Free
Research Platform
Issue Date: July 2015
Published Online: July 01, 2015
Updated: April 30, 2020
Automatic Assessment of Upper-Limb Function for Remote Monitoring During Home-Based Rehabilitation Using Videogames
Article Information
Hand and Upper Extremity / Neurologic Conditions / Rehabilitation, Participation, and Disability / Stroke / Assessment/Measurement
Research Platform   |   July 01, 2015
Automatic Assessment of Upper-Limb Function for Remote Monitoring During Home-Based Rehabilitation Using Videogames
American Journal of Occupational Therapy, July 2015, Vol. 69, 6911500088. https://doi.org/10.5014/ajot.2015.69S1-RP205C
American Journal of Occupational Therapy, July 2015, Vol. 69, 6911500088. https://doi.org/10.5014/ajot.2015.69S1-RP205C
Abstract

Date Presented 4/17/2015

Automatic in-game assessment of clinically validated, upper-limb function in stroke patients—generated using low-cost, commodity hardware and rehabilitation videogames—demonstrates the potential for remote monitoring of patients during videogame-delivered, home-based rehabilitation programs.

SIGNIFICANCE: Action videogames have been shown to provide structured home-based, upper-limb rehabilitation, but for implementation, there is also a requirement for clinically valid, remote monitoring by therapists. In this study, we attempted to answer the following research question: Can automatic assessments of upper-limb function be made during home-based rehabilitation using therapeutic videogames and a cloud-based analytical system that are clinically valid and sensitive to change?
METHOD: A longitudinal study was undertaken. Patients with hemiplegia after stroke played the rehabilitation game Circus Challenge daily, using commercial controllers that provide continuous three-dimensional (3D) data of upper-limb position and orientation. To derive the model, occupational therapists made eight blinded, clinical assessments per subject using the Chedoke Arm and Hand Activity Inventory (CAHAI) over a 12-wk period. The patients played the rehabilitation game in their homes, with data from game play relayed to a cloud-based analytical system. Clinical CAHAI assessments were undertaken at their home.
Ethical approval and written informed consent were obtained. A total of 40 patients (aged 33 to 81 yr) with hemiplegia after stroke (20 chronic: 36 to 414 wk poststroke; 20 acute: 1 to 6 wk poststroke) participated. Algorithms derived during game play were used to predict clinically assessed CAHAI scores that are valid and sensitive to change in upper limb-function. Kinematic variables obtained during gameplay on the same day that clinical assessments were made were used to derive linear models for CAHAI prediction for chronic and acute patients respectively, using the fitted R2 as the selection criterion. Model validity was examined with K-fold cross-validation. Cross-sectional validity was examined with the between-subjects correlation coefficient, and longitudinal validity was examined with the within-subjects correlation coefficient. Sensitivity to change was examined by comparing receiver operating characteristic (ROC) curves in classification of patients to acute and chronic groups for clinically assessed and model-derived CAHAI scores.
RESULTS: The models derived for the acute and for the chronic stroke patients use 14 covariates (two in common) and a baseline clinically assessed CAHAI. The models account for 96% (chronic) and 87% (acute) of the variability in clinically assessed CAHAIs. K-fold cross-validation gave root-mean errors of 3.64 (chronic) and 5.80 (acute): cross-sectional validity (r = .998, p < .001; chronic, r = .99; acute, r = .99) and longitudinal validity (r = .54, p < .001; chronic, r = .33, p < .001; acute, r = .63, p < .001). There was no difference between the model-derived and clinically assessed CAHAIs in classification of patients to acute and chronic groups (comparing ROC curves, p = .50), demonstrating similar sensitivity to change.
CONCLUSION: Automatic in-game assessment of upper-limb function—generated solely with low-cost, commodity hardware and action videogames and a cloud-based analytical system—demonstrates the potential for clinically valid, remote monitoring of patients during action videogame-delivered, home-based, upper-limb rehabilitation programs.