Objectives: We assessed the diagnostic performance of two-phase 64-slice cardiac computed tomographic angiography (CCTA) for the detection of a cardiac source of embolism in stroke patients using transesophageal echocardiography (TEE) as the reference standard. Background: Transesophageal echocardiography (TEE) is considered the reference standard method for the detection of potential sources of cerebral embolism. Although TEE is effective in detecting cardiac sources of embolism, a less invasive modality that is capable of assessing the cardiovascular system for embolic stroke is necessary. Methods: We selected 127 patients who had experienced a recent episode of stroke (onset within the previous seven days) and had undergone both two-phase 64-slice CCTA and TEE within a period of 5 days. A potential cardiac source of embolism detected at both CCTA and TEE was recorded, and echocardiographic findings were categorized into high- and medium-risk sources based on the TOAST (Trial of Org 10172 in Acute Stroke Treatment) classification. Results: Of 127 patients, 91 abnormal findings in 81 patients were found on TEE. In addition, 10 patients had more than one abnormal finding, and 46 patients had no abnormal finding on TEE. Using TEE as the reference standard, the overall sensitivity, specificity, accuracy, positive predictive value (PPV) and negative predictive value (NPV) of the 64-slice CCTA for detecting cardiac sources of embolism were 88%, 100%, 92%, 100% and 81%, respectively. For high-risk sources for cardiac embolic stroke, TEE detected a total of 47 abnormal findings, whereas CT detected 44 abnormal findings. However, The diagnostic accuracy was not significantly different between the CT and TEE for the detection of high-risk sources of embolism (p = 0.250). For medium-risk sources of cardiac embolic stroke, TEE detected a total of 44 abnormal findings, whereas, the CT detected 36 abnormal findings. The diagnostic accuracy was significantly different between the CT and TEE for the detection of medium-risk sources of embolism (p = 0.008). Conclusions: We found that two-phase 64-slice CCTA is a noninvasive and useful modality for detecting high-risk cardiac sources of embolism in stroke patients.