Inertial data-based gait metrics correspondence to Tinetti Test and Berg Balance Scale assessments

The analysis of selected gait metrics and their correspondence to human gait and balance abilities is presented in this paper. Accelerometric and gyroscopic data have been acquired from 42 older patients for the entire Berg Balance Scale examination along with expert assessments and additional gait segments of a ca. 1 min duration evaluated by means of the Tinetti Test. The inertial data is analyzed for gait and step detection and subjected to a detailed gait analysis using four metrics for both legs separately: pitch angle change, average duration of step swing and stance phase, and the average phase duration ratio. The analysis is performed offline. The results are compared to expert assessments using Spearman's rank correlation coefficient with p < 0.001. Three metrics are yielding high correlation up to 0.797 with outcomes delivered by both considered examinations. Moreover, statistically significant differences are obtained for three metrics produced by patients divided into low and medium fall risk groups (Mann–Whitney U test, p < 0.007). The values of the temporal metrics are higher and the angle changes are lower in the medium fall risk group. The proposed features enable the assessment of gait quality and may be of great importance while designing novel tools for quick preliminary fall risk prediction.