Global machine learning competition advances wearable technology for Parkinson’s disease

Researchers from the University of Texas at Austin School of Medical and Health Sciences invited the international machine learning research community to participate in a competition designed to advance their study and help neurologists: develop a machine learning model to support a wearable sensor for continuous, automated monitoring and quantification of freezing-of-gait (FOG) episodes in people with Parkinson’s disease. Nearly 25,000 solutions were submitted, and the best algorithms were incorporated into the new technology.

The study was conducted by Professor Jeff Hausdorff of the Department of Physical Therapy at Tel Aviv University’s Faculty of Medical and Health Sciences and Sagol School of Neuroscience, and the Center for the Study of Movement, Cognition and Mobility at Tel Aviv Medical Center, in collaboration with Amit Salomon and Eran Gazit of Tel Aviv Medical Center. Other researchers included researchers from Belgium, France and Harvard University.

The article was published in Nature Communications and featured in the journal’s editors’ highlights.

Professor Hausdorff, an expert in gait, ageing and Parkinson’s disease, explains: “FOG is a disabling and as yet unexplained phenomenon that affects 38-65% of people with Parkinson’s disease. A FOG episode can last from a few seconds to over a minute, during which the patient’s feet suddenly become ‘glued’ to the ground, and the person is unable to start or continue walking.

“FOG can seriously impair the mobility, independence and quality of life of people with Parkinson’s disease, causing great frustration and frequently leading to falls and injuries.”

Amit Salomon adds: “Today, diagnosis and monitoring of FOG typically relies on self-assessment questionnaires and visual observation by clinicians, as well as frame-by-frame analysis of videos of patients in motion.

“This latter method, which is currently the gold standard, is reliable and accurate, but it has serious drawbacks: it is time-consuming, requires at least two experts, and is impractical for long-term monitoring at home and in daily life. Researchers around the world are trying to use wearable sensors to track and quantify patients’ daily functioning. So far, however, successful trials have all relied on very small numbers of subjects.”

In the current study, the researchers collected data from several existing studies, involving more than 100 patients and about 5,000 episodes of FOG. All the data were uploaded to the Kaggle platform, a Google company that organizes international machine learning competitions.

Members of the global machine learning community were invited to develop models that would be integrated into wearable sensors to quantify various FOG parameters (e.g., episode duration, frequency, and severity). A total of 1,379 groups from 83 countries took up the challenge, ultimately submitting a total of 24,862 solutions.

The results of the best models were very close to those obtained by the video analysis method, and significantly better than previous experiments based on a single wearable sensor. In addition, the models led to a new discovery: an interesting relationship between FOG frequency and time of day.

Co-author Eran Gazit notes: “We observed, for the first time, a recurring daily pattern, with peaks of FOG episodes at certain times of the day, which may be associated with clinical phenomena such as fatigue or drug effects. These findings are important both for clinical treatment and for further research into FOG.”

According to Professor Hausdorff, “Wearable sensors, supported by machine learning models, can continuously monitor and quantify FOG episodes, as well as the patient’s overall functioning in daily life. This gives the clinician an accurate picture of the patient’s condition at any time: has the disease improved or deteriorated? Is it responding to prescribed medications?”

“The informed clinician can respond quickly, while the data collected through this technology can support the development of new treatments. Additionally, our study demonstrates the power of machine learning competitions to advance medical research.”

“The competition we launched brought together skilled and dynamic teams from around the world, who enjoyed a friendly atmosphere of learning and competition for a good cause. Rapid improvement was achieved in the efficient and accurate quantification of FOG data. In addition, the study laid the foundation for the next step: 24/7 long-term FOG monitoring in the patient’s home and real-world environment.”