Atrial thrombi are a potential source for cerebral and peripheral emboli. Objective of this study was to evaluate the diagnostic accuracy of 64-slice cardiac computed tomography (CT) for detection of atrial thrombi in comparison with transesophageal echocardiography (TEE) and cardiac surgery. Sixty-four patients were examined with ECG-gated multidetector CT coronary or pulmonary vein angiography. All patients underwent TEE. Cardiac surgery was performed in 31 patients. The Hounsfield units (HU) of atrial lesions were measured. The diagnostic accuracy of 64-slice CT for the detection of atrial thrombi was 77%: sensitivity 100% (9/9), specificity 73% (40/55), positive predictive value (PPV) 38% (9/24), and negative predictive value (NPV) 100% (40/40). All 15 false positive (FP) findings by CT were located in the left atrial appendage (LAA). Four characteristic imaging features suggesting incomplete filling of the LAA were noted in FP: "hypostatic layering," 5/15 (33%); "flow phenomenon," 9/15 (60%); "HU-run-off," 8/15 (53%); higher intralesional HU in FP when compared with thrombi (153.8 HU +/- 71 vs. 46.6 HU +/- 10; P < 0.0001). The diagnostic accuracy of CT in detecting atrial thrombi improved significantly (P = 0.03) to 86% after defining "typical filling defects" as "flow phenomenon/>180 HU" (sensitivity 100%; specificity 84%; PPV 50%; NPV 100%). On receiver operating curve (ROC) analysis, a threshold of 60.7 HU showed a specificity of 100% and a sensitivity of 86.7% to distinguish between FP and thrombi. Cardiac ECG-gated 64-slice CT is accurate to exclude atrial thrombi, which can be applied eg, in patients before radiofrequency (RF) ablation. Left atrial appendage "filling defects" cause a high number of false positive findings, and there are radiologic features, which are helpful to differentiate them from true thrombi.