Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

1 Introduction

Recent projects highlight how motor learning and a high level of attention control can potentially improve submaximal force production during recovery. Stroke patients might benefit as this pathology affects each year to 17 million people worldwide, being the second leading cause of disability [1]. We already provided preliminary evidence of the facilitation of learning processes from a kinematic point of view in healthy subjects with robotic devices [2]. Robotics can be used not only as standalone devices, but also in combination with virtual environments, providing the possibility of showing the patient and the physiotherapist the feedback of the performance of the task, which has been proven to improve rehabilitation [3]. In our previous work, we introduced the use of “tacit adaptability” (TA) –a symbiotic control strategy based on biomimetic mechanisms, which is based on “tacit learning” [4]– with a stroke patient [5], walking on a treadmill, modulating the compliance with attention levels estimated from EEG signals. Marchal–Crespo et al. [6] showed that random disturbances improved motor learning in the performance of a simple dorsi–plantarflexion task, because this variability may increase recovery by increasing the needed effort and attention into the task.

This study focuses on the assessment of detailed metrics of force production and position control and their correlation with submaximal force production control learning, as improving regularity in submaximal force production might help improve functionality and reduce disability [7] during a new task consisting in maintaining the position for early rehabilitation after stroke. Our aim is to characterize the capacity to perform the precision task of maintaining the position with a submaximal force production in healthy subjects, by using the game we have designed that addresses the important factors for motor training: precision, speed, and path directness toward the specific targets presented on the screen [8]. We provide a novel approach in using TA to modulate the compliance of a torque control, to in the end modulate the difficulty of the task and thus play around the concept of the “challenge point theory” [9] to try accelerate the rehabilitative process. We intend to improve the therapy for stroke by training the ability of maintaining the position of the foot, addressing the typical problem for these patients of drop foot, as we focus our actuation on dorsi/plantar–flexion.

2 Materials and Methods

In this section, we briefly present the platform we use for this study and expose the methodology.

2.1 Experimental Platform

A Motorized Ankle Foot Orthosis (MAFO) is used for this study, with a zero–torque control. This robotic platform permits to exert controlled torque profiles to the ankle joint of the subject. The TA module is explained in Fig. 1, as well as the rest of the platform.

Fig. 1.
figure 1

Experiment set–up. The controller of the robotic platform consists on a zero torque controller (PID), with the addition of the TA component (TA constant –\(K_{TA}\)– multiplied by alpha, angle between the robot and the subject, proportional to the interaction), modulating \(K_{TA}\) with the score. The user controls the position of the bird with the angular position of the ankle.

Full size image

2.2 Participant

One healthy subject, right–handed, 30 years old, was enrolled for this experiment.

2.3 Task

The experiment consisted in following the trajectories depicted via the visual paradigm, while the robot disturbed the movement by performing plantar and dorsiflex interleaved torque patterns (see Fig. 2 for explanation on the torque profiles). The aim was to improve the motor control by learning how to maintain the position to follow the trajectory in the screen, compensating the perturbations.

Fig. 2.
figure 2

Torque patterns. Torque to dorsiflexion (up) and plantarflexion (down) direction, with the behaviour of the TA module: (1) \(K_{TA} = 100\) leads to zero–torque control; (2) \(K_{TA} = 0\) leads to regular torque control, so up to 16 N \(\cdot \) m; and (3) \(K_{TA}\) between 0 and 100, closer to zero–torque control the higher \(K_{TA}\) is.

Full size image

(1) Task Description: The task consisted on five sessions, of 61 trajectories: (1) first trajectory to understand the dynamics of the exercise; (2) the next ten assessed the performance before the training (at fixed torque); (3) forty training trials were performed; and (4) finally, the last ten trials assessed the performance after the training. The five sessions were separated by one day between them.

(2) Data Analysis: We analyzed the results by looking into the score, and the error between performed and prescribed trajectories, for ten trajectories before (PRE) and after (POST) the 40 trials training. We provide standard deviation, and standard error, for score and root mean square error (RMSE) respectively, as a measurement of the dispersion on the performance over the assessments, meaning better motor control with higher homogeneity on the mean values, i.e. more similar values along the full assessment. We ran a 2–Way ANOVA of repetitive measures for the statistical analysis with assessment (PRE–POST) and session (1 to 5) factors and a Bonferroni’s post–hoc analysis, with a \(\textit{p}=0.05\).

3 Results

In the results we observed statistical (session factor, \(\textit{p}<0.05\)) increment on the score \((F(4,36)=15.8, \eta _p^{2}=64\%)\) and reduction on the RMSE \((F(4,36)=15.5, \eta _p^{2}=63\%)\). Post–hoc analysis revealed significant changes inter–sessions, as shown in Fig. 3. We found no intra–session significant changes.

Fig. 3.
figure 3

Scores and RMSE (average, and standard deviation for score, and standard error for RMSE) before (PRE) and after (POST) each training assessments. Dotted line represents the linear fitting of the data for PRE (black) and POST (red) assessments.

Full size image

4 Conclusions and Future Work

The reduction on the variability and the increase on the score and reduction on error along the sessions suggests that the protocol is a successful training for motor control; although the slight increment on the error and decrement on the score pre and post–training intra–session, not being significant though, may suggest that there is some attention decrease or fatigue.

Near future work includes enrollment of more subjects to statistically prove the validity of the platform and protocol for the training of motor control. Patients will be also enrolled to check the possibility of translating the platform to the clinical environment.