CFE 2019: Start Registration
View Submission - CMStatistics
B1275
Title: Functional registration of walking strides in high-density accelerometry data for estimation of gait asymmetry Authors:  Marta Karas - Johns Hopkins Bloomberg School of Public Health (United States) [presenting]
Ciprian Crainiceanu - Johns Hopkins Bloomberg School of Public Health (United States)
Jacek Urbanek - Johns Hopkins Bloomberg School of Public Health (United States)
Amy Bastian - Kennedy Krieger Institute and Johns Hopkins University School of Medicine (United States)
Ryan Roemmich - Kennedy Krieger Institute and Johns Hopkins University School of Medicine (United States)
Abstract: Ability to characterize human gait pattern has significant potential in health research and can help guide clinical decision making. For example, in stroke survivors, quantification of asymmetry in walking strides during and after an in-lab intervention is of particular interest. To address this problem, we use high-density, high-throughput wearable accelerometry data and propose a stride pattern registration framework. We use a two-parameter family of time-warping functions to estimate clinically relevant stride characteristics, including duration and asymmetry. To demonstrate the approach, we collect accelerometry data on a healthy adult walking on a split-belt treadmill under different conditions mimicking step-to-step asymmetry. To analyze the data, we first use ADaptive Empirical Pattern Transformation (ADEPT) - a fast and scalable method for strides segmentation. We then employ a parametrized stride pattern framework to further characterize segmented strides. We conclude that the parametrized adaptive pattern matching appears to be a promising approach for estimation of step asymmetry.