Performance of a multi-sensor wearable system for validating gait assessment: preliminary results on patients and healthy

Abstract

Real-world (RW) gait analysis is becoming increasingly important since in these conditions an individual’s mobility performance can be measured, in addition to what happens in a standardised lab environment, where mobility capacity is assessed. This is even more relevant in the presence of a motor disability, when continuous daily-life monitoring can unveil limitations in movement, better stratify disease progression and the efficacy of medical treatments [1]. Wearable inertial measurement units (IMUs) represent the most valid solution to tackle this challenge, but their validation requires RW gait data from a reliable Gold Standard (GS). This study evaluates the performance of a wearable multi- sensor system (INDIP) which was specifically conceived to be used as a RW GS [2]. In particular, the INDIP performance in estimating digital mobility outcomes (DMOs) was assessed against a stereophotogrammetric (SP) system, on healthy older adults and patients from five different cohorts, following a purposely designed experimental protocol [3]. The INDIP system includes three IMUs (sampling frequency (fs) =100 Hz) positioned on lower back and feet, two pressure insoles (16 sensing elements, fs=100 Hz) and two time-of-flight distance sensors (fs=50 Hz, range=0.2 m) [2]. Experiments were carried out on 107 participants from six different cohorts (Healthy Adults (HA), Parkinson’s Disease (PD), Multiple Sclerosis (MS), Chronic Obstructive Pulmonary Disease (COPD), Congestive Heart Failure (CHF), Proximal Femoral Fracture (PFF)) while performing a complex eleven-task protocol [3]. Walking bouts (WBs) were obtained from the INDIP data, according to the flowchart described in [2]. Then, for each WB, relevant DMOs were computed. Accuracy of the estimated DMOs was evaluated at cohort level in terms of median relative error (%). For sake of brevity, here we only report walking speed results. A total of 103 subjects were considered for the analysis (1234 WBs in total). Results at cohort level for WB walking speed are presented in Table 1. Median relative errors ranged from 2.02% to 2.58%. Results reported for the WB walking speed showed very low median relative errors across all cohorts. For each cohort, the distributions of values from the INDIP and SP were very similar as shown by the means and standard deviations in Table 1 . Such low differences are excellent, especially when considering the complexity of the tasks included in the experimental protocol. In conclusion, the INDIP system is an accurate reference system for RW gait analysis.