Deep learning, a kind of AI, applied to MRI scans improved ADHD detection significantly.

The University of Cincinnati College of Medicine and Cincinnati Children’s Hospital have developed an AI system that uses MRI scans to help detect ADHD.

Attention Deficit Hyperactivity Disorder (ADHD) is a pediatric mental disorder. It causes symptoms such as an inability to focus and difficulty assessing people’s emotions. According to the National Survey of Children’s Health, approximately 9.4% of U.S. children, ages 2 to 17 years (6.1 million) in 2016 were diagnosed with ADHD.

Detection of ADHD

There is no individual test or a medical scan that can effectively diagnose ADHD in a child. Doctors commonly diagnose ADHD after assessing a series of symptoms and behavioral tests. However, brain MRIs have shown potential in the diagnosis of the disorder.

Researchers at the University of Cincinnati College of Medicine and Cincinnati Children’s Hospital have gone beyond current procedures for the use of brain MRIs in the detection of ADHD. They adopted a more comprehensive and “multiscale method” that used multiple versions of the MRI scan.

They then built an AI, deep learning model that processed multiscale brain imaging data from 973 participants in the NeuroBureau ADHD-200 project. The model was also fed personal characteristics such as gender and IQ.

They found that deep-learning-aided MRI-based diagnosis could assist in implementing early interventions for ADHD sufferers. That’s because as important as it is for the correct diagnosis of ADHD, early detection is a huge step forward in its treatment. In fact, the University of Cincinnati AI system allows doctors to do so before the patient even develops ADHD.

Other potential benefits

Using AI with MRI is a process that can extend to medical conditions beyond ADHD. “This model can be generalized to other neurological deficiencies,” says Lili He, Ph.D., senior author of the study. “We already use it to predict cognitive deficiency in preterm infants. We scan them soon after birth to predict neurodevelopmental outcomes at two years of age.”

The above approach can apply to much larger neuroimaging data sets for a more accurate diagnosis. This is now possible with significant advances in deep learning models as well as computing power.

[Related Story:  Medtronic Launches the First Ever Artificial Intelligence-Based System for Colonoscopy  ]