Irritability and Neural Basis of Reward-Processing in ADHD

Poster Session F - Tuesday, April 16, 2024, 8:00 – 10:00 am EDT, Sheraton Hall ABC

Prerona Mukherjee¹ (p.mukherjee@ucdavis.edu), Saeedeh Komijani², Ian Farnsworth^1,3, Dipak Ghosal², Julie B. Schweitzer¹; ¹Department of Psychiatry and Behavioral Sciences, MIND Institute, University of California, Davis, ²Department of Computer Sciences, University of California, Davis, ³Center for Mind and Brain, University of California, Davis

Attention Deficit Hyperactivity Disorder (ADHD) is a neurodevelopmental disorder in children and adolescents characterized by elevated hyperactivity and impulsivity, and deficits in attentiveness. While irritability is not a diagnostic symptom of ADHD, temper outbursts and irritable moods are common in individuals with ADHD. Irritability is defined by proneness to anger and has been linked to reward processing. Reward processing abnormalities are also strongly implicated in ADHD. Our previous work revealed atypical connectivity within brain networks linked to reward processing in individuals with ADHD who exhibit irritability. Nevertheless, it is not known if reward processing drives irritability in ADHD. In this study, we harness Machine Learning (ML) to probe the link between irritability and co-activation patterns across different brain regions in response to a reward processing paradigm. Our dataset includes brain imaging data and clinical measures from 128 participants (ADHD N=64; Neurotypical N=64), ages 12-30. We will train an ensemble-based ML model utilizing average beta values from Regions of Interest (ROIs) selected based on a suitable meta-analysis to perform classification of ADHD/non-ADHD with/without co-occurrence of irritability. Additionally, we will investigate changes in ADHD symptom severity. We will address collinearity using hierarchical clustering. Model evaluation will be done with k-fold cross-validation, and accuracy measures will be reported. We will inspect the interpretability of the model using the importance of features. Our results will elucidate how brain activity patterns can predict ADHD diagnosis and irritability status and exemplify the application of ML techniques to answer impactful clinical questions using complex multimodal data.

Topic Area: OTHER