TOPOLOGICAL CHARACTERISTICS OF 5D SPATIALLY DYNAMIC BRAIN NETWORKS IN SCHIZOPHRENIA

The last decade of rich data-driven research on functional magnetic resonance imaging (fMRI) has provided novel in-sights into human brain function and aberrant behavior in brain disorders. Independent component analysis (ICA) is a widely-used technique for data-driven analysis of fMRI data. Spatial ICA is the most prominent variation of ICA and provides replicable and interpretable intrinsic connectivity network (ICN). It assumes common spatial activation across time. However, very recent studies indicate that there is utility in adopting a dynamic spatial activation modeling approach. Characterizing dynamics for both temporal and spatial domains means we have a multitude of decompositions of the already high-dimensional, multi-dataset, and multi-subject fMRI data. Hence making sense of the derived data becomes a significant issue. Here we use topological data analysis (TDA) to identify topological descriptors of the windowed spatially dynamic components of fMRI data. We discover and summarize differences in the spatial dynamics of controls and schizophrenia patients (SZs). We discover that SZs generally have lower Betti numbers and higher Wasserstein distance between spatiotemporal brain states, which provide intuitive summaries of the reduced dynamism SZs exhibit in resting-state fMRI studies.