A Wearable Magneto-Inertial System for Gait Analysis (H-Gait): Validation on Normal Weight and Overweight/Obese Young Healthy Adults

Abstract

Background: Wearable magneto-inertial sensors are being increasingly used to obtain human motion measurements out of the lab, although their performance in applications requiring high accuracy, such as gait analysis, are still a subject of debate. The aim of this work was to validate a gait analysis system (H-Gait) based on magneto-inertial sensors, both in normal weight (NW) and overweight/obese (OW) subjects. The validation is performed against a reference multichannel recording system (STEP32), providing direct measurements of gait timings (through foot-switches) and joint angles in the sagittal plane (through electrogoniometers). Methods: Twenty-two young male subjects were recruited for the study (12 NW, 10 OW). After positioning body-fixed sensors of both systems, each subject was asked to walk, at a self-selected speed, over a 14-m straight path for 12 trials. Gait signals were recorded, at the same time, with the two systems. Spatio-temporal parameters, ankle, knee, and hip joint kinematics were extracted analyzing an average of 89 ± 13 gait cycles from each lower limb. Intraclass correlation coefficient and Bland-Altmann plots were used to compare H-Gait and STEP32 measurements. Changes in gait parameters and joint kinematics of OW with respect NW were also evaluated. Results: The two systems were highly consistent for cadence, while a lower agreement was found for the other spatio-temporal parameters. Ankle and knee joint kinematics is overall comparable. Joint ROMs values were slightly lower for H-Gait with respect to STEP32 for the ankle (by 1.9° for NW, and 1.6° for OW) and for the knee (by 4.1° for NW, and 1.8° for OW). More evident differences were found for hip joint, with ROMs values higher for H-Gait (by 6.8° for NW, and 9.5° for OW). NW and OW showed significant differences considering STEP32 (p = 0.0004), but not H-Gait (p = 0.06). In particular, overweight/obese subjects showed a higher cadence (55.0 vs. 52.3 strides/min) and a lower hip ROM (23.0° vs. 27.3°) than normal weight subjects. Conclusions: The two systems can be considered interchangeable for what concerns joint kinematics, except for the hip, where discrepancies were evidenced. Differences between normal and overweight/obese subjects were statistically significant using STEP32. The same tendency was observed using H-Gait.

Keywords: H-Gait; STEP32; gait analysis; joint kinematics; magneto-inertial sensors; obese; overweight; spatio-temporal parameters; wearable; young.

Figure 1

H-Gait and STEP32 sensor positioning. The images show the (a) frontal, (b) lateral and (c) rear view of a subject prepared for the bi-instrumented gait analysis: the MIMU sensor positioned below the medial malleolus is shown in panel (d), the footswitches in panel (e).

Figure 2

Bland-Altman plots: spatiotemporal parameters. Bland-Altman plots are shown for (a) cadence; (b) stance; (c) swing and (d) double support, for normal weigh subjects (blue dots) and overweight/obese subjects (red dots). In each plot, the mean and LoA (mean ± 1.96 SD) are shown for both groups. Values from left and right lower limbs were represented as separate dots, except for cadence, where left/right values were averaged.

Figure 3

Bland-Altman plots: joint ROMs. Bland-Altman plots are shown for (a) ankle ROM; (b) knee ROM and (c) hip ROM; for normal weigh subjects (blue dots) and overweight/obese subjects (red dots). In each plot, the mean and LoA (mean ± 1.96 SD) are shown for both groups. Values from left and right lower limbs were represented as separate dots.

Figure 4

Joint kinematics. Ankle, knee, and hip kinematics are compared between normal weight (NW) and overweight/obese subjects (OW), both for H-Gait and STEP32 systems. Mean ± SD is reported.