The BRADLEY DEPARTMENT of ELECTRICAL and COMPUTER ENGINEERING

Objectively detecting ADHD | ECE | Virginia Tech

ECE NEWS

Objectively Detecting ADHD

An undergraduate researcher wearing EEG electrodes looks up from a computer.
In a mock session, undergraduate researcher Sarah Hanson (BSEE '17) tests out the attention-focused computer task that had been administered to the children participating in the longitudinal study. "We use some of our experimental data to train our machine-learning algorithms, and other data to test it," she said.

Despite the growing number of children and adolescents identified as having Attention Deficit Hyperactivity Disorder (ADHD), there is no objective diagnosis protocol. Current methods depend heavily on subjective observations by parents, teachers, and physicians.

"It would be very useful to provide a method of hands-off diagnosing," said ECE Professor Louis Beex. So he's building one.

Beex has spent the last 15 years researching and developing a method to diagnose ADHD based on data alone, which may not even require a physician to be present.

Two researchers watch EEG data in a darkened room, while their subject can be seen looking at a computer monitor in the next room.
In the Cognition, Affect, and Psychophysiology Lab directed by Professor Martha Ann Bell, psychology doctoral student Tashauna Blankenship attached the electrodes and monitored Hanson's brain waves through the EEG.

Current ADHD diagnoses involve extensive interviewing of the children in question, their parents, and their teachers. The process is time-consuming and can be unreliable. A more reliable diagnosis may finally involve functional magnetic resonance imaging (fMRI), but that method is expensive, said Beex. And rightly or wrongly diagnosed, a child will still be affected by medications like Ritalin, which is commonly prescribed for ADHD.

"Recent literature has suggested that this medicine can have a long-term affect on the heart," he said.

An interdisciplinary collaboration

In collaboration with Professor Martha Ann Bell in the Department of Psychology, Beex began working on a new approach that involves using electroencephalogram (EEG) data to discern the brain wave patterns associated with ADHD.

During their first attempt, Beex and his students analyzed EEG data of children at rest, with a success rate of around 60 percent. "Not very good," he noted.

Now they are trying a different test. Beex and his team are currently working with more recent data collected by Bell and her students in the Cognition, Affect, and Psychophysiology (CAP) Lab. This EEG data was recorded while children, some of whom had already been diagnosed with ADHD, were actively engaged in computer tasks that required focus and motor skills.

Children between the ages of 6 and 9 visited the research lab as part of an ongoing longitudinal study in the CAP Lab. After the electrodes were placed on their heads, the children were prompted to follow simple instructions based on visual cues, while the EEG tracked and recorded their brain wave patterns.

Interpreting brain wave patterns

"Brain signals can be thought of as rhythms or ribbons of frequency bands," said Beex. "EEGs measure the signals as they change over time and frequency."

By passing the signal through low- and high-pass filters, decimation, and multi-resolution parametric spectral estimation, Beex and his team can look at these brain wave patterns on very fine timescales—so fine that they hope to trace where in the brain a signal originates as well as the pathways it travels.

Machine-learning techniques like K-Nearest Neighbor, Gaussian Mixture Models, and Universal Background Model classifiers helped the researchers select a combination of channels, that would best show whether or not the subject had ADHD.

"We found that only two or three channels were relevant to ADHD detection if we were monitoring the brain during the attention tasks," said Beex.

After analyzing the different channels and modeling the data, Beex and his students compared their findings to each child's diagnosis. And although the research is limited by the amount of data available, the team has achieved 85-95 percent accuracy rates so far.

Looking further, looking closer

While much work remains, Beex believes this technology could revolutionize the way ADHD is detected/diagnosed and subsequently streamline how it is treated.

"We'll be able to diagnose ADHD in a school cafeteria with a couple of electrodes hooked up to a laptop," said Beex. If a diagnosis is positive, he hopes that the same methods will also be used to test the effectiveness of behavioral therapy—interventions that can help children with ADHD manage their symptoms—with no side effects and no long-term risks.

But ultimately, Beex is driven toward a farther-reaching goal: he wants to know why humans develop this disorder in the first place and what, if anything, can be done to address it.

"By looking at the data in great detail and identifying the pathways involved, we are investigating ADHD, not as an inherent property in the architecture of the brain, but as something that developed," said Beex.