Characterizing intersection variability of butterfly diagram in post-stroke gait using Kernel Density Estimation

• Published in: Gait & posture 2020-02, Vol.76, p.157-161
• Main Authors: Lee, Yun-Ju, Liang, Jing Nong
• Format: Article
• Language: English
• Subjects: Butterfly diagram; Center of pressure; Gait; Kernel Density Estimation; Locomotor variability
• ISSN: 0966-6362; 1879-2219
• DOI: 10.1016/j.gaitpost.2019.12.005

•Conventional butterfly diagrams have limitations in describing gait variability.•Novel approach using Kernel Density Estimation analysis.•Characterizing highly variable hemiparetic gait in individuals post-stroke. Center of pressure (COP) trajectory during treadmill walking have been commonly presented using the butterfly diagram to describe gait characteristics in neurologically intact and impaired individuals. However, due to the large amount of displayed information, the butterfly diagram is not an efficient solution to visualize locomotor variability. The purpose of this study was to evaluate post-stroke locomotor variability by applying Kernel density estimation (KDE) on the intersections of the butterfly diagram, and to compare KDE derived metrics with conventional metrics of gait symmetry and variability. Bilateral toe-off (TO) and initial contact (IC) points of the butterfly diagram were determined to calculate the COP symmetry index and the intersections of bilateral TOIC. Subsequently, the intersections during the walking window were used to evaluate its density and variability by Kernel density estimation. Standard deviations of step width and step length were compared between groups. Using the KDE surface plots we observed 4 characteristically different patterns with individuals post-stroke, which were associated with functional status quantified using walking speed and lower extremity Fugl-Meyer scores. However, locomotor variability quantified using standard deviations of step width and lengths did not differ between groups. Significance & Novelty: This paper presents a novel approach of using KDE analysis as a better and more sensitive method to characterize locomotor COP variability in individuals with post-stroke hemiparesis, compared to conventional metrics of gait symmetry and variability.