Lenhoff Mark W, Zanini Silvia, Chingas Gregory, Patel Snehal, DiGiacomo Robert, Hillstrom Howard J
Department of Rehabilitation, Hospital for Special Surgery, 535 E 70th ST, NY 10021, USA.
Department of Rehabilitation, Hospital for Special Surgery, 535 E 70th ST, NY 10021, USA.
Med Eng Phys. 2025 Aug;142:104375. doi: 10.1016/j.medengphy.2025.104375. Epub 2025 Jun 4.
Markerless motion capture is faster than a traditional marker-based method, but its reliability is not well established. The limitations of this new technology must be understood to ensure its appropriate utilization.
Does a markerless motion capture system exhibit intra-device and inter-device reliability with an Intraclass Correlation Coefficient (ICC) > 0.70 during functional tasks?
Twenty test subjects were asked to squat and walk three times to compute the test-retest (repeated task with the same scaled skeletal model applied) and remove-replace (repeated task with two independently generated skeletal models applied) intra-device reliability for each of three devices and the inter-device reliability between the devices. Two devices were configured in a standard volume, and one device was configured sub-optimally in a larger volume.
For intra-device reliability, the average squat range of motion (ROM) test-retest ICC was 0.86 and average squat remove-replace ICC was 0.74. The average walk ROM test-retest ICC was 0.70 and the average walk remove-replace ICC was 0.72. For inter-device reliability between the two devices in the standard volume, the average ICC was 0.82 for squat and 0.72 for walk. With the sub-optimally configured device included for a three-device analysis the average inter-device reliability ICC dropped to 0.55 for squat and 0.46 for walk.
Larger field of view, suboptimal camera placement and tasks performed away from the center of the field of view presented challenges to the DARI markerless mocap system. Intra-device and inter-device reliability can achieve a level of ICC > 0.7 using DARI markerless mocap technology using an optimized setup, but the transverse plane kinematics exhibit the weakest performance. Optimal camera placement (i.e., field of view) is critical for reliable results.
无标记运动捕捉比传统的基于标记的方法更快，但其可靠性尚未得到充分确立。必须了解这项新技术的局限性，以确保其合理应用。
在功能任务期间，无标记运动捕捉系统的组内相关系数（ICC）>0.70时，是否表现出设备内和设备间的可靠性？
20名测试对象被要求进行三次深蹲和行走，以计算三种设备各自的重测（对同一缩放骨骼模型应用重复任务）和移除-替换（对两个独立生成的骨骼模型应用重复任务）设备内可靠性以及设备间的可靠性。两个设备以标准体积配置，一个设备以次优方式在更大体积中配置。
对于设备内可靠性，平均深蹲运动范围（ROM）重测ICC为0.86，平均深蹲移除-替换ICC为0.74。平均行走ROM重测ICC为0.70，平均行走移除-替换ICC为0.72。对于标准体积中两个设备之间的设备间可靠性，深蹲的平均ICC为0.82，行走的平均ICC为0.72。将次优配置的设备纳入三设备分析时，深蹲的平均设备间可靠性ICC降至0.55，行走的降至0.46。
更大的视野、次优的相机放置以及在视野中心以外执行的任务给DARI无标记运动捕捉系统带来了挑战。使用优化设置的DARI无标记运动捕捉技术，设备内和设备间可靠性可以达到ICC>0.7的水平，但横向平面运动学表现出最弱的性能。最佳相机放置（即视野）对于获得可靠结果至关重要。
