TASK-SPECIFIC VIRTUAL TRAINING FOR IMPROVED PATTERN RECOGNITION-BASED PROSTHESES CONTROL

virtual reality;prosthesis;motor learning;Biomedical Engineering

Lee, InhwaThakor, Nitish V. ;

Johns Hopkins University

关键词: virtual reality;    prosthesis;    motor learning;    Biomedical Engineering;   

Others  :  https://jscholarship.library.jhu.edu/bitstream/handle/1774.2/37246/LEE-THESIS-2014.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: JOHNS HOPKINS DSpace Repository

【 摘 要 】

The emergence of dexterous prostheses presents the potential to significantly improve amputees’ quality of life. The use of intuitive pattern recognition algorithm is among the most promising control strategy for dexterous prostheses, with the demonstration of near perfect classification accuracies in laboratory settings. However, recent literatures show a weak correlation between classification accuracy and usability of the prostheses. External factors such as varying limb positions affect electromyography signals and consequently deteriorate usability of the prostheses; therefore, task-specific user training is proposed to enhance usability of the pattern recognition-based prostheses. Eight able-bodied subjects and one transradial amputee subject participated in the study to validate the efficacy of task-specific virtual training and examine the relationship between the virtual reality and real-world environment performance of prostheses use. Subjects were evaluated in 2 functional tests, Modified Box and Block Test and Reach-Grasp-Release Test, in both virtual and real-world environment, and received five sessions of one-on-one virtual training that lasted for one hour. Subjects were evaluated once again after completing five virtual training sessions and showed a significant improvement in functional tests. The amputee subject, despite the fact that he had been a pattern recognition- based prosthesis wearer for 5 months, also showed improvement upon virtual training, especially in the test that enforced him to use his prosthesis in postures that are outside of his usual range. In addition, no statistically significant difference was observed between the performance in virtual reality and real-world environment, indicating the potential for virtual reality evaluation to be a diagnostic tool to determine individual’s usability of pattern recognition-based myoelectric prostheses. It was shown that high classification accuracy alone does not guarantee proficiency in prostheses control; rather, it only represented the capacity of one’s prostheses control. To effectively prepare amputees for pattern recognition-based myoelectric prostheses control in activities of daily living, task-specific virtual training should be administered prior to prosthesis fitting. For future study, the integration of accurate, stable motion tracking system with head-mounted display is suggested for more immersive experience that enables users to practice proper positioning of the terminal device, an essential skill for object interaction with prostheses.