Prediction of ADHD in Adolescents Utilizing fMRI-Based Individual Cortical Thickness Measurements

Monday, June 12, 2023

2:00 PM-4:00 PM

BIOMED Master's Thesis Defense

Title:
Prediction of ADHD in Adolescents Utilizing fMRI-Based Individual Cortical Thickness Measurements

Speaker:
Julia Dengler, Master's Candidate
School of Biomedical Engineering, Science and Health Systems
Drexel University
 
Advisors:
John Medaglia PhD
Assistant Professor
Department of Psychological and Brain Sciences
College of Arts and Sciences
Drexel University

Hasan Ayaz, PhD
Associate Professor
School of Biomedical Engineering, Science and Health Systems
Drexel University

Details:
Attention-deficit/hyperactivity disorder (ADHD) is a neuropsychiatric disorder primarily affecting youth defined by symptoms of inattention and impulsivity. Current diagnosis tools consist of questionaries and tasks which are subject to biases. A quantitative and objective measurement for diagnosis does not exist. The use of fMRI-derived metrics, such as cortical thickness measurements, has gained popularity in diagnosing disease. Since cortical thickness differences exist in adolescents with ADHD, prediction algorithms can be written using this metric to diagnose ADHD. To determine regions most likely to predict ADHD, cortical thickness measurements were extracted in functional regions associated with symptoms of inattention and impulsivity, as well as for the whole brain.

Four different types of classification models were created and tested across the three different functional conditions: inattention, impulsivity, and whole brain. 238 ADHD and 244 healthy adolescent participants from the ADHD 200 publicly available dataset were used to test and train the models. Across all models, the average accuracy measure was 63.83% (+/- 2.2%). There were no differences in accuracy values between the different functional networks associated with symptoms and the whole brain analysis. Support vector classification (SVC) predicted ADHD the most accurately across the different functional network conditions. Specificity was found to be slightly better than sensitivity across SVC models.

Overall, cortical thickness provides a new avenue of possible prediction in diagnosing ADHD. Further work evaluating hemispheric cortical thickness differences and extracting the functional networks most responsible for accurate prediction needs to be completed to better understand how the models work.

Contact Information

Natalia Broz
njb33@drexel.edu