Researchers from the Cincinnati Children’s Hospital Medical Center are utilizing multichannel deep neural network model (mcDNN) in conjunction with MRI to predict attention deficit hyperactivity disorder (ADHD) in children, according to a study recently published in Radiology: Artificial Intelligence.
In the United States, a total of 6.1 million children have been diagnosed with ADHD. Many children with ADHD also struggle with at least one other mental, emotional, or behavioral condition, and 30 percent of youth suffer from anxiety. To lessen the symptoms, many children undergo a combination of behavioral therapy and medication. However, there’s no defined imaging exam to effectively determine if a child has ADHD or not. Instead, they’re typically assessed with psychological testing.
In their study, the group of researchers led by Lili He, PhD, used data from 973 patients from the Neuro Bureau ADHD-200 to better understand neurological differences in children who will develop ADHD. They created multi-scale functional brain connectomes using anatomic and functional criteria. Connectomes are created with spatial regions on MRI scans. To identify ADHD, the mcDNN model analyzed the connectome data and personal characteristic data. They examined the classification performance with cross-validation and hold-out validation methods using metrics such as accuracy, sensitivity, and specificity.
The researchers found that this multiple connectome maps using numerous brain parcellations was much more effective in generating results. “This model can be generalized to other neurological deficiencies,” said Dr. He. “We already use it to predict cognitive deficiency in pre-term infants. We scan them soon after birth to predict neurodevelopmental outcomes at two years of age.”