Parameters and Measures in Assessment of Motor Learning in Neurorehabilitation; A Systematic Review of the Literature

Abstract

Upper limb function, essential for daily life, is often impaired in individuals after stroke and cerebral palsy (CP). For an improved upper limb function, learning should occur, and therefore training with motor learning principles is included in many rehabilitation interventions. Despite accurate measurement being an important aspect for examination and optimization of treatment outcomes, there are no standard algorithms for outcome measures selection. Moreover, the ability of the chosen measures to identify learning is not well established. We aimed to review and categorize the parameters and measures utilized for identification of motor learning in stroke and CP populations. PubMed, Pedro, and Web of Science databases were systematically searched between January 2000 and March 2016 for studies assessing a form of motor learning following upper extremity training using motor control measures. Thirty-two studies in persons after stroke and 10 studies in CP of any methodological quality were included. Identified outcome measures were sorted into two categories, “parameters,” defined as identifying a form of learning, and “measures,” as tools measuring the parameter. Review's results were organized as a narrative synthesis focusing on the outcome measures. The included studies were heterogeneous in their study designs, parameters and measures. Parameters included adaptation (n = 6), anticipatory control (n = 2), after-effects (n = 3), de-adaptation (n = 4), performance (n = 24), acquisition (n = 8), retention (n = 8), and transfer (n = 14). Despite motor learning theory's emphasis on long-lasting changes and generalization, the majority of studies did not assess the retention and transfer parameters. Underlying measures included kinematic analyses in terms of speed, geometry or both (n = 39), dynamic metrics, measures of accuracy, consistency, and coordination. There is no exclusivity of measures to a specific parameter. Many factors affect task performance and the ability to measure it—necessitating the use of several metrics to examine different features of movement and learning. Motor learning measures' applicability to clinical setting can benefit from a treatment-focused approach, currently lacking. The complexity of motor learning results in various metrics, utilized to assess its occurrence, making it difficult to synthesize findings across studies. Further research is desirable for development of an outcome measures selection algorithm, while considering the quality of such measurements.

Machine Learning Methodology in a System Applying the Adaptive Strategy for Teaching Human Motions

“…Many types of parameters, such as position matching errors, movement time, velocity, motion range, and average torques or forces [1,52], can be considered. Moreover, with digitized motion signals, we can easily calculate parameters defined in the time and frequency domain.…”

“…Moreover, with digitized motion signals, we can easily calculate parameters defined in the time and frequency domain. For instance, in the time domain, the RMSE (Root Mean Squared Error) [52,53] can be calculated, whereas in the frequency domain, we can compute the mean amplitude related to certain frequency band, or the parameter IPNS (Integral of the Power spectrum density of Normalized Speed) [53]. The proper use of these parameters requires the establishment of ranges in time or space in which the parameters should be calculated.…”

“…To evaluate learning quality, we used three different parameters. The first factor was based on the RMSE parameter [53] and estimated the accuracy of the motion learned by the person [1,52]. We calculated the RMSE from each preprocessed one-dimensional signal S i = (s 1 i , s 2 i , .…”

“…The last parameter, denoted by E 3 , was used to evaluate the progress of learning. This can be expressed by a difference between the values of selected parameters that correspond to the testing phase and a certain stage before learning (or its early phase) [52]. The simple difference can be extended by a weight factor that determines the impact of the earlier phase.…”

Kinematic analysis of motor learning in upper limb body-powered bypass prosthesis training

“…Fortunately, motor learning and performance is well understood in the broader rehabilitation population. Accuracy, quickness, smoothness, and efficiency have been quantified using motion capture as a distance from a target location, movement velocity, normalized jerk, and path length respectively in numerous patient populations [3,[10][11][12]. In the stroke patient population for example, normalized jerk has been shown to decrease, indicating smoother movement, with training during a reaching task [13].…”