Segment-Then-Connect: Change Point Dynamic Connectivity for Early MCI Detection

The most widely used inputs in classification models for resting-state functional magnetic resonance imaging (rs-fMRI) data are estimates of static-based functional connectivity (SFC) and sliding window dynamic functional connectivity (swDFC). Although these methods are computationally convenient, the representations they yield are highly simplified portrayals of a deeply integrated and dynamic process. Change point dynamic functional connectivity (cpDFC) methods offer an alternative to swDFC approaches with many advantages. In this study, we consider a classification task between controls and patients with eMCI using rs-fMRI data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) studies, ADNI2 and ADNIGO. Our results indicate that the DFC methods are generally superior to the SFC methods when used as inputs into the classification model. Most importantly, we find that cpDFC os generally superior to swDFC. We discuss how cpDFC methods offer greater parsimony of network features and ease of interpretability. Our empirical results indicate that functional brain network representations are dynamic, multiscale, and subject-specific, underscoring the need for a learning paradigm tailored to these properties.