Accuracy of Temporo-Spatial and Lower Limb Joint Kinematics Parameters Using OpenPose for Various Gait Patterns With Orthosis

Abstract

A cost-effective gait analysis system without attachments and specialized large environments can provide useful information to determine effective treatment in clinical sites. This study investigates the capability of a single camera-based pose estimation system using OpenPose (OP) to measure the temporo-spatial and joint kinematics parameters during gait with orthosis. Eleven healthy adult males walked under different conditions of speed and foot progression angle (FPA). Temporo-spatial and joint kinematics parameters were measured using a single camera-based system with OP and a three-dimensional motion capture system. The limit of agreement, mean absolute error, absolute agreement (ICC 2, 1 ), and relative consistency (ICC 3, 1 ) between the systems under each condition were assessed for reliability and validity. The results demonstrated that most of the ICC for temporo-spatial parameters and hip and knee kinematics parameters were good to excellent (0.60 -0.98). Conversely, most of the ICC for ankle kinematics in all conditions were poor to fair (<0.60). Thus, the gait analysis using OP can be used as a clinical assessment tool for determining the temporo-spatial, hip, and knee sagittal plane angles during gait.

Reliability and validity of OpenPose for measuring hip-knee-ankle angle in patients with knee osteoarthritis

“…The accuracy of OpenPose estimation in this study was higher than in previous reports 20 , 24 , 35 . Ota et al reported an ICC (2, 1) of 0.8 for maximum knee flexion angle in squatting and 0.62–0.67 for maximum knee flexion angle in walking on a treadmill with regard to the measurement validity of OpenPose compared with a 3D motion analysis system 20 , 24 .…”

“…Ota et al reported an ICC (2, 1) of 0.8 for maximum knee flexion angle in squatting and 0.62–0.67 for maximum knee flexion angle in walking on a treadmill with regard to the measurement validity of OpenPose compared with a 3D motion analysis system 20 , 24 . Yamamoto et al 35 similarly reported an ICC (2, 1) of 0.54–0.67 for maximum knee flexion angle during walking on a treadmill. Squatting differs from walking in that the positional relationship between the subject and the camera is less likely to change.…”

Verification of gait analysis method fusing camera-based pose estimation and an IMU sensor in various gait conditions

“…In the OP method, the TLA was calculated from the angle between the vector joining the dominant hip and small toe keypoints and a perpendicular line passing through center of the camera image. Knee joint angles obtained using OP methods tend to be underestimated by approximately 5° 21 , 36 ; therefore, we compensated by adding 5° to the knee joint angle using OP methods.…”

“…Some previous study reported that OP can measure temporo-spatial and kinematic parameters in the sagittal plane during gait without substantial error 19 , 20 . Our previous study revealed that single RGB camera gait analysis by OP can measure several temporo-spatial parameters and sagittal plane joint angles with good to excellent agreement and consistency compared with optoelectronic 3DMC systems; however, the lower limb joint angle error increased in some gait conditions, including out of the 2D image plane excessive motion, such as increasing foot progression angle (FPA) 21 . In particular, the peak ankle joint angles had substantial error.…”

“…The secondary aim was to clarify temporo-spatial parameters and lower limb joint angles during four overground gait conditions with combinations of self-selected and slow speeds and with their normal FPA and with a larger FPA, as measured by 3DMC and OP. Although our previous study examined the accuracy of temporo-spatial parameters and joint kinematics by OP, we measured only treadmill gait 21 . Some previous studies have compared 3DMC and markerless gait analysis systems on overground gait 10 , 24 ; however, the four gait conditions were not measured, including out of 2D image plane excessive motion, such as large FPA.…”