Key components of mechanical work predict outcomes in robotic stroke therapy

Abstract

Background: Clinical practice typically emphasizes active involvement during therapy. However, traditional approaches can offer only general guidance on the form of involvement that would be most helpful to recovery. Beyond assisting movement, robots allow comprehensive methods for measuring practice behaviors, including the energetic input of the learner. Using data from our previous study of robot-assisted therapy, we examined how separate components of mechanical work contribute to predicting training outcomes.

Methods: Stroke survivors (n = 11) completed six sessions in two-weeks of upper extremity motor exploration (self-directed movement practice) training with customized forces, while a control group (n = 11) trained without assistance. We employed multiple regression analysis to predict patient outcomes with computed mechanical work as independent variables, including separate features for elbow versus shoulder joints, positive (concentric) and negative (eccentric), flexion and extension.

Results: Our analysis showed that increases in total mechanical work during therapy were positively correlated with our final outcome metric, velocity range. Further analysis revealed that greater amounts of negative work at the shoulder and positive work at the elbow as the most important predictors of recovery (using cross-validated regression, R2 = 52%). However, the work features were likely mutually correlated, suggesting a prediction model that first removed shared variance (using PCA, R2 = 65-85%).

Conclusions: These results support robotic training for stroke survivors that increases energetic activity in eccentric shoulder and concentric elbow actions.

Trial registration: ClinicalTrials.gov, Identifier: NCT02570256. Registered 7 October 2015 - Retrospectively registered.

Keywords: Energetics; Neurorehabilitation; Outcomes; Robotic therapy; Stroke; Upper limb.

Conflict of interest statement

The authors declare that they have no competing interests.

Figures

Fig. 1

Experimental design. a Stroke survivors performed self-directed motor exploration by moving the robot handle in the horizontal plane. Measurements of their limb motion and the interaction forces were used to estimate the positive (concentric) and negative (eccentric) mechanical work exerted in different directions of shoulder and elbow joint motion. b The probability distribution of each individual's movement velocities during unassisted motor exploration (top; blue indicates lower probability, red indicates higher probability, black contour line represents the 90th percentile velocity coverage) formed the basis for the design of customized training forces (bottom; red arrows indicate the direction and relative magnitude of forces applied, colored contour lines represents Gaussian model fit to velocity data)

Fig. 2

Correlation analysis. a The total mechanical work performed during motor exploration force training significantly correlated with changes in velocity coverage. Each data point represents an individual participant. The size of each data point is proportional to that participant’s velocity coverage during Baseline (session 2). For the Force group (black closed circles), participants with greater initial velocity coverage (evident of larger data points) tend to exert more total work during Training and showed greater gains in velocity coverage. This trend was not observed in the Control group (black open circles). b The breakdown of work reveals subcomponents that significantly correlated with changes in velocity coverage. A single pair of open (Control group) or closed (Force group) red triangle (Positive Work) and blue circle (Negative Work) along the x axis (the Training axis on each plot) represents an individual participant. Regression lines only shown for statistically significant correlation (α < 0.05) observed in the Force group

Fig. 3

Model predictions and feature selection. Predictions of patient recovery, in terms of changes in velocity coverage, using multiple regression analysis. Each gray dot represents a single repeat of a cross-validation staggered for easy visualization by fitting a probability density function. Each black dot and bar represents the mean R2 ± SD. Positive (concentric) and negative (eccentric) work features are indicated in red and blue, respectively. The negative work in shoulder adduction and positive work in elbow flexion and extension features were selected most often by the LASSO model across the cross-validation repeats. The successive removal of the four most selected features resulted in a diminishing return of model accuracy. The full model equation is represented as y = [4.85A + 1.46B + 2.75C - 0.41D + 0.16E + 0.09F + 211.0] × 10− 3, where model coefficients assigned to each feature were averaged across cross-validation repeats