Relationship between normalized distributional pattern and functional outcome in patients with acute cardiogenic cerebral embolism

This study aimed to elucidate spatial characteristics for magnetic resonance imaging (MRI) of cardiogenic cerebral embolism, to determine imaging biomarkers predicting patient outcome and cerebral herniation in cardioembolic stroke. This retrospective study assessed 90 patients with cardiogenic cerebral embolism. All images from MRI were normalized using a voxel-based symptom lesion mapping technique. Patients were categorized into two subgroups based on the outcome and presence of cerebral herniation. Each subgroup was assessed individually. The distribution map of all analyzed patients revealed accumulated ischemic lesions in bilateral middle cerebral artery areas. Ischemic lesions for the poor outcome group accumulated at the corona radiata on the right side and throughout the entire left hemisphere. Receiver operating characteristic (ROC) analysis suggested that a normalized ischemic volume of 62.8 mL allowed optimal differentiation between good and poor outcomes (sensitivity, 0.923; specificity, 0.923; area under the curve (AUC), 0.91) for left-side-dominant infarction. The distribution map for the cerebral herniation group revealed large ischemic areas in the left hemisphere. The analysis of differential involvement map with random permutation analysis showed that left anterior circulation infarcts were associated with midline shift. Receiver operating characteristic analysis revealed that a normalized infarction volume of 192.9 mL was highly predictive of cerebral herniation (sensitivity, 0.929; specificity, 0.750; AUC, 0.895). The medial frontal and occipital lobes, caudate head and basal ganglia were significantly involved in those patients who developed cerebral herniation. Ischemic volume contributed to outcomes and cerebral herniation. Ischemic lesions of the anterior and posterior cerebral arteries and basal ganglia in addition to middle cerebral artery area were identified as differences on MRI images between with and without cerebral herniation patients.