Skip to main content

Artificial Intelligence and Wearable Sensors Put the Gait Lab in the Patient’s Pocket

UCSF orthopaedic surgeon Stefano Bini, MD, in collaboration with Google’s Advanced Technologies and Products (ATAP) division, is developing technology that will revolutionize how providers measure joint function and quantify joint disease. This groundbreaking approach uses cutting-edge machine-learning algorithms and data collected from inexpensive, wearable inertial sensors to accurately replicate gait lab outputs for a fraction of the cost of traditional methods, and in any context. Doing so will allow researchers, surgeons, patients and payers to better understand the drivers of clinical function and develop objective, personalized and dynamic outcome measures of function that were previously not possible.

“We’re re-creating the gait lab in the patient’s pocket,” Bini said. “We will be able to collect several months of patient data in real time and while patients are doing their daily activities in the environments in which they live. Currently, the best we can do is 10 seconds or so of data captured in a lab with patients doing highly specific tasks. This data is rather artificial, expensive and not accessible. Once the system is fully developed, we will be able to get an accurate, dynamic measure of joint function in four dimensions - the fourth being time - and reference it to a large data set of normal function. It’s an objective measure of outcome we have never had access to before.”

The use of highly affordable sensors costing less than $100 combined with Google’s algorithms and training data, which is licensed for free to users, make this a low-cost solution that will provide gait-lab-quality kinetic and kinematic data to anyone. Bini and his collaborators see a massive opportunity for developers to create both patient- and physician-facing apps that track and measure clinical outcomes to assist with patient care and triage using quantifiable, objective metrics.

Patient data captured 24-7

One example of how this data will be used is to provide detailed metrics to patients recovering from joint replacement surgery that can be compared to data from others at the same stage of recovery. The objective nature of the information, coupled with established patient-reported outcome measures, will help physicians and their patients assess progress and adjust physical rehabilitation plans far more accurately than in the past.

“While we will start with arthroplasty patients, we see applications for tracking and measuring patient function following knee ligament surgery, trauma surgery, or simply following joint injury or arthritis progression,” Bini said. Eventually, this technology can be applied to any joint in the body.

Furthermore, Bini and his team expect that the sensor-derived data will help to pinpoint the root cause of pain. “We know that certain gait patterns correlate to problems in specific muscles or the back, which can in turn cause pain or poor function in the knees,” Bini said. “We can make some inferences from the data.” Patients may have weak quadriceps, hip problems or other issues that impact knee function and, if correctly identified, can be addressed through physical therapy. 

“This is a new measurement tool that changes the game,” Bini said. “Soon, instead of measuring outcomes in the clinic once or twice following surgery, we can get objective measures of function 24-7 on any patient wearing sensors in their socks or shoes.”

Blueprint for a digital twin

When the data set is larger, Bini expects to be able to generate personalized outputs that correlate to a patient’s height, weight, gender, fitness level, social determinants of health and other factors. “The data could be normalized to the patient’s peers, giving us the ability to create personalized outcome metrics as opposed to the ‘one size fits all’ approach we have to measuring outcomes today,” Bini said.

“We are getting close to creating a digital twin, basically a representation of how a patient’s knee moves through space,” he added. The data could be integrated into the patient’s personal health record and linked to an app that translates the information for the patient in a gamified context.

The data will also be valuable for a host of potential users, from pharmaceutical researchers needing to quantify the impact of a drug on patient gait to orthopaedic device manufacturers looking to measure the results associated with a specific design change to an implant. “I haven't stopped coming up with ways of using this information,” Bini said.

“At the core of our work is the ability to replicate the gait lab with a simple, inexpensive sensor that can be easily worn in the shoe or around the knee,” he continued. “This is different from what anyone else is doing because of the sheer number of complex variables we are reproducing with incredible accuracy and at very low cost.”

To learn more

UCSF Arthritis and Joint Replacement Center

Phone: (415) 353-2808  | Fax: (415) 885-3862

Refer a patient

Clinical trials