In neuroscience, the notion has emerged that the brain abides by two principles: segregation and integration. Segregation into functionally specialized systems and integration of information flow across systems are basic principles that are thought to shape the functional architecture of the brain. A measure called integration, originating from information theory and derived from mutual information, has been proposed to characterize the global integrative state of a network. In this paper, we show that integration can be applied in a hierarchical fashion to quantify functional interactions between compound systems, each system being composed of several regions. We apply this method to fMRI datasets from patients with low-grade glioma and show how it can efficiently extract information related to both intra- and interhemispheric reorganization induced by lesional brain plasticity.