Background and purpose Several studies have reported that sleep disordered breathing (SDB) is a risk factor for stroke. A prior study demonstrated that white matter lesions on CT in stroke patients were associated with severe SDB. White matter lesions are thought to reflect ischemic damage to the small deep cerebral vessels. Therefore, white matter lesion should be strongly related with cerebral microbleeds (MBs), deep subcortical white matter hyperintensity (DSWMH), and periventricular hyperintensity (PVH) on MRI. Therefore, we hypothesized that severe SDB patients had more frequently such MRI lesions than non-severe SDB patients. The aim of the present study was to examine whether presence of MBs, DSWMH, and PVH on MRI were associated with severe SDB in stroke and TIA patients. Methods Between April 2010 and June 2011, patients with stroke and TIA patients within 24 hours from onset were prospectively enrolled. Subject underwent respiratory sleep studies within 7days after admission, and measured the respiratory disturbance index (RDI). Patients were classified into 2 groups, severe SDB group had RDI>30 and non-severe SDB group had RDI<30. The present study performed MRI on admission, and assessed the presence of cerebral MBs on T2* weighted imaging, the percentage of DSWMH and PVH grade (Grade0-1 or 2-3) according to Fazekas classification on FLAIR imaging. Univariate and multivariate logistic regression analysis were performed on clinical characteristics including vascular risk factors, and MRI findings between severe SDB and non-severe SDB patients. Results Two hundred and twenty patients were enrolled (mean age 71.8 ± 11.9 years, male 61%, TIA 16%, ischemic stroke 65%, and hemorrhagic stroke 19%). 66 (27.2%) patients had severe SDB and 154 (72.8%) had non-severe SDB. MBs (58% for severe SDB group vs. 40% for non-severe SDB group, p=0.019), PVH grade 2-3 (44.4% vs. 27.1%, p=0.021), atrial fibrillation (30.3% vs. 16.4%, p=0.028), body Mass Index (BMI) (23.8 ± 3.3kg/m2 vs. 22.3 ± 3.9kg/m2, p=0.012) were significantly higher in the severe SDB group than in the non-severe SDB group. DSWMH Grade 2-3 was not different between two groups (43.2% vs. 31.3%, p=0.142). Multivariate logistic regression analysis demonstrated that MBs (OR 2.2, 95%CI 1.1-4.3, p=0.022), PVH grade 2-3 (OR 2.0, 95%CI 1.1-3.8, p=0.046), and BMI>23kg/m2 (OR 2.6, 95%CI 1.1-5.2, p=0.008) were independently associated with severe SDB. Conclusions MBs and PVH should be associated with severe SDB in stroke and TIA patients.