A University at Buffalo researcher led a new study that identified brain connectivity as a biomarker of attention deficit hyperactivity disorder (ADHD). This is a result of specific communication between brain regions.
The research relied upon a deep architecture using machine learning classifiers to identify with 99 percent accuracy adults who had been diagnosed with ADHD as children.
“This suggests brain connectivity is a stable biomarker of ADHD, at least through childhood,” said Chris McNorgan (an assistant professor of psychology at the UB College of Arts and Sciences and the study’s principal author).
The journal published the findings. Frontiers in PhysiologyThese can be used to help diagnose ADHD, a common but difficult disorder that is hard to diagnose. They can also be used to help clinicians target treatments by helping them understand where patients fit on a continuum.
McNorgan, a neuroimaging expert and computational modeler, said that ADHD medications react differently to certain pathways. Understanding the differences can help you make informed decisions about which medication you should take.
Attention deficit disorder, the most common type of psychological disorder among school-aged children is difficult to identify. Multiple subtypes can complicate ADHD’s clinical diagnosis.
A patient’s clinical diagnosis of ADHD can change if they return for a second evaluation.
McNorgan stated that while patients may display symptoms that are consistent with ADHD, they might not be showing the same symptoms or even the same degree days later. It could be the difference between a good and bad day.
“But ADHD’s brain connectivity signature appears to be more stable.” We don’t see the diagnostic flip-flop.”
Cary Judson, a UB undergraduate researcher, and Dakota Handzlik, a Department of Computer Science and Engineering research volunteer, worked together to create a multidisciplinary research team. John G. Holden, an assistant professor of psychology at the University of Cincinnati used archival fMRI data of 80 adult participants who were diagnosed as having ADHD as children.
Four snapshots of activity were taken during a task that tested the subject’s ability not to respond to an automatic response using machine learning classifiers.
Focused analysis of individual runs yielded 91% diagnostic accuracy, while collective analysis achieved close to 99%.
McNorgan said, “It is by far the most accurate rate I’ve seen reported anywhere – it is leagues above anything that has come prior to it, well beyond anything that was achieved with behavioral assessments.” “We are able to classify better because of many factors.”
Previous research has suggested that ADHD and brain connectivity are related. This was done using linear classification. This research examines the relationships between something and what it is predicting, such a relationship between coffee and performance.
For many ranges, a direct linear classification is efficient, but the relationship between caffeine and performance, like ADHD symptoms or behavioral symptoms, is not linear. While a cup of coffee may increase performance, it can also cause problems. McNorgan states that nonlinear relationships can occur when you have “too much or too little” of a good quality.
Deep learning networks are excellent for detecting nonlinear conditional relationships. In the current study, ADHD was predicted based on the patterns of communication among brain areas.
McNorgan’s model also distinguishes individuals with ADHD who perform well on the Iowa Gambling Task (IGT). The IGT is a behavioral paradigm, similar to a card game. It offers both high-risk and lower-risk options. It is widely used to diagnose ADHD.
Traditional methods can’t distinguish between two classifications at once. McNorgan’s method elegantly ties ADHD diagnosis and performance on the IGT to provide a bridge that may explain why both are related to brain wiring.
People with ADHD are more likely to make riskier decisions in the IGT. However, this is not a universal determinant. People without ADHD can also make more risky choices than others.
McNorgan stated, “This approach of differentiating both dimensions provides a mechanism to sub-classify people with ADHD in ways which can allow for targeted treatment.” “We can see where people stand on the continuum.”
He explained that different brain networks can be implicated in people at each end of the continuum. This opens up the possibility of developing therapies that focus specifically on these brain networks.