Analyzing gait data measured by wearable cyborg hybrid assistive limb during assisted walking: gait pattern clustering

Abstract

Introduction: The wearable cyborg Hybrid Assistive Limb (HAL) is a therapeutic exoskeletal device that provides voluntary gait assistance using kinematic/kinetic gait data and bioelectrical signals. By utilizing the gait data automatically measured by HAL, we are developing a system to analyze the wearer's gait during the intervention, unlike conventional evaluations that compare pre- and post-treatment gait test results. Despite the potential use of the gait data from the HAL's sensor information, there is still a lack of analysis using such gait data and knowledge of gait patterns during HAL use. This study aimed to cluster gait patterns into subgroups based on the gait data that the HAL automatically collected during treatment and to investigate their characteristics.

Methods: Gait data acquired by HAL, including ground reaction forces, joint angles, trunk angles, and HAL joint torques, were analyzed in individuals with progressive neuromuscular diseases. For each measured item, principal component analysis was applied to the gait time-series data to extract the features of the gait patterns, followed by hierarchical cluster analysis to generate subgroups based on the principal component scores. Bayesian regression analysis was conducted to identify the influence of the wearer's attributes on the clustered gait patterns.

Results: The gait patterns of 13,710 gait cycles from 457 treatments among 48 individuals were divided into 5-10 clusters for each measured item. The clusters revealed a variety of gait patterns when wearing the HAL and identified the characteristics of multiple sub-group types. Bayesian regression models explained the influence of the wearer's disease type and gait ability on the distribution of gait patterns to subgroups.

Discussion: These results revealed key differences in gait patterns related to the wearer's condition, demonstrating the importance of monitoring HAL-assisted walking to provide appropriate interventions. Furthermore, our approach highlights the usefulness of the gait data that HAL automatically measures during the intervention. We anticipate that the HAL, designed as a therapeutic device, will expand its role as a data measurement device for analysis and evaluation that provides gait data simultaneously with interventions, creating a novel cybernics treatment system that facilitates a multi-faceted understanding of the wearer's gait.

Keywords: cybernics treatment; gait analysis; hierarchical clustering; hybrid assistive limb (HAL); neuromuscular diseases; wearable devices.

Figure 1

Conceptual diagram of the use of gait data measured by HAL. The process includes treatment, data accumulation, and analysis using HAL. In addition to the conventional intervention using HAL, a more sophisticated treatment system will be constructed by using the gait time-series data that HAL automatically measures during the assistance.

Figure 2

Patterns of ground reaction forces for each subgroup. The loads are normalized by the wearer's weight. The n in the graph is the number of gait cycles in the cluster. The solid lines are the average patterns of all gait cycles in the cluster, while the colored shadings are the mean ± standard deviation range. The magenta and cyan colors represent the values measured by the sensors on the heel and toe side of the reference leg (the leg that begins bearing weight at 0% of the gait cycle), respectively. The red and blue colors represent the heel and toe side of the opposite leg, respectively.

Figure 3

Patterns of hip joint angle for each subgroup. Positive angles indicate flexion, negative angles indicate extension, and zero indicates the hip joint is straight. The n in the graph is the number of gait cycles in the cluster. The solid black line is the average pattern of all gait cycles in the cluster, and the gray shading is the mean ± standard deviation range.

Figure 4

Patterns of knee joint angle for each subgroup. Positive angles indicate flexion, negative angles indicate extension, and zero indicates the knee joint is straight. The n in the graph is the number of gait cycles in the cluster. The solid black line is the average pattern of all gait cycles in the cluster, and the gray shading is the mean ± standard deviation range.

Figure 5

Patterns of trunk pitch angle for each subgroup. The trunk pitch angle represents the tilt of the HAL control system behind the wearer's lumbar at the sagittal plane. Positive trunk pitch angles indicate anterior tilt, negative trunk pitch angles indicate posterior tilt, and zero indicates upright posture with no tilt. The n in the graph is the number of gait cycles in the cluster. The solid black line is the average pattern of all gait cycles in the cluster, and the gray shading is the mean ± standard deviation range.

Figure 6

Patterns of trunk roll angle for each subgroup. The trunk roll angle represents the tilt of the HAL control system behind the wearer's lumbar at the frontal plane. Positive trunk roll angles indicate tilt toward the reference leg side, negative trunk roll angles indicate tilt toward the opposite leg side, and zero indicates upright posture with no tilt. The n in the graph is the number of gait cycles in the cluster. The solid black line is the average pattern of all gait cycles in the cluster, and the gray shading is the mean ± standard deviation range.

Figure 7

Patterns of hip joint torque produced by HAL for each subgroup. Positive torque indicates a force in the flexion direction, negative torque indicates in the extension direction, and zero indicates no torque is applied. The n in the graph is the number of gait cycles in the cluster. The solid black line is the average pattern of all gait cycles in the cluster, and the gray shading is the mean ± standard deviation range.

Figure 8

Patterns of knee joint torque produced by HAL for each subgroup. Positive torque indicates a force in the flexion direction, negative torque indicates in the extension direction, and zero indicates no torque is applied. The k in the graph is the number of gait cycles in the cluster. The solid black line is the average pattern of all gait cycles in the cluster, and the gray shading is the mean ± standard deviation range.