Dynamic coordination among three large-scale functional brain networks – the default mode network (DMN),
the central executive network (CEN), and the salience network (SN) – has been found to be aberrant in many
neuropsychiatric disorders, including obsessive-compulsive disorder (OCD). However, this line of research
typically relies on a group-based definition of the networks of interest: a standardized anatomical or functional
atlas or parcellation or a group-based independent component analysis (g-ICA). These group approaches
do not allow for a full accounting of individual variation in the structure of and relationships between these
large-scale networks. Discounting inter-individual heterogeneity, especially spatial variation (which has
been shown to be greater in several clinical populations, though not previously in OCD), may lead to failure
to detect significant effects not because they are not present, but because we did not target the right nodes
for every individual. We aim to address this problem in a systematic way, using existing data from patients
with OCD and in matched healthy controls.
Two approaches to remedy this problem have been proposed. An individualized probabilistic ICA
approach defines brain networks individually for each subject and then enters the individualized measures
into a second-level random effects analysis. A hierarchical probabilistic group ICA approach provides
model-based estimation of brain functional networks at both the population and subject level
simultaneously. Properties of the data (i.e., relations between subject-level and population-level variance
in variables of interest) determine whether hierarchical or non-hierarchical modeling will produce superior
results. We will analyze previously collected resting-state fMRI data from six studies (total 253 individuals
diagnosed with OCD, 148 of them were not on any medication, and 271 healthy controls without any
psychopathology, HC). We will derive subject-specific network maps and time courses using both individual
and hierarchical methods and use them to examine how subjects’ diagnosis, continuous clinical measures,
and non-clinical characteristics relate to (1) individual variability in topographical and functional organization
of DMN, SN, and CEN and (2) individual variability in functional coupling among these networks.
Prior research has demonstrated that rs-fMRI metrics of these brain-wide networks vary across OCD
dimensions and can serve as predictors of treatment response. However, calculation of these metrics does not
generally account for cross-subject topological heterogeneity, which can be misinterpreted as variations in
coupling. Our research will produce new and more precise markers of cross-patient heterogeneity, and de-
bias promising existing biomarkers of treatment response.
