Pamela K. Douglas-Gutman, Ph.D.

Fellow in Child & Adolescent ADHD

Project Details


Mark S. Cohen, Ph.D.


University of California Los Angeles


Data Mining and Interpretation of the ADHD 200 Data Set


Data Mining and Interpretation of the ADHD 200 Data Set


In 2012, the global ADHD200 machine learning (ML) competition challenged the neuroscientific and data mining communities to develop pattern classification tools to predict ADHD diagnosis based on structural and functional MRI data. This competition provided one of the largest neuroimaging data consortiums to date with over 1000 subjects and potentially the most valuable public neuroimaging resources for studying ADHD. As a participant in this competition, thousands of potential neuroimaging “features” or variables were derived from these data. The goal, here, is to accelerate the process of knowledge gain by further analyzing the neural features which were identified as potentially prognostic in a high throughput analysis of this large data set. Structural features will be analyzed using novel quantitative differential geometry tools, and functional connectivity between these selected regions will be assessed on scrubbed data. Lastly, clustering techniques will be used to see how groups of variables are altered together in ADHD and each of its subtypes.

Like many behaviorally-diagnosed neurodevelopmental disorders, it is likely that multiple factors influencing several neural pathways can all lead to the ADHD phenotype. We hypothesize that this detailed analysis may identify and link novel biomarkers to provide new insights into the heterogeneous symptom clusters unique to ADHD subtypes. Downstream this knowledge may lead to quantitative metrics to aid in diagnosis and individualized treatment regimens.

Read Researcher’s Biography