Abstract
Introduction: There are evidences that the nervous system produces motor tasks using a low-dimensional modular organization of muscle activations, known as motor modules. Previous studies have identified characteristic motor modules across similar tasks in healthy population. This study explored the generalizability of motor modules across two families of walking-based (level-walking, downhillwalking and stair-decent), in-place ascending (sit-to-stand, squat-to-stand), and in-place descending (stand-to-sit and stand-to-squat) motor tasks in a group of six individuals undergone total knee replacement (TKR) surgery. Methods: Motor modules were extracted from the EMG data of CAMS-Knee dataset using non-negative matrix factorization technique. A distribution-based approach, employing three levels of k-means clustering, was then applied to find the shared and task-specific modules, and assess their representability among the whole task-trial data. Results and Discussion: Results indicated a four- and a seven-subcluster arrangement of the shared and task-specific motor modules, depending upon the membership criteria. The first arrangement revealed motor modules which were shared across all tasks (min coverage index: 76%; modules' distinctness range: 7.08-8.91) and the latter among tasks of the same family mainly, although there remained some interfamily shared modules (min coverage index: 81%; modules' distinctness range: 7.17-9.89). It was concluded that there are shared motor modules across walking-based and in-place tasks in TKR individuals, with their generalizability and representability depending upon the analysis method. This finding highlights the importance of the analysis method in identifying the shared motor modules, as the main building blocks of motor control.