No abstract

No abstract

No abstract

No abstract

Introduction: Assessment of the left atrial appendage (LAA) in patients with atrial fibrillation (AF) undergoing cardioversion for thromboembolic risk stratification is inherently qualitative. Oftentimes, interobserver assessment may disagree, leading to differences in risk stratification and patient management. We propose a novel method of quantitative LAA assessment and examine its predictive value for future thromboembolic events. Hypothesis: Evaluation of LAA using ImageJ pixel density quantification software in patients with AF will be superior to qualitative assessment by expert readers in assessing and predicting future thromboembolic risk. Methods: A total of 124 patients (mean age 59 ± 13 years, 70% male) undergoing transesophageal echocardiography to exclude LAA thrombus prior to radiofrequency pulmonary vein isolation or electrical cardioversion for AF were retrospectively studied. LAA were examined by two expert readers and identified as thrombus ( n = 9), sludge ( n = 11), spontaneous echocontrast ( n = 90), or normal ( n = 138). LAA were then separately examined using ImageJ to calculate a gain-independent ratio between the average pixel density of the LAA cavity and the average pixel density of the LAA wall (C/W ratio). Results: There was significant disagreement between the two expert readers in LAA assessment (weighted Kappa 0.510, p < 0.0001). Qualitative LAA analysis was significantly related with C/W ratio ( p < 0.0001) and thromboembolic events ( p < 0.0002). C/W ratio was also significantly related with thromboembolic events (OR 1.60, 95% CI 1.095 - 2.347, p < 0.0152). According to Area Under the Curve (AUC) of the ROC curve, C/W ratio (AUC 0.73, 95% CI 0.60 - 0.86) was a more reliable predictor for future thromboembolic when compared to qualitative LAA assessment (AUC 0.72, 95% CI 0.53-0.90). Conclusions: Quantitative LAA assessment using the C/W ratio provided by ImageJ software is more reliable than qualitative assessment by expert readers for identifying future thromboembolic events while reducing reader variability.