Automated Threshold-Based 3D Segmentation Versus Short-Axis Planimetry for Assessment of Global Left Ventricular Function with Dual-Source MDCT

Abstract

The purpose of this study was to evaluate software for threshold-based 3D segmentation of the left ventricle in comparison with traditional 2D short axis-based planimetry (Simpson method) for measurement of left ventricular (LV) volume and global function with state-of-the-art dual-source CT. Fifty patients with known or suspected coronary artery disease underwent coronary CT angiography. LV end-diastolic, end-systolic, and stroke volumes and ejection fraction were determined from axial images to which 3D segmentation had been applied and from short-axis reformations from 2D planimetry. Interobserver variability was assessed for both approaches. Threshold-based 3D LV segmentation had excellent correlation with 2D short-axis results (end-diastolic volume, R = 0.99; end-systolic volume, R = 0.99; stroke volume, R = 0.90; ejection fraction, R = 0.97; p < 0.0001). Bland-Altman analyses revealed systematic underestimation of LV end-diastolic volume (-7.4 +/- 8.9 mL) and LV end-systolic volume (-7.0 +/- 4.4 mL) with the 3D segmentation approach and 2.8 +/- 3.3% overestimation of LV ejection fraction. Interobserver variation with 3D segmentation analysis was significantly (p < 0.001) less (e.g., LV ejection fraction, 0.1 +/- 1.7%) than with the 2D technique, and mean analysis time was significantly shorter (172 +/- 20 vs 248 +/- 29 seconds; p < 0.05). Automated threshold-based 3D segmentation enables accurate and reproducible dual-source CT assessment of LV volume and function with excellent correlation with results of 2D short-axis analysis. Exclusion of papillary muscles from LV volume results in small systematic differences in quantitative values.