IndivSTATIS: A multivariate approach to analyze brain network configurations with individualized parcellation

Abstract

A critical step in the analysis of large-scale functional brain networks in neuroimaging is parcellation, which defines the nodes of a brain network. Group or atlas-based parcellation schemes use a shared common space, ensuring that each individual has the same number of brain parcels, which facilitates standard analytic approaches. However, studies reveal individual differences in the boundaries of brain areas. Extracting signals using atlas-based schemes can result in varying levels of blurring of signals across homogeneous areas within a specific individuals brain. Individualized parcellation schemes can be obtained when sufficient data are available; however, these approaches introduce a significant analytical challenge: the number of parcels and networks differ across individuals. Here, we introduce IndivSTATIS, a new multivariate method based on the STATIS framework, designed to integrate individualized parcellation schemes while maintaining comparability across participants in a shared component space. The resulting network/node component scores can be used to predict individual differences measures (e.g., age, behavior). By allowing individualized parcellations to be compared within a common component space, IndivSTATIS provides a solution for incorporating individual network variability into larger studies, with potential to improve the sensitivity and interpretability of functional brain markers across both basic neuroscience and clinical applications.