Dec. 14, 2004 -- Editor's Note: A noninvasive brain-computer interface (BCI) can allow people who are totally paralyzed to communicate by operating a computer cursor, according to the results of a study published in the Dec. 21 issue of the Proceedings of the National Academy of Sciences and posted online Dec. 7.
Earlier BCI studies involved implantable sensors, but this study used external scalp electrodes to record EEG signals from four individuals, and an adaptive algorithm was used to identify and
focus on the electroencephalographic features that the person was best able to control. After several training sessions facilitating further improvement in that control, participants were able to achieve multidimensional point-to-point movement control falling within the range of that reported with invasive methods in monkeys.
Movement time, precision, and accuracy were similar to those reported with invasive BCIs, suggesting that people with severe motor disabilities could use brain signals to operate a robotic arm or a neuroprosthesis without the need for electrodes implanted in their brains. Interestingly, the two participants with spinal cord injuries confining them to wheelchairs appeared to master the technique more quickly than the two nondisabled participants.
To learn more about the implications of this study and potential applications for this BCI system, Medscape's Laurie Barclay interviewed lead author Jonathan R. Wolpaw, MD, chief of the Laboratory of Nervous System Disorders at the Wadsworth Center of the New York State Department of Health and the State University of New York in Albany.
Medscape: Please describe the BCI device and how it works.
Dr. Wolpaw: This is a noninvasive brain-computer interface. We record EEG from the scalp using electrodes very similar to regular EEG electrodes. We analyze the signals in real time, and we derive certain features that are then converted into movement of a cursor on a screen. This is a skill that people develop over a period of training. We look at features that tend to be similar across people, but that are not exactly the same across people. The program and the analysis are adjusted to some extent for each person, and the software adjusts itself continually to optimize each person's performance. This is the result of an adaptive or a learning interaction between the person who's learning to control the process and the system that's learning to interpret the EEG signals to produce cursor movement. So this is very definitely a skill; this is not mind reading or looking in on thoughts.
Medscape: How does the software distinguish the pertinent signals from extraneous EEG signals not related to the intended movement?
Dr. Wolpaw: We look at the EEG from the scalp, and using a variety of filtering and analysis methods, we focus on particular features of the EEG signal coming from particular parts of the cortex, parts that are most directly involved in sensation and movement, the sensorimotor cortex. People initially use various kinds of motor imagery to change the signals, that is, to make a particular rhythm larger or smaller. They do this by trial and error. The amplitude of a rhythm is translated into movement of a cursor, initially in one dimension, to hit targets at the top or bottom of a screen. They learn what kind of movement imagery will move the cursor in one direction and what will move it in the other direction.
Medscape: Is it similar to biofeedback but more complex?
Dr. Wolpaw: [Patients] receive feedback on the results of whatever motor imagery they're using. Some people can think about moving or not moving; they can think about running or even shooting baskets. Just by trial and error, they determine what will move the cursor in one direction and what will move it in another direction. They usually get better as they do it more, and control develops. Particularly when people advance to two-dimensional control, it tends to become a more natural, less effortful activity. At that point, people often can't tell you exactly what they're doing. They do it the way you raise your arm — you can't really tell people how you raise your arm. BCI use becomes more like a normal motor skill, except that it doesn't involve muscles.
Medscape: How sensitive and specific is the control that develops, and how long does the training usually take?
Dr. Wolpaw: Most people develop basic control in a few initial sessions. Each session consists of 24 minutes of practice, and people generally do two or three sessions a week. Over several weeks most get some initial control. But to get to two dimensions it takes longer. The people in this study had at least 22 sessions of two-dimensional training, so it takes a few months for two-dimensional control. It develops slowly, like any complex skill.
Medscape: Is two-dimensional control sufficient to operate a keyboard, for example?
Dr. Wolpaw: It should be -- the better the control is, the faster that operation can be. The factors are how fast can you move the cursor, how accurately can you move, and how often do you make a mistake. If you make mistakes, then you have to be able to erase those mistakes, and that slows things down. You can do word processing, but the speed of the word processing depends on how good your control is. You can even do word processing with one-dimensional control, but it's going to be relatively slow.
Medscape: In terms of practical applications, are people actually using this to communicate?
Dr. Wolpaw: We collaborate with Drs. Birbaumer and Kuebler's group in Germany. They have been working for a long time with people who are extremely disabled — people with Lou Gehrig's disease or brainstem stroke who may be nearly totally locked into their bodies. We're trying to provide these people with very simple communication capability — the ability to say yes and no or do very slow word processing. This group published a paper a number of years ago showing a letter that one of these individuals wrote using a [BCI]. It took him a long time to do it, but he was able to do it eventually, and he was very pleased.
At this point, BCI technology has been applied in a very limited, largely research fashion in a small number of patients. There are other people engaged in similar work. Dr. Pfurtscheller in Austria is working with someone who's using a neuroprosthesis to open and close his hand. That's a very simple kind of control: on/off; open/close, so it's being used in relatively simple ways in a small number of people at present.
Medscape: How many patients would you estimate have tried this in one fashion or another?
Dr. Wolpaw: We've had quite a few people with various levels of disability come into our lab, and they've studied quite a number in Germany. It would probably be somewhere close to 100 in one form or another. Some of these have participated very briefly, for just a few sessions, while some have participated much longer.
Medscape: If patients want to use this at home, how expensive is it?
Dr. Wolpaw: The equipment cost isn't the problem. The hardware is not that expensive; it's basically PC hardware and standard EEG amplifiers. The software we give away for research purposes. Our new general-purpose BCI software, BCI2000, has already been given to about 30 labs around the world for research. But the problem is the expert oversight that's needed, both initially and on a continuing basis. That's really the sticking point in terms of applying this more widely — BCI technology has to be made much less labor-intensive on the part of the technicians and the scientists.
A lot more of the capability has to be incorporated into the software, so that a reasonably informed person could do this either for him or herself, or help someone else to do it. It's not really at that point yet. There's a lot of oversight that's necessary; a lot of adjustments in the system; a lot of troubleshooting. All that is very expensive and very time consuming, so that's the primary limitation right now.
We have a research program at the Wadsworth Center, and we work closely with the Birbaumer/Kuebler group in Germany, but there are relatively small numbers of people involved. There are groups elsewhere, but there's nothing like the kind of personnel that would be necessary to support widespread applications to this point. The technology really has to be developed to the point where it doesn't require as much oversight and maintenance and adjustment as it does now. We're moving in this direction; we're already starting to make it less dependent on oversight and trying to move it more out of the lab, but that's a long process.
Medscape: In addition to the word processing and communication capability, can patients achieve environmental control using this system?
Dr. Wolpaw: Yes -- once you can control a cursor, you can do anything that can be done with a computer. So you can do environmental controls, you can play games, and there are simple programs they developed in Germany for surfing the Internet.
Medscape: Are there potential ethical issues?
Dr. Wolpaw: Not really; this is not mind reading, so there are no privacy issues. This is skill development, so it's not something that can be done to a brain passively. You can't use our methods to listen in on the brain. They require active involvement and they initially require a good deal of concentration, so it's a skill, it's definitely not mind reading.
Medscape: Because this interface is not bidirectional, is it correct to assume there are also no "mind-control" issues?
Dr. Wolpaw: Stimulating electrodes are implanted in the brain for Parkinson's disease and related disorders, but that's for totally different, very specific therapeutic purposes. We're not involved in that. In any case, I don't believe that those endeavors or our own involve anything that could be called "mind control."
Medscape: Is there anything you would like to add?
Dr. Wolpaw: Until recently it has been widely assumed that noninvasive BCIs were only useful for very simple applications — slow word processing, indicating yes or no, very simple environmental control, and things like that. The assumption was that if you were going to get control of complex movements, of a neuroprosthesis or a robotic arm, for example, or full cursor control and mouse selection, you would need to have implants in the brain. What we're showing here is that this may not be necessary, that this assumption is not correct. Noninvasive methods appear to be a good deal better than people gave them credit for. So in many situations, it may not be necessary to actually stick something into the brain to get pretty complex control.
Finally, in regard to the potential practical value of BCI research in general, I think it's very important to point out that recent studies show that people who are severely paralyzed, even nearly locked in, can lead lives that they consider enjoyable if they have reasonable physical and social environments and are well cared for. Recent studies show they are not much more likely to be depressed than people without disabilities. This reality gives additional impetus to efforts to provide these patients with BCI communication and control technology.
Furthermore, physicians caring for people, with ALS for example, who are approaching total paralysis, need to consider these new data when they advise patients concerning accepting artificial ventilation and other life-support technology.
Disclosures: None were reported by Dr. Wolpaw. This study was supported in part by the National Institutes of Health (National Center for Medical Rehabilitation Research of the National Institute of Child Health and Human Development, National Institute of Biomedical Imaging and Bioengineering, and the National Institute of Neurological Disorders and Stroke) and the James S. McDonnell Foundation.
Proc Natl Acad Science USA. 2004;101:17849-17854
Reviewed by Gary D. Vogin, MD
Medscape Medical News © 2004 Medscape
Cite this: Laurie Barclay. Brain-Computer Interface Allows "Locked-In" Patients to Communicate via Computer - Medscape - Dec 14, 2004.