Topological integration and segregation of the conscious connectome as biomarkers for state, behavior, and prognosis

Abstract

Consciousness is hypothesized to emerge from a sophisticated balance between functional integration and segregation across large-scale brain networks. However, how these topological properties are disrupted in disorders of consciousness (DoC), and whether they can predict long-term recovery, remain to be established. Using resting-state functional MRI and graph-theoretical modeling in patients with DoC (N = 117) and healthy controls (HCs; N = 30), we mapped the macroscale reconfiguration of the “conscious connectome” to delineate disruptions in functional integration and segregation. Associations between topological features and behavioral consciousness, as measured by the Coma Recovery Scale-Revised, were further investigated. In addition, we developed a machine-learning framework to evaluate the prognostic value of network topology for neurological outcomes. Compared with HCs, patients with DoC exhibit widespread disruptions in both functional integration and segregation, with more pronounced deficits in segregation. Functional integration, but not segregation, is positively correlated with behavioral consciousness, particularly within the somatomotor network. Furthermore, the machine-learning framework leveraging macroscale network topology, notably the default mode network, predicts long-term recovery with a precision–recall area under the curve (PR-AUC) of 0.61. The inclusion of clinical variables further improved model performance (PR-AUC = 0.84). These findings suggest that functional segregation is more sensitive to the topological disruption characterizing pathological conscious state, whereas residual functional integration better explains interindividual variability in conscious behavior; together, they provide a topological scaffold for predicting neurological prognosis. Collectively, this framework provides objective and individualized biomarkers for the assessment and neuroprognostication of patients with DoC. After a severe brain injury, some patients show very limited signs of consciousness. Better tools are needed to understand their condition and estimate their chances of recovery. In this study, we analyzed brain scans from more than one hundred patients with disorders of consciousness. We examined how different brain regions work together and how specialized brain systems remain organized. We found that both types of brain organization were disrupted in patients. However, preserved communication between brain regions was more closely related to remaining signs of consciousness. These brain network features also helped predict long-term recovery. Our findings provide insight into how consciousness is supported by the human brain and may contribute to more accurate diagnosis and prognosis for patients with severe brain injuries. Zhu et al. characterize large-scale brain network topology in disorders of consciousness using resting-state functional MRI and graph-theoretical analyses. They show that disrupted network topology characterizes consciousness impairment, reflects behavioral responsiveness, and predicts long-term recovery in patients.