Hand Articulating Neuro-Training Device (HAND).

A-B. Participants’ hand fitting in the device, posture recorded with mounting distance and angle. C. Illustration of finger joint motion to Cartesian coordinates in the virtual 3D space.

The 3D finger individuation paradigm.

A-B. Natural trajectory task. A. Screenshot of the task for one finger: participant control a white dot by exerting isometric forces towards one of the 6 directions in the 3D virtual space and move the dot between the home position (gray sphere) and a virtual wall. B. Example force traces recorded from index finger during natural trajectory task in a healthy participant and a stroke patient’s paretic hand; C-D. Finger individuation task. C. Screenshot of the task for one finger: participant control a white dot in the virtual space and try to hit a target by following the specified path (thick black line) estimated from that finger’s natural trajectory (shown in B) while attempting to minimize forces from uninstructed fingers (red bar). D. Example force traces recorded from all five fingers in a healthy participant’s left hand and a stroke survivor’s paretic hand during the individuation task when the participants attempted to move their left index finger towards a target (red dot).

Patient characteristics.

Data indicate patients’ gender (M, male; F, female), age (years), time since stroke (in months), handedness (L, left; R, right), affected side (L, left; R, right), Montreal Cognitive Assessment (MoCA, maximum 30), Fugl-Meyer Assessment for upper extremity (FMA, maximum 66), and Action Research Arm Test (ARAT), and grip strength (in pounds) if applicable.

Individuation Index derivation and results.

A. Illustration of derivation of Individuation Index: overall net force trajectories (first two panels) and the function of mean deviation net force from the uninstructed fingers as a function of the force in the instructed finger towards the instructed direction (3rd panel). Individuation index is -log(slope) of the regression line of the function; B. Individuation Indices of healthy and stroke patients in cartesian spaces; C. Individuation Indices summarized in joint space for healthy controls and patients; D. Reduction of Individuation Indices (paretic subtracted from non-paretic) for patients with mild (FMA>=40) and severe (FMA<40) impairment.

Pattern similarity analysis of finger coactivation/enslavement patterns using cosine distances. A larger distance indicates distinct shapes of finger coactivation patterns.

A. An illustration of representation similarity matrix (RDM) computed across coactivation/enslavement patterns for one instructed finger (finger 2) exerting force in two different directions: +X vs. +Y. B. Mean RDMs for each instructed finger direction averaged across all non-paretic and paretic hands. C. Direct comparison of mean distance values for each participant at each instructed finger. D. An illustration of RDM computed across coactivation/enslavement patterns for two different fingers (finger 3 & 4) exerting force in the same target direction (+X). E. Mean RDMs for each target direction averaged across all non-paretic and paretic hands. F. Direct comparison of mean distance values for each participant at each instructed target direction.

Pattern similarity analysis of the finger coactivation/enslavement patterns using Cosine distances with mildly (FMA>=40) and severely (FMA<40) impaired patients separated.

A. Mean RDMs for each instructed finger direction averaged across all non-paretic and paretic hands. B. Sum of mean distance values across all participants for each instructed finger. C. Mean RDMs for each target direction averaged across all non-paretic and paretic hands. D. Sum of mean distance values across all participants for each instructed target direction.

Pattern similarity analysis of the finger coactivation/enslavement patterns using Euclidean distances.

A. Mean RDMs for each instructed finger direction averaged across all non-paretic and paretic hands. B. Direct comparison of mean distance values for each participant at each instructed finger. C. Mean RDMs for each target direction averaged across all non-paretic and paretic hands. D. Direct comparison of mean distance values for each participant at each instructed target direction.

Pattern similarity analysis of finger coactivation/enslavement patterns using Euclidean distances, with mildly (FMA>=40) and severely (FMA<40) impaired patients separated.

A. Mean RDMs for each instructed finger direction averaged across all non-paretic and paretic hands. B. Sum of mean distance values across all participants for each instructed finger. C. Mean RDMs for each target direction averaged across all non-paretic and paretic hands. D. Sum of mean distance values across all participants for each instructed target direction.

Bias measures.

A. Mean biases for each subject for all fingers plotted in the same 3D space. Paretic data are separated by severity: (FMA>=40 vs. (FMA<40)). B. Bias summary in the flexion (-Y and -Z), extension (+Y and +Z), and abduction/adduction (+X/-X) directions for the three groups.

Scatter plots of Bias Difference between paretic and non-paretic (intrusion of flexor bias) and distance measures. A-B. Cosine Distances by Finger and Target Direction, respectively, and C-D. Euclidean Distances by Finger and Target Direction, respectively. r-values are Pearson correlation coefficients.

PCA analysis results.

Variance explained by the number of PCs in finger coactivation/enslavement patterns for the healthy, non-paretic, and paretic hands. Note that there are a maximum of 15 PCs due to the nature of the individuation task.