Applying LDA-based pattern recognition to predict isometric shoulder and elbow torque generation in individuals with chronic stroke with moderate to severe motor impairment.

Only 29 of 35 recruited participants were analysed, and six were excluded for corrupted data or profound external rotation weakness, so results do not apply to people who cannot produce external rotation torque. The task was maximum isometric torque in a supported, casted arm, not functional reaching or a real-time wearable device. Classification was tested with recorded data and trial-wise cross-validation, so it does not show how a device would perform during therapy. EMG was recorded from eight major upper-extremity muscles but not rotator-cuff muscles, which the authors say may have limited accuracy. Some participants were very inaccurate in the four most confused directions, including internal rotation as low as 3%. No significant correlation was found between accuracy and Fugl-Meyer score or reaching distance, so the study does not explain which patients would be best classified.

The supplied publication types list NIH extramural research support. No author conflict-of-interest declaration is provided.

Do not conclude that a wearable shoulder device can now be controlled during daily reaching. This was a test of recorded isometric torque data in a supported, casted arm, and only 20 of 29 participants were classified above 90% by the load cell classifier.