Estimating Maximal Effort Movement during Rehabilitative Training of Humans and Rodents.

摘要

The success of motor training in health and following neurologic injuries such as stroke and spinal cord injury depends on the level of challenge presented to the learner during training. Movement rehabilitation is accomplished with wearable sensors, virtual environments, and/or robotic training devices that allow patients to interact with computer games during training rather than requiring continuous, direct supervision from a therapist. However, few algorithms have been developed for automatically measuring the capability of a trainee and then controlling the level of challenge experienced during motor training based on this capability. This dissertation developed and tested three devices and two algorithms for estimating maximal effort movement capability and challenging the trainee during computer-based rehabilitation training.;The first algorithm addressed the problem of determining movement capability when the performance of the trainee at the rehabilitation game is unknown. The algorithm developed was derived from an optimization framework that used only motion sensor data recorded during game play. In testing with 15 individuals with a stroke, this algorithm accurately estimated each subject's maximum acceleration capability as they played a Wii-like game with the Wimplifier, a Wiimote like acceleration-sensing gripper developed for this project. We extended the algorithm for use in two degrees of freedom with the Gesture Therapy 2.0, a computer vision system for reach and grasp exercise that was developed as part of this dissertation.;The second calibration algorithm assumed availability of the trainee's success or failure at the rehabilitation task, but the unique problem here was that it was impossible to instruct the trainee to exert a maximal effort, since the trainees were rats, an important scientific model for neural recovery research. To train and estimate the strength of the animals we implemented a rat robotic grip strength training device. In testing with 15 rodents over 10 months, we found that the algorithm converged to a reliable strength estimate in each session. We also found that the maximal strength estimated through this device was significantly greater than, and uncorrelated with, the Grip Strength Meter; we hypothesize that this is because the new technique better motivates the animal.