A novel method to demonstrate thrombus formation of the left atrial appendage in patients with persistent atrial fibrillation by cardiac computed tomography.

Background Transesophageal echocardiography (TEE) is currently the gold standard technique for diagnosing left atrial appendage (LAA) thrombi. Cardiac computed tomography (CT) has been expected to become an alternative method to TEE; however, a reliable quantitative evaluation method has not been established. Methods and results We enrolled 177 patients with persistent atrial fibrillation who underwent both cardiac CT and TEE before catheter ablation. The patients were classified into two groups according to the TEE results: the thrombus group (13 patients) and non-thrombus group (164 patients). The Hounsfield unit (HU) density at the proximal LAA (LAAp) and distal LAA (LAAd) was measured on cardiac CT images. The LAAd/LAAp HU ratio and standard deviation of HU density (HU-SD) at the LAAd were evaluated. We created an algorithm by decision tree analysis to predict LAA thrombus formation using the HU ratio and HU-SD. Definite absence of LAA thrombus (Category-I) was diagnosed for 139 patients by combining the first and second branching of the decision tree (Category-Ia: HU ratio of ≥0.26, Category-Ib: HU ratio of <0.26, HD-SD of ≥26.94). Definite presence of LAA thrombus (Category-Ⅱ) was diagnosed for 3 patients using the third branching of the decision tree (Category-Ⅱ: HU ratio of <0.26 and HU-SD of <13.85). Highly possibility of LAA thrombus (Category-III), but not definite, was diagnosed for the remaining 35 patients; therefore, these patients required diagnostic TEE. The diagnostic accuracy of this algorithm was 0.95. Conclusion We have proposed a reliable algorithm to diagnose LAA thrombus formation using the HU ratio and HU-SD.