Funded Grants

A non-invasive brain-computer interface for prosthesis control

Many disorders disrupt the channels through which the brain normally communicates with and controls its external environment. Amyotrophic lateral sclerosis (ALS, or Lou Gehrig's disease), brainstem stroke, brain or spinal cord injury, cerebral palsy, muscular dystrophies, multiple sclerosis, and numerous other degenerative diseases impair the neural pathways that control muscles or impair the muscles themselves. These disorders affect nearly two million people in the United States, and far more around the world. Those most affected may lose all voluntary muscle control, including eye movements and respiration, and may become completely "locked-in" to their bodies, unable to communicate in any way whatsoever. Conventional assistive communication devices require some voluntary muscle control, and thus they are often useless for people in this condition. At the same time, modern life-support technology, such as home ventilators, can allow most individuals, even those who are locked-in, to live long lives, so that the personal, social, and economic burdens of their disabilities are prolonged and severe. The capacity to communicate is critically important, and the absence of this capacity greatly reduces quality of life. Indeed, people with ALS often elect to die when they lose the power to communicate or to exert any control over their environment.

Over the past 15 years, studies in many laboratories have shown that electroencephalographic activity (EEG, or brain waves) recorded from the scalp can provide new non-muscular communication and control channels. These brain-computer interfaces (BCIs) do not depend on voluntary muscle control, and thus they can be used by people who are severely paralyzed or even totally locked-in. Encouraged by new understanding of the brain, by the advent of powerful low-cost computers, and by growing recognition of the needs and potentials of people with disabilities, BCI research focuses on developing new assistive communication and control technology that allows severely disabled people to express their wishes to caregivers or even operate word processing programs or neuroprostheses. Science 299:496-499 (2003) provides an excellent short review of BCI research and development.

Present-day BCIs use a variety of brain signals. These signals are translated in real-time into commands that operate a computer display or other devices. The user learns to encode commands in these signals and the BCI decodes the commands from the signals. BCIs can be non-invasive (i.e., they use scalp-recorded EEG signals) or invasive (i.e., they use neuronal activity or field potentials recorded in or on the cortex). Non-invasive BCIs are commonly perceived to be suited only for simple communication, while invasive BCIs are thought to be applicable to more demanding applications such as real-time control of neuroprostheses or robotic devices. (A neuroprosthesis is a device that can stimulate the muscles of a paralyzed limb so that the person can use the limb again.) On the other hand, invasive BCIs face greater technical difficulties and entail substantial clinical risks.

The motivation for this project is a strong conviction that non-invasive BCIs can in fact provide real-time control of complex movements. This conviction has two sources. First, recent studies in our laboratory have improved non-invasive BCI methods to a point suggesting that, with continued improvement, non-invasive BCIs should be able to provide multidimensional real-time control of a robotic arm or other prosthetic device. Second, while prosthesis operation has been conceived of as a problem of process control (i.e., controlling the details of movement), it is more realistically conceived of as a problem of recognizing intent. This is certainly true on a conscious level: we think "I want to pick up that book" rather than "I want to activate these muscles at these strengths in this sequence so that my hand reaches out and picks up that book." Furthermore, due to the wellestablished distributed organization of brain function, the real-time details of movement control are handled largely at subcortical, brainstem, and spinal cord levels, rather than exclusively by the cortical neurons that are the focus of invasive BCIs. Thus, the development of invasive BCIs that vest detailed movement control in cortical neurons is not only technically difficult and clinically risky, but also artificial and simply unnecessary. It is more realistic and practical to focus BCI research on recognizing intent, and having downstream software manage the process that achieves that intent. Once intent is known, process control is a relatively straightforward engineering problem, whether it is control of a robotic arm or a neuroprosthesis. Recent studies show that non-invasive BCI methods can indeed convey intent rapidly and consistently.

This proposal's central hypothesis is that non-invasive BCIs can enable users to control threedimensional movements of a robotic arm (i.e., a prototype prosthesis) at speeds nearly comparable to normal movements. The goal is to show that normal volunteers and people with disabilities can master such control. The specific aims are: (1) to improve current scalp-recorded EEG-based BCI control to move a robotic arm rapidly and accurately in two dimensions; (2) to extend that control to rapid and accurate movement in three dimensions; (3) to further extend that control to functions such as moving an object from one place to another or pouring water into a cup; (4) to show that people with severe disabilities (e.g., locked in by late-stage ALS or brainstem stroke) can use this technology and that it improves their quality of life.

In accord with the Foundation's 21st Century Initiative, this project spans neural, cognitive, and behavioral levels of analysis and fits into the "Applied Research" category. It is interdisciplinary, involving neurophysiology, neuropsychology, signal analysis, engineering, and clinical rehabilitation. Our laboratory has expertise in all these areas, extensive experience in BCI research, and encouraging preliminary data. We are confident that, with the Foundation's support, we can achieve the specific aims and thereby greatly extend the capacities of non-invasive BCI technology to provide important new control capacities and improved quality of life to people with severe disabilities. This work should also yield new understanding of the interactions between conscious intent, as formatted in cortex and associated areas, and its execution by lower-level natural or artificial structures.