high precision, but they’re expensive, require a lab setting, and are useless for understanding how someone actually bends, twists or sits during a normal day at work or home. Fitbit-style wrist accelerometers, meanwhile, are cheap and wearable but can’t directly capture spinal strain the way a sensor on the back itself can. Motion Tape is designed to split the difference: portable and inexpensive enough for someone to wear for days at a stretch in the real world, while still capturing strain direction and magnitude at the site that matters most for low back pain.
Teaching an Algorithm to Trust Noisy Data
The tradeoff for that portability is signal quality. Because Motion Tape is a newer and less mechanically stable sensor than lab-grade equipment, its datasets tend to be smaller and noisier than researchers would like. In a paper published in the journal Sensors in February 2026, the UC San Diego team introduced what they call the Motion-Tape Augmentation Inference Model, a deep learning pipeline specifically built to classify low back movements from Motion Tape recordings despite that noise and limited training data. A separate 2026 study from the same group examined multi-sensor, multi-movement data to classify low back pain-related movement patterns, and another explored grouping patients into movement-based subgroups using biomechanical and causal feature engineering — work aimed at eventually predicting which patients will respond to which specific physical therapy interventions.
The Case for Skepticism
It’s worth being clear-eyed about where this technology actually stands: it is a federally funded academic research program, not a commercial product with FDA clearance or a defined path to market. The published results so far are proof-of-concept classification studies — demonstrating that a deep learning model can identify movement patterns from Motion Tape data — rather than clinical trials showing that using the system actually improves patient outcomes or reduces pain compared with standard physical therapy. Wearable sensors glued to the skin also raise practical questions about comfort, skin irritation and consistent placement over multi-day wear that lab researchers rarely have to solve, and any clinical deployment would need to prove the system holds up across different body types, skin sensitivities and movement styles far more varied than a typical research cohort.
The Bigger Idea Behind the Tape
Farcas has described the underlying goal in blunt terms: “This research will support remote monitoring of the patient’s posture and movement throughout the day, with the ultimate goal of enabling personalized physical therapy treatments and improving health outcomes.” Loh has pointed to the unusually cross-disciplinary makeup of the team — structural engineers, physical therapists and machine learning researchers working from the same dataset — as what makes the approach possible at all. If the models keep improving, the long-term vision is a therapist who can look at a dashboard between appointments and see not just whether a patient did their prescribed stretches, but whether their spine is actually moving differently at the grocery store or the office than it did a month ago.
What’s Next
The UC San Diego team has said that, pending continued positive results, they hope to extend the Motion Tape approach beyond low back pain to conditions like spinal cord injury and stroke recovery, where continuous, low-cost movement tracking outside a clinical setting could be even more valuable. For now, the work remains firmly in the research pipeline rather than on pharmacy shelves, but it points toward a version of physical therapy where the sensor doing the most useful work isn’t a smartwatch at all, but a strip of smart fabric nobody else can see under a shirt.
Photo: PNW Production / PEXELS via Pexels
