Combining static and dynamic functional connectivity analyses to identify male patients with obstructive sleep apnea and predict clinical symptoms

Patients with obstructive sleep apnea (OSA) experience chronic intermittent hypoxia and sleep fragmentation, leading to brain ischemia and neurological dysfunction. Therefore, it is important to identify features that can differentiate patients with OSA from healthy controls (HC) and provide insights into the underlying brain alterations associated with OSA. This study aimed to distinguish patients with OSA from healthy individuals and predict clinical symptom alterations using cerebellum-whole-brain static and dynamic functional connectivity (sFC and dFC, respectively), with the cerebellum as the seed region. Sixty male patients with OSA and 60 male HC matched for age, education level, and sex were included. Using 27 cerebellar seeds, sliding-window analysis was performed to calculate sFC and dFC between the cerebellum and the whole brain. The sFC and dFC values were then combined and used in multiple machine-learning models to distinguish patients with OSA from HC and predict the clinical symptoms of patients with OSA. Patients with OSA showed increased dFC between cerebellar subregions and the superior and middle temporal gyri and decreased dFC with the middle frontal gyrus. Conversely, increased sFC was observed between cerebellar subregions and the cerebellar lobule VI, cingulate gyrus, middle frontal gyrus, inferior parietal lobules, insula, and superior temporal gyrus. Combined dynamic-static FC features demonstrated superior classification performance with a support vector machine in discriminating OSA from HC. In clinical symptom prediction, FC alterations contributed up to 30.11 % to cognitive impairment, 55.96 % to excessive sleepiness, and 27.94 % to anxiety and depression. Combining cerebrocerebellar sFC and dFC analyses enables high-precision classification and prediction of OSA. Aberrant FC patterns reflect compensatory brain reorganization and disrupted cognitive network integration, highlighting potential neuroimaging markers for OSA. •Combined dFC-sFC features have superior classification performance using a support vector machine.•Combining cerebrocerebellar sFC and dFC analyses enables high-precision classification and prediction of OSA.•Aberrant FC patterns reflect compensatory brain reorganization and disrupted cognitive network integration.