Mobilise-D insights to estimate real-world walking speed in multiple conditions with a wearable device

Abstract

This study aimed to validate a wearable device's walking speed estimation pipeline, considering complexity, speed, and walking bout duration. The goal was to provide recommendations on the use of wearable devices for real-world mobility analysis. Participants with Parkinson's Disease, Multiple Sclerosis, Proximal Femoral Fracture, Chronic Obstructive Pulmonary Disease, Congestive Heart Failure, and healthy older adults (n = 97) were monitored in the laboratory and the real-world (2.5 h), using a lower back wearable device. Two walking speed estimation pipelines were validated across 4408/1298 (2.5 h/laboratory) detected walking bouts, compared to 4620/1365 bouts detected by a multi-sensor reference system. In the laboratory, the mean absolute error (MAE) and mean relative error (MRE) for walking speed estimation ranged from 0.06 to 0.12 m/s and - 2.1 to 14.4%, with ICCs (Intraclass correlation coefficients) between good (0.79) and excellent (0.91). Real-world MAE ranged from 0.09 to 0.13, MARE from 1.3 to 22.7%, with ICCs indicating moderate (0.57) to good (0.88) agreement. Lower errors were observed for cohorts without major gait impairments, less complex tasks, and longer walking bouts. The analytical pipelines demonstrated moderate to good accuracy in estimating walking speed. Accuracy depended on confounding factors, emphasizing the need for robust technical validation before clinical application.Trial registration: ISRCTN - 12246987.

Figure 1

Overview of (a) the TVS protocol, (b) the analytical pipeline applied to estimate walking speed from the wearable device data (WD), (c) the approach to validating walking speed estimated from the analytical pipeline.

Figure 2

Residual plots of walking speed for all true-positive WBs recorded in the laboratory (left) and during the real-world recording (right). The margin plots represent the overall speed and error distributions. The margin plots are further grouped by the performed tests for the laboratory and by the cohort for the real-world recordings. The light blue bars around the Limits of Agreement (LOA) (dashed horizontal lines) represent their bootstrapped confidence intervals. The dashed black line represents the result of a linear regression on all datapoints. The grey area around the regression line represents the bootstrapped 95% confidence intervals.

Figure 3

Residual plots for the walking speed combined over all identified WBs. For the laboratory tests the median over all WBs within one motor task is taken (left). For the real-world recording the median over all WBs in the entire real-world assessment is shown (right), where each datapoint represents an individual participant. The margin plots represent the overall speed and error distributions. The margin plots are further grouped by the performed tests for the laboratory and by the cohort for the real-world recordings. The light blue bars around the Limits of Agreement (LOA) (dotted horizontal lines) represent their bootstrapped confidence intervals. The dashed black line represents the result of a linear regression on all datapoints. The grey area around the regression line represents the bootstrapped 95% confidence intervals.

Figure 4

The dependency of the absolute walking speed error of all true-positive WBs from the real-world recording on the WB duration reported by the reference system. In the top, WB errors are grouped by various duration bouts. In the bottom the number of bouts within each duration group is visualized.

Figure 5

The walking speed estimations from the real-world recording of the reference system and the wearable device, from all WB within the respective duration bouts. The boxplots show the distribution over all WBs. The bars in the upper plot show the absolute difference between the medians of the distributions (see right y-axis). The bottom plot shows the number of WBs in each duration bout.

Figure 6

The dependency of the absolute walking speed error on the different defined complexity tasks (see text). The results are split by patient cohort. The “All” group represents the statistics over all WBs independent of the cohort.

Figure 7

Overview over the different algorithmic steps of the analytical pipeline with short explanations of the intermediate and final outputs of each of the algorithmic blocks; gait sequence detection (GSD), initial contact detection (ICD), cadence estimation (CAD) and stride length estimation (SL). The algorithm column indicates the used algorithms for the two pipelines P1 (HA, COPD, CHF). (MS, PD, PFF) and P2 (MS, PD, PFF) Short citations for the algorithms are provided below the figure. For more details see Table 1 in.