Features on initial computed tomography scan of infarcts with a cardiac source of embolism in the NINDS Stroke Data Bank.

BACKGROUND AND PURPOSE: The lack of valid criteria for the clinical diagnosis of cardiogenic embolism is a major problem in both patient care and research. The aim of this study was to identify features on the initial computed tomogram of the brain that discriminate between patient groups with and without a cardiac source of embolism. To gain insight into the neuroradiological features relevant to the diagnosis of cardiac embolic stroke, we studied the initial computed tomogram of the 1,267 patients with ischemic stroke and such a scan in the National Institute of Neurological Disorders and Stroke (NINDS) Stroke Data Bank. METHODS: We analyzed the initial computed tomographic data from 1,267 patients with ischemic stroke in the NINDS Stroke Data Bank. Based solely on the presence of cardiac sources of embolism, we defined groups with high (n = 244), medium (n = 165), and low (n = 858) risk for cardiogenic embolism and compared the features on the initial computed tomogram among these three groups. RESULTS: Patients in the high-risk group were significantly more likely (p < 0.001) to have infarcts involving one half lobe or larger or infarcts involving both superficial and deep structures than patients in the medium- or low-risk groups. In contrast, deep small infarcts had a negative association (p = 0.004) with the presence of a cardiac source of embolism. There was no significant trend across risk groups in the percent with hemorrhagic infarction, regardless of whether patients with anticoagulant use at the time of the stroke were excluded. CONCLUSION: Although some features of the initial computed tomogram had highly significant associations with the presence of a cardiac source of embolism, the predictive value of these features for an embolic source was low.