ѻý

Deep-Learning Model Shows Promise for ADHD Diagnosis in Teens

<ѻý class="mpt-content-deck">— AI algorithm identified significant differences in nine brain white matter tracts
MedpageToday

CHICAGO -- An artificial intelligence (AI) deep-learning model pinpointed significant differences in nine brain white matter tracts in adolescents with attention deficit-hyperactivity disorder (ADHD), researchers reported here.

Using brain imaging data from 1,704 participants in the Adolescent Brain Cognitive Development Study, fractional anisotropy (FA) measurements were significantly elevated in nine of 30 white matter tracts in patients with ADHD compared with those without ADHD, said Justin Huynh, MS, a research specialist at the University of California San Francisco and a medical student at the Carle Illinois College of Medicine at Urbana-Champaign, during the Radiological Society of North America annual meeting.

"The deep-learning algorithm was designed to recognize differences in brain patterns that were different enough to be considered, by the trained algorithm, statistically significant with a value of P=0.05," Huynh told ѻý. "The areas of the brain where these nine significant differences were found were mainly related to connections to the frontal cortex, which is related to executive functioning and attention."

"We also found some connections to speech comprehension and speech production. There were also some connections in the occipital cortex and parietal cortex," he added.

"ADHD often manifests at an early age and can have a massive impact on someone's quality of life and ability to function in society," Huynh said in a . "It is also becoming increasingly prevalent in society among today's youth, with the influx of smartphones and other distracting devices readily accessible."

He said the goal of the research was to develop a tool that can objectively diagnose ADHD -- diagnosis currently depends on self-reported behavioral surveys -- and also to delve into exactly what causes ADHD.

"Today, and for the last 20 years or so, ADHD diagnoses have been very subjective," he noted. "We really don't know, on a neurological basis, what exactly ADHD is. We really don't know what is happening in the brain. We think this tool will help us with that."

Max Wintermark, MD, chair of neuroradiology at the University of Texas MD Anderson Cancer Center in Houston, told ѻý that "finding an imaging biomarker that would allow us to reliably identify adolescents with ADHD would definitely be helpful, and the artificial intelligence algorithm may detect patterns that we may not be able to see with our own eyes when we look at these imaging studies."

"I would want to see confirmation of these results in a different population of ADHD patients," he said. "These imaging studies tend to be nonspecific and there are a lot of potential issues with reproducibility that limit its widespread application."

In performing the study, Huynh and colleagues accessed data from the multi-institutional Adolescent Brain Cognitive Development Study, which includes brain imaging data, clinical surveys, and other data on over 11,000 adolescents from 21 research sites in the U.S.

From that group, the researchers included 1,704 brain imaging scans from adolescents with and without ADHD. Using diffusion-weighted imaging, they extracted FA measurements -- a measure of how water molecules move along the fibers of white matter tracts -- along 30 major white matter tracts.

The FA values from 1,371 patients were used as input for training the deep-learning AI model, which was then tested on 333 patients, including 193 with ADHD and 140 without. ADHD diagnoses were determined by the Brief Problem Monitor assessment, a rating tool used for monitoring a child's functioning and their responses to interventions.

Huynh said the next steps for his team are to use intergroup findings to see if the deep-learning model can be used to individualize diagnoses. "Right now, if you take a large group of people, you can see, on average, significant differences," he noted. "We think these findings can set us on the path to actually detect changes in individuals."

"I feel that our results are pretty promising," he added. "If you look at the progress that has occurred across the board with artificial intelligence in the last 7 years, we believe it would not be unreasonable to think that this program could be helping identify people with ADHD within the next 5 years."

The researchers are currently seeking to add more individuals from the overall study to enhance their findings. However, the current number of patients that are included in the algorithm is "an order of magnitude greater than all the other previous studies that have looked for a signal," Huynh noted.

  • author['full_name']

    Ed Susman is a freelance medical writer based in Fort Pierce, Florida, USA.

Disclosures

Huynh and Wintermark disclosed no relevant relationships with industry.

Primary Source

Radiological Society of North America

Huynh J, et al "Unsupervised discovery of microstructural deviations in diffusion MRI tractometry of adolescents with ADHD" RSNA 2023.