Validity and reliability of deep-learning-based 3D markerless motion capture to measure functional movement

The Ergo had a high degree of validity compared with the measured values of marker-based motion capture system. The linear regression analysis showed a high to excellent validity and the RMSEs were ranging from 2.33±0.46 ° to 6.25±1.65° in all time series joint angles. The intraclass correlation coefficient showed an excellent validity in all joint peak angles and bland-Altman analysis revealed no significant bias. Furthermore, The Ergo showed high convenience and efficiency results in post-question surveys.