Comparison of Two Video-Based Methods for Knee Joint Angle Measurement: A Preliminary Study.

The measurement of the range of motion of human lower limb joints is of fundamental importance for biomechanical assessment. To date, there are numerous video-based methodologies for joint angle analysis, such as vision systems with passive markers and vision systems using deep learning algorithms. This study conducted a comparative analysis between two video-based measurement methods for analyzing the knee angle in the sagittal plane in locomotion activities. A healthy adult subject was involved in this study. Two motor conditions were analyzed: slow walking (3.6 km/h) and fast walking (5,0 km/h) on a treadmill. Two RGB cameras were used, one placed sagittally to the subject and the other at 45°. A 6-camera optoelectronic system was used as the gold standard. The knee angle in the sagittal plane was analyzed using a deep learning method (TC-Former) and a passive marker method (Blob Analysis). The results for slow walking on the treadmill show a relative error expressed in percentages of 14.7±11.0 and 7.5±3.5 for the first and second methods respectively. The results for the fast walking task show a relative error of 20.0±7.7 and 8.9±4.0 for the first and second methods, respectively.