Assessing inertial measurement unit locations for freezing of gait detection and patient preference.

Only seven people walked the sensor course, and only four wore the eleven-sensor set, so model performance may be unstable. The survey participants and walking participants were not the same people, and the survey asked them to extrapolate home wearability from clinic experience. Walking was collected in a clinic on a course, not in the home, and participants were off medication and deep brain stimulation. All participants had known freezing of gait, so results may not apply to people without freezing or with different treatment. Models were trained mainly on freezing during continuous walking, not freezing at gait initiation. The survey sample was skewed toward people who wanted to help research. The raw-data models underperformed some previously reported hand-engineered-feature models, and comparisons across the six- and eleven-sensor sets should not be treated as absolute.

Funded by the National Institute of Neurological Disorders and Stroke, the National Institute of Biomedical Imaging and Bioengineering, the National Science Foundation, a Stanford Bio-X Bowes Graduate Fellowship, and an Inventec Stanford Graduate Fellowship. The supplied text does not report additional author conflicts of interest.

Do not conclude that these sensors can already monitor freezing of gait reliably at home or that wearing one ankle sensor will change therapy. The walking data came from seven people with known freezing in a clinic course, off medication and deep brain stimulation, and the survey asked 16 people to imagine 12-hour wear based on clinic experience.