Advanced User Interfaces For Upper Limb Functional Electrical Stimulation

Spinal cord injury (SCI) paralyzes approximately 12,000 people each year in the United States. Individuals with an injury at and above the sixth cervical vertebra (C6) lose function in the upper and lower limbs. To provide greater independence to this population, the restoration of reaching and grasping movements is critically important. Functional electrical stimulation (FES) is currently the only clinical approach for reanimating paralyzed muscles. While it has been used with great success in this population, especially for the restoration of hand grasp, the user interfaces currently available are not sufficient to control high-dimensional, dexterous movements of the hand and arm. Some neuroprostheses use purely logic-based command signals to switch between preprogrammed grasp patterns, while others use logic to select features such as grasp type in combination with proportional control of hand opening and closing. Furthermore, many of these approaches rely on unnatural actions not directly related to the desired movement and place considerable cognitive burden on the user. There is a need to develop "effortless" or low-cognitive-burden interfaces that allow users to control FES systems in a natural manner. Recent advances in physiological recording technologies and signal processing may provide solutions to this problem.The cortical areas that are involved in the normal generation of movement commands are still intact in SCI patients and may provide a more natural user interface. By recording directly from the cortex it may be possible for a user to control many degrees of freedom of the hand and arm using methods similar to those that are already enabling paralyzed patients to use computer interfaces and perform robot control tasks.A further way to improve control may to combine information from multiple signal sources depending on the needs of the individual patient: the brain, residual movements, electromyograms (EMGs), and gaze direction. In the context of Bayesian statistics, an optimal combination of the user's control signals with prior knowledge of the probabilistic nature of the desired movement can be formulated. For example, if we want to predict a reaching movement, we might have some knowledge of the likely reach target. Since the reach trajectory is generally stereotypical, the reach target informs us about the likely trajectory of the arm. By using this information to improve our estimate of the intended movement, less input may be required from the user.This chapter reviews currently available interfaces and addresses aspects of these promising new technologies with a view to improving control of reaching and hand grasp for FES neuroprostheses. We focus on the development of more natural user interfaces that may enable more effortless control, while recognizing that the needs and abilities vary widely between individual patients.