Fixing the Brain-Computer Interface
Researchers are addressing the sizable population for whom BCI doesn't work
Photo: East Tennessee State University Brain-Computer Interface Laboratory
16 June 2011—Brain-computer interface (BCI) technology, which allows users to perform basic computer tasks through brain activity alone, has quickly become one of the most promising areas of neuroscience. Researchers have focused primarily on clinical applications, such as communication programs for patients with amyotrophic lateral sclerosis (ALS) and other medical conditions that prohibit speaking and writing, and BCI technology clearly has potential for both gaming and telecommunications as well.
But BCIs have a big problem. For reasons that are not entirely clear, as much as 20 to 30 percent of people who try BCI systems can’t get them to work. Until recently, scientists simply excluded the "nonperformers" from their studies. This approach made sense when the technology was new—researchers had to establish how BCI works, and including nonperformers would have skewed the results. But now that the proof-of-concept phase is over, a growing number of researchers are beginning to tackle the usability issue.
There are several types of BCI systems, each with its own usability problems, but the ones most likely to be widely used outside of the clinic employ electroencephalography (EEG), a noninvasive technology that measures the brain’s electrical activity through the scalp. Within this category, there are two main systems: the P300 systems and the sensorimotor rhythm systems. The P300 systems are so named because a wave of voltage usually occurs within 300 milliseconds of a perceived event. These systems automatically convert EEG recordings into letters on a virtual keyboard.
The sensorimotor rhythm, or motor imagery, systems allow users to control cursor movement with their brains. In this system, the electrodes are attached to the scalp above the sensorimotor cortex, the brain region involved with perceiving sensory information and planning voluntary movements.
Researchers are finding that with the P300 BCI systems, user success largely depends on external factors, such as fatigue. For example, a study at East Tennessee State University, published earlier this spring, found that users performed better with the P300 when they completed "mindfulness" exercises for 6 minutes beforehand. Psychological factors can also play a role, says Eric Sellers, the lead researcher involved in the East Tennessee study.
Sellers recalls one control subject who performed well on the P300 only when she was being watched. "She was only performing with 30 to 50 percent accuracy when she was alone," he says. "But when I stayed in the room to observe, her accuracy suddenly shot up to 85 to 90 percent." A 2010 study by researchers at the University of Tübingen, in Germany, found that monetary rewards can also help boost P300 performance.
Conversely, the usability problems associated with motor imagery BCI systems appear to be more physical in nature. These systems may offer more possibilities in terms of communication and other applications, but they also require more effort from the user and are more difficult to master. Some users can produce the requisite EEG signals to calibrate the BCI device but are unable to control the cursor; others can’t even get past the calibration stage.
Last year, a group of German scientists set out to address both types of user problems with a coadaptive motor imagery BCI system. They extended the amount of time that subjects received feedback from the device to increase the probability that their imagined movements would match up with an EEG pattern the computer could understand. The researchers then made the computer algorithm more flexible to make it easier for the device to find a usable EEG pattern. "This way, the BCI system learns from the user’s EEG, and the user learns from the feedback he or she receives once they start moving the cursor," says Carmen Vidaurre, a computer scientist at the Berlin Institute of Technology and one of the researchers involved in the project.
In a study published in the Journal of Neural Engineering in April, Vidaurre and her colleagues tested their coadaptive motor imagery BCI system on 14 participants who had previously been unable to achieve cursor control and got it to work for 10 of them within 15 minutes. Still, the system had no effect on two of the participants, and it’s unclear if more training would have helped them, says Vidaurre. "Everything depends on the performance a person can achieve, and there are undoubtedly situations where motor imagery BCI does not make sense," she says.
But even if Vidaurre, Sellers, and the other researchers in this field can improve BCI usability by only a few percentage points, it could be useful for clinical applications, given the limited options for patients who are physically unable to communicate. What’s more, their research is providing neuroscientists with new information about how the human brain works. Indeed, some researchers have already started using BCI performance as a measure of cognitive function, implying that the technology could also have a future as a research tool.
About the Author
Erica Westly is a freelance science writer based in Brooklyn, N.Y. In July 2010, she wrote for IEEE Spectrum about how engineers are using technology to demystify the black art of barbecue.