Validation And Clinical Application Of A Novel Markerless Motion Capture System For Gait Assessment

Objective:Gait abnormality is one of the most common Sequelae of stroke,and it’s essential to evaluate the efficacy of rehabilitation and to develop rehabilitation plans of comprehensive and objective gait analysis.A commonly used laboratory gait assessment tool,the large infrared motion capture systems,which have a high accuracy but expensive equipment and demanding experimental environments,making it difficult to be widely used in the clinical,and Gait assessment Scales do not provide objective quantification of gait function.For this reason,it’s urgent to have a low-cost and portable gait assessment tool for clinical.The objectives of this study are:(i)Validating the accuracy of the markerless motion capture system based on Azure Kinect for gait assessment,using the infrared 3D motion analysis system as the gold standard.(ii)To analyze the feasibility for clinical test,using the markerless motion capture for gait assessment in stroke patients and comparing it with clinical gait assessment scales.Methods: Firstly,we measured the gait of 27 healthy subjects simultaneously by Large infrared 3D motion capture Qualysis system and Azure Kinect markerless motion capture system.Then,we acquired the spatiotemporal and kinematic parameters,calculated the Intraclass Correlation Coefficient(ICC values)and Root Mean Square Error(RMSE)between systems,and plotted Bland-Altman plots and linear regression plots of spatiotemporal parameters to verify the accuracy of Azure Kinect system.On top of this,Azure Kinect was used to analyze the gait characteristics of 23 stroke patients,and spatiotemporal、kinematic parameters and gait symmetry were correlation analysis with the TGA,a gait analysis scale commonly used in clinical practice for stroke.Results: Compared with the Qualysis,Azure Kinect markerless motion capture system measured a high degree of consistency in stride,step length,gait speed and step width(ICC >0.75,p<0.001),with RMSE≤5cm and RMSE of step speed was2.6cm/s;B-A plot found that 90.91%,94.8%,94.8%,96.1% and 90.91% of left and right stride length,step length,step width and step speed points were distributed within the 95% Lo A,respectively;linear regression plots found slopes of 0.97,0.90,0.97,0.54 and 0.94 for left and right stride length,step length,step width and step speed,respectively,and with coefficients of determination R~2 of 0.78,0.83,0.60,0.51 and 0.88,respectively.There was no significant difference between the ROM measured in the left and right sides of the two systems,the ROM measured in the hip and knee joints were excellent consistency between systems(ICC values >0.75,p<0.001),with RMSE <5°;the correlation between the left and right ankle joints were good consistency(ICC values>0.6,p<0.001),with RMSEs of 16.45°and18.67°,respectively.The gait cycle was divided into four gait phases: initial double support phase DLS,single support phase SLS,terminal double support phase DLS and swing phase SW,and it was found that all hip ROMs were excellent consistency;The knee was excellent consistency except for the right side,which had a good consistency in the SLS;The ankle was no correlation on the left side of the initial DLS,fair consistency on the left side of the SLS and a good consistency on the right side of the SW,the rest were excellent consistency.The hip in the final DLS and the knee in the SW had a larger error,RMSE was slightly greater than 5°,while the rest RMSE were less than 5°,and the RMSE of the ankle in all four phases were greater than 5°.Gait parameters measured using the Azure Kinect in stroke patients revealed no significant differences in stride and gait speed between the healthy and affected side(p>0.05),while significant differences in hip and knee ROM(p < 0.01).The correlation between TGA scale and stride,gait speed was moderately,with r=0.68 and0.58.There were weak or moderate collection between TGA and hip、knee ROM,with r= 0.37,0.61,0.48 and 0.57,respectively.None of symmetry parameters of stride and gait speed correlated with the scale,and of the symmetry parameters of hip and knee angles,only the direction of the knee cyclogram showed a moderate negative correlation with TGA(r =-0.42),the rest did not correlate.Conclusion:The markerless motion capture system based on Azure Kinect for the measurement of spatiotemporal gait parameters and sagittal plane full gait cycle hip and knee ROM in gait analysis,which can quantify gait parameters scientifically,but the measurement accuracy for the ankle joint has yet to be improved.Meanwhile,the gait parameters measured by Azure showed simultaneous validity with the gait assessment scales commonly used in clinical,and the sensitivity of the basic gait parameter indexes and symmetry indexes responding to the patient’s gait function was not used,which should be applied comprehensively to make a comprehensive and objective evaluation of gait function.With further upgrading of the system and improvement of tracking performance in the future,the Azure system can be widely used as a low-cost tool for quantitative assessment of assisted gait in clinical,community and home rehabilitation.