Entropy-based integration index for quantifying network integration in resting-state functional MRI

Abstract

Independent component analysis (ICA) is widely used in resting-state fMRI to identify large-scale functional networks; however, existing approaches provide limited means of quantifying how network representations are distributed across independent components. We introduce an entropy-based network integration framework that characterizes the organizational architecture of canonical resting-state networks by quantifying the distribution of ICA-derived functional contributions within Yeo atlas networks. Spatial overlap between independent components and network templates is normalized to generate a probability distribution, from which Shannon entropy and a normalized integration index are derived. The resulting metric provides a continuous measure of network representational integration, ranging from specialized configurations dominated by a small number of components to distributed configurations involving multiple functional modes. The framework was evaluated and validated using resting-state fMRI data from healthy controls, Parkinsons disease patients with normal cognition, and Parkinsons disease patients with mild cognitive impairment. Global entropy and integration measures were complemented by network-specific analyses, dominance profiling, principal component analysis (PCA), and multivariate centroid-distance assessments. The proposed framework revealed selective alterations in Ventral Attention and Limbic network organization associated with cognitive-status differences, while preserving overall within-group heterogeneity. Group-wise PCA independently further identified these networks as major contributors to altered network organization, and centroid-distance analyses demonstrated that observed differences reflected coherent shifts in network architecture rather than increased variability. By quantifying the distribution of network representations across ICA-derived functional modes, this framework provides a simple, interpretable, and generalizable measure of large-scale brain organization, offering a complementary approach for studying network reorganization in health and disease.