Validation of markerless motion capture for spatiotemporal gait measures in people with Parkinson's disease

Spatiotemporal gait measures are increasingly important for the characterization of gait impairments in people with Parkinson's disease (PD). Hence new assessment methods, like markerless motion capture using Theia3D, are of rising interest. This study aims to validate markerless motion capture in people with PD to retrieve spatiotemporal gait measures during overground and treadmill walking. Overground- and treadmill walking at comfortable speed were assessed in 30 individuals with PD. Markerbased and markerless motion capture were carried out in randomized order with the markerbased method using a model with markers at the heel, the lateral ankle and the toe. The markerless data (Theia Markerless Inc., Kingston, ON, Canada) was analyzed 1) with the same model as the markerbased data and 2) within the automatic report pipeline. Agreement between the measurements was assessed via mean differences, Pearson correlation coefficients, intraclass correlation coefficients and Bland-Altman methods. The results show good to excellent agreement (ICC > 0.75) between spatiotemporal gait measures of the two measurements for overground walking, except for double support time (ICC < 0.5), when analyzed with our custom algorithm. For treadmill walking all outcomes had excellent agreement (ICC > 0.9). Mean differences were below the minimal clinical differences for gait speed and step length. For the results of the automatic report during overground gait agreement between methods was poor except for gait speed and step length. Overall, the markerless motion capture system used in this study provides valid results for the assessment of spatial and temporal gait variables in people with PD, when using a gait detection algorithm that is suitable for this population.