Brain Over Behavior

A more reliable model of ADHD could improve not just diagnosis but treatment too.

An adolescent with a mind full of tangled thoughts.

Attention deficit hyperactivity disorder (ADHD) is the most commonly diagnosed psychological disorder among school-aged children, but it can be hard to diagnose correctly. Now, a new study using machine learning to identify a brain-based biomarker for the condition has achieved unprecedented accuracy in detecting ADHD.

The University at Buffalo psychologist who led the breakthrough research says the findings can also help clinicians target the most effective treatments for their young patients.

A more stable diagnostic tool

ADHD is a diagnostically slippery disorder, as it is based solely on behavioral assessments. Multiple subtypes further complicate a clinical determination.

“A patient may be exhibiting behavioral symptoms consistent with ADHD one day, but even days later, might not,” says Chris McNorgan, an assistant professor in UB’s Department of Psychology. “It could just be the difference between a good day and a bad day.”

Instead of focusing on behaviors, McNorgan looked at specific communication among different brain regions, known as brain connectivity, and found it to be a much more reliable indicator. “We don’t see the diagnostic flip-flop,” he says.

Highest reported accuracy rate

To test the approach, a multidisciplinary research team used data from 80 adult participants who were diagnosed with ADHD as children. Machine learning—a form of artificial intelligence—was applied to four snapshots of activity during a task designed to test the subject’s ability to inhibit an automatic response.

The results identified with 99% accuracy those adults who had received a childhood diagnosis of ADHD many years earlier.

“It’s by far the highest accuracy rate I’ve seen reported anywhere—well beyond the consistency we would see with a behavioral assessment,” McNorgan says.

Targeted treatments

Previous research suggesting a relationship between brain connectivity and ADHD used direct linear classification. But in ADHD, relationships between different areas of the brain are complex and intertwined. For example, a strong connection between two areas of the brain might be predictive of ADHD, but not if those regions are also strongly connected to a third area.

Machine learning is ideal for detecting this kind of non-linear relationship, McNorgan says. It “provides a mechanism for subclassifying people with ADHD in ways that can allow for targeted treatments. We can see where people are on the continuum.”

And because different brain networks are implicated in people at either end of the continuum, he adds, it also opens the door for developing therapies that focus on specific brain networks.