Could AI Be the Future of ADHD Diagnosis?

At the State University of New York at Buffalo, researchers are exploring a new way of diagnosing ADHD in adults.

Chris McNorgan, PhD, and his colleagues applied the principles of machine learning in identifying how brain connectivity—communication among regions of the brain—can be used as a biomarker for ADHD. They analyzed archival MRI data and behavioral data of 80 participants who had completed the go/no-go task from the longitudinal follow-up of the Multimodal Treatment Study of ADHD.

The results show that task-based functional connectivity is a biomarker of ADHD. In the study, they write that the framework they created provides a template approach that explicitly ties behavioral assessment measures for ADHD to both clinical diagnosis and functional connectivity, the communications within the brain.

They also found that brain connectivity differences remain through childhood into adulthood and across environments—home, work, school, social life, and other places of life.

“This suggests that brain connectivity is a stable biomarker for ADHD… even when an individual’s behavior had become more typical, perhaps by adapting different strategies that obscure the underlying disorder,” says Dr. McNorgan. A clinical diagnosis of ADHD can change, whether it’s measured across a group of years or even daily. Measuring brain connectivity can provide a more accurate result regardless of those changes between years or days. “A patient may be exhibiting behavioral symptoms consistent with ADHD one day, but even days later, might not present those symptoms, or to the same degree. But the brain connectivity signature of ADHD appears to be more stable. We don’t see the diagnostic flip-flop.”

Machine learning and brain connectivity

Machine learning is computer programing that allows computers to make predications based on data they receive. It is a form of artificial intelligence and is often used by researchers to study and interpret a great deal of information more quickly. In this study, the researchers used machine learning and the Iowa Gambling Task, which is used to investigate decision making, to assess how the study participants’ brain processed information. Our brains gather and analyze information in different sections and those sections must work together to understand information and then allow us to make decisions. Machine learning allowed the researchers to map that decision-making activity and observe how well those sections of the brain were able to communicate.

The connections made during the analyzing phase resemble more of a spider web rather than a straight-line approach. Researchers were able to identify specific patterns within that web of connections. An ADHD diagnosis was predicted from the patterns of communication between groups of brain areas during the study. The type of pattern could also be used for identifying the ADHD presentation (inattentive, hyperactive, or combined) and the ways in which ADHD affects the specific individual.

“This approach by differentiating both of these dimensions provides a mechanism for sub-classifying people with ADHD in ways that can allow for targeted treatments,” Dr. McNorgan says. “We can see where people are on the continuum.”

What this could mean in the future

This study is not considered diagnostic, nor can it be used to help evaluate someone for ADHD at this time. Dr. McNorgan and his colleagues suggest it helps us to better understand ADHD as a brain-based condition with specific actions occurring within the brain.

In the future, machine learning approaches could be incorporated into a medical review during an evaluation for ADHD, especially if the science can help medical providers to better determine which ADHD presentation someone may have. They could also help to improve treatment options by supporting a more tailored approach for the individual person.

“Our multiple constraint network analysis is generalizable to other behavioral assessments and domains, and may guide development of more efficacious intervention strategies,” Dr. McNorgan writes.

By applying machine learning to ADHD and similar conditions, he says, researchers and medical professionals can develop new treatments and strategies that better meet the needs of those with ADHD.

Interested in more research?

Join the discussion: What research approaches would you be interested in learning more about?