Date

7-24-2015

Document Type

Thesis

Degree Name

M.P.H.

Department

Department of Public Health & Preventive Medicine

Institution

Oregon Health & Science University

Abstract

Background – ADHD is one of the most prevalent neurodevelopmental disorders in our society. One of the hallmark features of ADHD is increased reaction time variability (RTV), which is particularly significant among children, as RTV is implicated in many goal-­‐directed behaviors and later academic achievement. Recent brain imaging studies have revealed that those with ADHD have atypical functional brain signatures compared with those without ADHD, suggesting a neurobiological basis for ADHD that can be measured using brain imaging.

Methods – This is a historical cohort study of 32 children (7-­‐14y) with ADHD and 40 children without ADHD. We administered a test of reaction time variability at study entry and then at least at one other time-­‐point in the study. We use support vector machine-­‐based multivariate pattern analysis to determine whether connection features derived from resting state functional connectivity MRI (rs-­‐fcMRI) were able to predict longitudinal changes in a test of RTV.

Hypothesis – We hypothesized that baseline rs-­‐fcMRI measurements would predict RTV in the sample as a whole (i.e. ADHD and typically developing children as a single group). Furthermore, we expected that predictions of change in reaction time variability would become more robust when considering typically developing children and children with ADHD separately, as there are likely to be predictive systems that are distinct in these two groups, as well.

Results – Connectivity between consensus features positively predicted changes in RTV when children with ADHD and TDC were evaluated as a single group (Adjusted R2 = 0.10, p = 0.0027), and predictive functional networks include the cinguloopercular and ventral attention networks. When children with ADHD were considered alone, connectivity between consensus features improves in its ability to positively predict changes in RTV (Adjusted R2 = 0.3183, p < 0.0001), and predictive functional networks include the cinguloopercular, default, visual, and ventral attention networks. When considering TDC alone, predictive functional networks include the cinguloopercular, default, and dorsal attention networks, and consensus features also positively predict changes in RTV (Adjusted R2 = 0.2813, p < 0.0001). Finally, group comparison between predictive networks in ADHD vs. TDC reveals a degree of overlap (particularly between the cinguloopercular and default networks), but also highlights specific subregions that are distinct within these networks, including the anterior insula/frontal operculum (within the cinguloopercular network), the inferior parietal subregion (within the frontoparietal network), the lateral parietal and inferior temporal subregions (within the default network), and the ventral vs. dorsal attention networks.

Conclusion – We apply multivariate statistics along with non-­‐invasive brain imaging to use baseline functional connectivity measurements to predict longitudinal change in RTV in children with and without ADHD. The functional connections that predict RTV in these two groups overlap in some regions, but are significantly distinct in other regions. While there are functional similarities in the brains of children with ADHD and TDC, there are also innate differences in the functional connectivity of these two populations, and these differences play a significant role in predicting longitudinal changes in RTV across these two groups. Ultimately, this work has the potential to characterize an approach aimed at identifying children at high-­‐risk of deficits in future cognitive performance and thus associated life outcomes such as academic achievement.

Identifier

doi:10.6083/M4QN65KV

School

School of Medicine

Available for download on Tuesday, July 24, 2018

Included in

Public Health Commons

Share

COinS