Advances in wearable tech and AI open up new prospects for stroke patients recovering

Gustavo Balbinot, Associate Professor of Neurorehabilitation at the University of British Columbia, says a new study opens the door to the development of innovative technologies to improve stroke rehabilitation.

In an interview, he stressed that the findings from this study are not limited to treating stroke survivors, but could also be applied to people at risk of falls due to balance loss due to other causes such as dizziness or spinal injuries.

The study, published in the UK Journal of Clinical Rehabilitation after a rigorous peer-review process, used body-worn sensors to monitor more than 50 stroke patients while they performed mobility tests.

From the data collected, the team developed motion simulation models, which helped analyze and better understand how patients move during recovery.

Balbinot likened the process to throwing a stone into water and watching the ripples that spread out. Each person, he said, has a unique frequency of movement, a measure that can reflect the stability or irregularity of gait and balance.

The researchers found that people recovering from strokes tended to have smoother, slower movements, suggesting caution, while healthy people tended to have faster, stronger, and more “jerky” movements.

To enhance fall risk detection, the team developed proprietary software that can break down motion data into 3-second intervals.

The software automatically analyzes whether the user’s movements are too erratic or unstable, which Balbinot says is key to providing early warning when signs of danger are detected, preventing falls that could cause serious injury to the recovering person.

The team's goal is to integrate the software into smart wearable devices such as watches or bracelets, allowing users to receive real-time alerts when there are signs of loss of balance.

“If the software detects unusual movement, it can send a warning signal like, ‘You should sit down and rest for a bit,’ to help the user avoid getting into a dangerous situation,” he explained. Continuous monitoring every three seconds allows for near-instantaneous responses.

The software is also designed to “learn” from user data through machine learning. Over time, the system will gain a better understanding of each patient’s individual movement patterns, allowing it to make more accurate predictions about fall risk.

“This software will get smarter and smarter, and for each individual, it will know what is normal and what are signs to pay attention to,” Balbinot asserts.

The sensors used in the study not only measure speed but also track direction of movement. Balbinot said the technology is now advanced enough to integrate these sensors into clothing, making it more convenient for users to use on a daily basis without the need for bulky equipment.

In particular, the study also showed that this software is not only useful for patients but also helpful for clinicians. Intuitive and easy-to-interpret data from the system can support doctors to make more accurate treatment decisions.

The study concludes: “Integrating machine learning algorithms will open up the possibility of personalizing rehabilitation strategies, by analyzing individual movement patterns and predicting fall risk for each specific patient.”

The report also calls for further studies to validate the technology's long-term efficacy in clinical settings and its applicability to different patient groups.