UC San Diego

Clinical evaluations of neuromuscular disorders, such as hypertonia, have been based on perception and can vary from physician to physician. In efforts to provide an objective assessment a multimodal sensing glove was developed, measuring the force and motion trajectories, recording the muscle's resistance to evaluate severity. However, due to grip and evaluation speed differences, variations in the output magnitude between raters require calibration for effective cross comparison. In this work, we demonstrate a multivariate regression model to address interrater variability across clinicians. The model uses both the acceleration and force recorded from the glove to calibrate each rater's assessment. In addition, it effectively increased the kernel density overlapping area between clinicians after calibration, reducing variability and increasing force output consistency. Ultimately, this work will enhance hypertonia monitoring accessibility to support rehabilitation specialists.