Body-Machine Interface for Recovering Muscle Control

When suffering from conditions affecting the central nervous system, such as spinal cord injury (SCI), stroke or neurodegenerative disorders, two pathways are available for regaining independence and quality of life. One way is through clinical interventions, including therapeutic exercises, often in combination with pharmacological agents. The other is provided by assistive technologies, such as wheelchairs or robotic systems. These two approaches have conflicting characteristics. While rehabilitation exercises challenge patients to use the most affected parts of their musculoskeletal apparatus, assistive technologies are typically designed to bypass the disability. This has led to divergent research domains. In both fields there are three major gaps that we plan to address in the investigator's research:

1. High cost of technology and the limited amount of available hospital-based rehabilitation;
2. Lack of adaptability of currently available assistive technologies, such as head switches and sip-and puff devices, that require users to overcome a hard learning barrier;
3. Inadequate criteria for assessment of effectiveness of therapy, with common techniques still relying on subjective approaches that are inadequate considering the current state of biomedical science and technology.

We will address all of these issues by developing a new generation of body-machine interfaces (BoMI) supporting both assistive and rehabilitative goals. BMIs will translate movement signals and muscle activities of the user into control signals for assistive devices and computer systems. State-of-the-art systems for surface electromyography (EMG) and movement recording (IMU) will be integrated through machine learning techniques to facilitate sensorimotor learning while providing the means to promote or reduce the use of targeted muscles. New comprehensive assessment techniques will be developed by integrating standard measure of function - as the manual muscle test - with EMG analysis and non-invasive magnetic brain stimulation (TMS) (Magstim 200 Bistim, Whitland, UK). The development will be organized in three specific aims.

AIM 1: To develop a BMI integrating muscle activities and motion signals for operating external devices and performing rehabilitation exercises. EMG signals derived from multiple muscles in the upper body (e.g. deltoid, pectoralis, trapezius, triceps, etc.) will be integrated with motion signals to generate control signals for external devices (e.g. the coordinates of a cursor on a computer monitor or the speed and direction commands to a powered wheelchair). Both linear (PCA) and nonlinear maps (auto encoder networks) will be explored, although current preliminary evidence suggests that non-linear auto encoders (AE) are likely to better facilitate user learning1.

AIM 2: To enable targeting and modulating recruitment of specific muscles and muscle synergies during the practice of games and functional tasks. To enhance or reduce the role of a muscle or synergy, the output of the BoMI will be modulated in proportion to the deviation of the measured muscle activity from the desired level. The effectiveness of the approach will be tested at different times following training, both by tracking of motions and EMG activities during the performance of selected activities of daily living (ADL) and trough the assessment of muscle responses evoked by non-invasive brain stimulation.