This proposal investigates the use of Brain-Computer-Interfaces (BCIs) as a command interface for Functional Electrical Stimulation (FES) neuroprosthetic systems. In the absence of efferent cortical command signals, FES applies spatially and temporally stereotyped patterns of electrical activation to peripheral nerves to achieve movement restoration in spinal cord injured (SCI) persons. Upper-extremity FES systems typically require SCI persons to retain some volitional movement such as contraction of muscles or movement of joints above the line of injury. As a result, persons with high cervical spinal cord injury, resulting in complete tetraplega, are not able to control current FES systems. BCIs, however, may make it possible to give persons with high cervical SCI a natural and volitionally controlled FES command interface for regaining functional control of the arm and hand. The role of the brain in controlling arm and hand movements has been extensively studied in able-bodied non-human primates, but less so in humans with prolonged and severe paralysis. The central hypothesis of this proposal is that cortical signals in humans with impaired movement reliably encode specific parameters of reaching-to-grasp arm movements and these signals can be reliably decoded to provide real-time control of a functional reaching-to-grasp movement in an FES neuroprosthetic system.
The aims of the proposal will investigate this hypothesis by examining three components of a typical reaching-to-grasp movement.
Aim one develops models for discriminating imagined and performed functional hand grasps based upon the neural activity patterns of primary motor cortex (M1). These models will discern which cortical frequency bands are most useful for hand grasp discrimination. Electrocorticography (ECoG) signals will be recorded from the cortical surface of M1 while human subjects imagine and perform a series of hand grasps. Offline discriminant analysis (DA) models will assess how well the neural activities of these imagined and performed hand postures can be differentiated. Subjects will then attempt to use these DA models in a closed-loop task to cortically control the grasp postures of a virtual hand. The number of correct postures selected, as well as the time taken to select the correct posture will be used as metrics to assess subjects' cortical control capabilities.
Aim two examines how control of grasp force is encoded in cortical signals. Cortical activity will be recorded from persons with previously implanted ECoG recording arrays while they perform a force tracking task using power and lateral pinch hand grasps. Cortical activity will be recorded as the grasp force is varied between 10% and 50% of maximal voluntary grasp (MVG) force. The continuous grasp force will be related to the cortical activity through linear decoding methods such as multiple-input-single- output System Identification (SID) or Kalman Filtering (KF), or if necessary, nonlinear decoding methods such as time-delayed artificial neural networks. Subjects will then use the optimized decoding models to control virtual grasp force directly using brain signals in a force target reaching task. Success of control will be measured by the number of successful target acquisitions, as well as the time necessary to hit each target.
Aim three investigates cortical control of arm and hand positioning and grasping by paralyzed persons chronically implanted with a microelectrode array in M1 as part of the BrainGate2 Clinical Trial. The relationship of single unit activity and power in local field potential frequeny bands to kinematics of observed arm reaching movements will be characterized in the effort to build neural decoders, based upon SID or KF techniques. Participants will use these decoders to control virtual arm reaching in a 3D center-out-center task, in intrinsic (joint) and extrinsi (global) coordinate frames. Metrics to measure success of control will include time to target, number of targets acquired, and information transfer rate. Successful completion of these three aims will lead to a greater understanding of the neural representation of reaching-to-grasp, and how to best decode these arm movement components for controlling an upper-extremity FES neuroprosthesis.
Although about 11% of Americans are Veterans, they comprise almost 20% of all individuals with spinal cord injuries. Individuals who have experienced high cervical spinal cord injury, or other severely paralyzing impairments, such as amyotrophic lateral sclerosis (ALS), have few options for regaining the independence needed to perform basic activities of daily living. This is in large part because current assistive devices require them to be able to volitionally generate standard independent command sources for control. These individuals do not have that ability. However, they usually retain their cognitive ability and mental faculties. Hence, it would be highly beneficial to this population to be able to use these retained brain signals as control inputs for their assistive devices. This is the overarching motivation for the research described i this proposal. Advancements towards this end resulting from the proposed work could significantly improve the quality of life and care of Veterans that is expected after one experiences an impairment resulting in severe paralysis.
