Advancing the Study of Functional Connectome Development

Journal Title
Journal ISSN
Volume Title
A better understanding of functional changes in the brain across childhood offers the potential to better support neurodevelopmental and learning challenges. However, neuroimaging tools such as functional magnetic resonance imaging (fMRI) and electroencephalography (EEG) are vulnerable to head motion and other artifacts, and studies have had limited reproducibility. To accomplish research goals, we need to understand the reliability and validity of data collection, processing, and analysis strategies. Neuroimaging datasets contain individually unique information, but identifiability is reduced by noise or lack of signal, suggesting it can be a measure of validity. The goal of this thesis was to use identifiability to benchmark different methodologies, and describe how identifiability associates with age across early childhood. I first compared several different fMRI preprocessing pipelines for data collected from young children. Preprocessing techniques are often controversial due to specific drawbacks and have typically been assessed with adult datasets, which have much less head motion. I found benefits to the use of global signal regression and temporal censoring, but overly strict censoring can impact identifiability, suggesting noise removed must be balanced against signal retained. I also compared several different EEG measures of functional connectivity (FC). EEG can be vulnerable to volume conduction artifacts that can be mitigated by only considering shared information with a time delay between signals. However, I found that mitigation strategies result in lower identifiability, suggesting that while removing confounding noise they also discard substantial signal of interest. Individual experiences may shape development in an individually unique way, which is supported by evidence that adults have more individually identifiable patterns of FC than children. I found that across 4 to 8 years of age, identifiability increased via increased self-stability, but without changes in similarity-to-others. In the absence of ground truth, it is difficult to argue for or against analysis decisions based solely on a theoretical framework and need to also be validated. My work highlights the importance of not thinking about techniques in a valid-invalid dichotomy; certain methods may be sub-optimal while still being preferable to alternatives if they better manage the trade off between noise removed and signal retained.
Graff, K. (2023). Advancing the study of functional connectome development (Doctoral thesis, University of Calgary, Calgary, Canada). Retrieved from