Longevity & AgingResearch PaperOpen Access

Wearable sensors and depth cameras agree on key gait measures in multiple sclerosis

In 26 people with MS, a wearable sensor system and a Kinect depth camera closely agreed on cadence, stride time and speed, but not on gait phases.

Thursday, November 19, 2026 0 views
Published in Mult Scler Relat Disord
Person with MS walking down a bright clinic hallway wearing small body sensors while a depth camera records on a cart

Summary

Gait impairment drives disability in multiple sclerosis (MS), yet common scales such as the patient-reported PDDS are coarse and miss subtle change. This single-center study compared two objective tools in 26 adults with MS walking 6 meters twice in a clinic hallway: wearable inertial sensors (Clario OPAL) and a Microsoft Kinect depth camera. The devices agreed strongly on cadence, stride duration and, to a lesser degree, gait speed. They agreed only moderately on gait-cycle phases, stride length (right leg) and mid-swing foot elevation. Gait speed correlated moderately with PDDS (r about -0.61), and the other metrics correlated weakly. This is a cross-sectional agreement study. It did not test change over time, treatment response, or accuracy against a gold-standard motion-capture system.

Detailed Summary

Walking problems are a leading cause of disability in MS, and most patients need a gait aid within about 15 years of diagnosis. The usual disability scales (EDSS, and its patient-reported cousin PDDS) are stepwise and nonlinear, and they are insensitive to small changes, especially in people with mild disease. Objective gait measurement that works in a routine clinic, and possibly at home, could catch early decline. This study asks whether a cheap consumer depth camera can match a validated wearable sensor system.

Twenty-six adults with MS (median age 54, 69% female, median disease duration 65 months; 13 relapsing-remitting, 9 secondary progressive, 2 primary progressive, 2 isolated syndromes) walked 6 meters, turned and walked back, twice, at their own pace. They wore seven OPAL inertial sensors (sternum, lumbar, wrists, feet) and were recorded at the same time by one Azure Kinect depth camera. The turn was excluded from the camera analysis. Disability was self-reported on the PDDS (median 1, range 0-5).

Time-based metrics agreed best. Cadence correlated strongly between devices (r=0.94-0.95), with the camera reading about 5 steps/min lower. Stride duration correlated at r=0.97, with mean differences of 0.03-0.06 s. Gait speed correlated at r=0.94 on the left and 0.78 on the right, with mean differences of 0.04 and 0.15 m/s. Stride length agreed on the left (r=0.8) but not the right (r=0.41). Gait-cycle phases (swing, stance, double support) agreed only moderately (r=0.4-0.7), with mean differences of -7.5% to 4.1%. Mid-swing foot elevation correlated poorly (r=0.44), though the average difference was under 1.5 cm. The authors suggest that the devices detect gait events differently (foot-ground contact versus toe position relative to the body), and that camera recalibration after the turn may explain the weaker right-leg results.

Against PDDS, gait speed correlated most strongly (about -0.61 for both devices), and other metrics correlated more weakly. The authors read this as evidence that the objective tools capture more than the coarse scale does. The study did not directly test that claim.

The main caveats are the small sample, a single center, and a short walk. There was no motion-capture or pressure-mat reference, so agreement between devices does not establish accuracy. Neither device was tested for change over time or for the home setting, and the study used PDDS rather than clinician-rated EDSS. Clinically meaningful thresholds for most gait metrics are still undefined.

Key Findings

  • In 26 people with MS, cadence (r=0.94-0.95) and stride duration (r=0.97) agreed strongly between wearable sensors and a depth camera.
  • Gait speed agreed well on the left (r=0.94) and less well on the right (r=0.78), with mean differences of 0.04-0.15 m/s.
  • Gait-cycle phases agreed only moderately (r=0.4-0.7) and mid-swing elevation poorly (r=0.44); right-leg stride length agreed weakly (r=0.41).
  • Gait speed correlated best with self-reported PDDS disability (about r=-0.61); other gait metrics correlated weakly with PDDS.
  • Agreement was measured only between devices, with no gold-standard reference, no longitudinal data and no home testing.

Methodology

Cross-sectional, single-center (Mayo Clinic) agreement study of 26 consecutively offered people with MS, each completing two 6-meter walk-turn-walk trials with simultaneous OPAL sensors and a Kinect depth camera. Agreement was assessed with Pearson correlation and Bland-Altman analysis, and associations with PDDS with Spearman correlation and PCA. Camera data excluded the turn and were processed offline.

Study Limitations

The sample was small and from a single center, with mostly mild disability (median PDDS 1), and the gait task was short. Neither device was checked against motion capture or a pressure walkway, so agreement does not equal accuracy. Disability was patient-reported rather than clinician-rated EDSS, and the claim of greater sensitivity than PDDS was not tested directly. Right-leg metrics and gait phase measures agreed less well.

Enjoyed this summary?

Get the latest longevity research delivered to your inbox every week.

Enter your email to subscribe: