Limbic-Visual Disintegration and Salience-Control Specialization Characterize Tinnitus Network Topology

Subjective tinnitus (ST) has been hypothesized to arise from large-scale network reorganization, but the affected circuits and their symptom scaling remain unclear. In a normal-hearing cohort (N = 114; 57 ST, 57 matched controls), we combined resting-state fMRI graph topology, ROI-to-ROI connectivity, voxel-based morphometry (VBM), and multivariate modelling using a harmonized 50-ROI parcellation from the CONN atlas, spanning 15 functional networks. Node-wise analyses (covarying age, sex, education, and motion; multiple-comparison control) showed selective, not global, reconfiguration: reduced integration centered on the left posterior parahippocampal gyrus (lower global efficiency and degree), increased segregation/clustering in inferior frontal and anterior insular hubs, longer path length in parahippocampal and frontal regions, and elevated local efficiency in the right amygdala. Network-based revealed hyperconnectivity in fronto-salience-language-cerebellar circuits and hypoconnectivity across default mode and dorsal attention/temporo-parietal pathways. Symptom coupling was convergent and dissociable: Higher tinnitus severity/duration tracked reduced integration in medial visual/limbic regions with increased integration/degree in right frontal-temporal opercular nodes, whereas higher anxiety related to increased integration/clustering in subcallosal, cerebellar, and occipito-limbic territories alongside decreases in putamen. VBM demonstrated widespread white-matter reductions (inferior frontal, temporal pole, insula, inferior temporal gyrus, and putamen) with more focal gray-matter effects, and a multivariate GLM confirmed a robust omnibus group difference. These multimodal, symptom-linked signatures provide strong evidence that ST reflects targeted network reorganization reduced medial temporal/visual-limbic integration accompanied by increased local specialization within salience and control hubs and yield actionable circuit markers for patient stratification and mechanism-guided treatment targeting.