Yanli Zhang, Junqiang Lei, Shunlin Guo, Tianhong Wang
Carbon monoxide poisoning (COP) often leads to acute brain injury and delayed neuropsychiatric sequelae (DNS), resulting in heavy social and economic burdens. Conventional imaging can detect gross structural lesions, and acute brain lesions on routine MRI are an established biomarker for predicting DNS. However, it fails to identify early occult injury and interpret pathological mechanisms. Advanced multimodal neuroimaging techniques enable comprehensive assessment of COP-associated brain damage. Structural changes are dominated by prefrontal and limbic lobe atrophy. Diffusion tensor imaging and diffusion kurtosis imaging can both assess microstructural damage of cerebral white matter; accumulating observational data suggest that kurtosis-derived metrics may yield higher sensitivity for detecting subtle white matter injuries. Abnormalities of the default mode network are commonly detected on resting-state functional MRI (rs-fMRI), while DNS patients show more severe damage in deep brain areas and the brainstem. Nuclear medicine imaging confirms basal ganglia and lobar hypoperfusion, and clarifies the unique pattern of striatal dopaminergic injury in COP-related parkinsonism. Magnetic resonance spectroscopy and glutamate chemical exchange saturation transfer characterize metabolic disturbances and excessive glutamate release, which are closely correlated with cognitive impairment. Advanced rs-fMRI approaches contribute greatly to exploring DNS pathogenesis, yet high-quality evidence across different modalities is insufficient, and no unified predictive biomarkers have been recognized to date. Given the limitations of existing single-center, small-sample and cross-sectional studies, further multicenter, longitudinal and multimodal investigations are needed to optimize predictive models and improve the diagnosis, treatment and mechanistic research of COP.