Disrupted Integration-Segregation Balance in the Intact Hemisphere in Chronic Spatial Neglect
Spatial neglect is a common and disabling consequence of right hemisphere stroke, characterized by a failure to attend to the contralesional left space, and frequently persists into the chronic stage. There is robust evidence on the role of right-hemisphere frontoparietal dysfunction, interhemispheric structural disconnection and maladaptive activity in the left hemisphere in the persistence of neglect. However, the specific impact of right frontoparietal dysfunction on the functional (re)organization of the left hemisphere remains poorly understood. In this study, we introduce a novel application of functional connectivity gradient analysis to investigate macroscale functional reorganization in the non-lesioned left hemisphere of patients with chronic left spatial neglect. Focusing on resting-state fMRI data, we demonstrate that abnormal segregation patterns in the left frontoparietal and default mode networks are robustly associated with neglect severity and spatial attentional bias. Notably, the principal gradient--typically capturing a global unimodal-to-transmodal hierarchy--was altered in these patients, suggesting a reorganization favoring lateralized unimodal networks. Single-subject analyses confirmed the presence of this pattern in 11 of the 13 patients included in the study. We also show that the structural integrity of the left inferior fronto-occipital fasciculus (IFOF) plays a key role in shaping these functional dynamics. These findings reveal a previously overlooked aspect of neglect pathophysiology: the maladaptive dominance of the non-lesioned hemispheres intrinsic architecture. By combining innovative gradient-based metrics with classical lesion approaches, our study offers a new framework for understanding neglect as an emergent property of large-scale network imbalance, with clinical implications for diagnosis and intervention, and theoretical consequences for models of hemispheric asymmetries and conscious access.