Arraiza M, Bastarrika G, Zudaire B, Pueyo J, Villanueva A
Servicio de Radiología, Clínica Universidad de Navarra, Pamplona, España.
Radiologia. 2009 Mar-Apr;51(2):148-55. doi: 10.1016/j.rx.2008.01.001. Epub 2009 Mar 14.
To evaluate the interobserver agreement in quantifying left ventricular function and mass and to assess the accuracy of conventional manual contour tracing compared to semiautomatic segmentation analysis software.
Twenty consecutive subjects who underwent cardiac DSCT with retrospective ECG-gating were included. Two different multiphase image reconstructions were done in 5% steps throughout the entire cardiac cycle (0-95% of the R-R interval) with effective slice thickness of 1mm in the axial plane and 8mm in the short-axis orientation. Left ventricular function and mass were calculated by two independent observers, tracing endocardial and epicardial borders manually and using a semiautomatic software tool (Circulation II, Siemens). Ejection fraction (EF), end-diastolic volume (EDV), end-systolic volume (ESV), stroke volume (SV), cardiac output (CO), and myocardial mass were evaluated by two independent observers blind to each other's assessments. The interobserver agreement and the reliability of the different segmentation methods were calculated. The time required for manual contouring and semiautomatic contour tracing was also registered.
We found an excellent correlation (r>0.94; p<0.05) between the two independent observers for the quantification of left ventricular function and mass. Left ventricular functional parameters derived from semiautomatic contour software and conventional manual tracing method were not significantly different (p>0.05). The semiautomatic contour detection algorithm overestimated LV mass significantly compared with the manual contouring method (mean difference 29.45+/-1.64g; p<0.05). The time needed to calculate these parameters with the semiautomatic tool was significantly lower (248.85+/-99.8s) than with manual contouring (452.7+/-73.92s) (p<0.05).
Interobserver agreement for quantifying left ventricular function and mass using DSCT is excellent. Despite overestimating left ventricular mass, the semiautomatic software tool allows cardíac parameters to be quantified with the same reliability as the conventional manual method in half the time.
评估观察者间在量化左心室功能和质量方面的一致性，并评估与半自动分割分析软件相比，传统手动轮廓追踪的准确性。
纳入连续20例接受心脏双源CT（DSCT）及回顾性心电图门控的受试者。在整个心动周期（R-R间期的0-95%）以5%的步长进行两种不同的多期图像重建，轴面有效层厚1mm，短轴方向有效层厚8mm。由两名独立观察者通过手动追踪心内膜和心外膜边界并使用半自动软件工具（Circulation II，西门子）来计算左心室功能和质量。两名独立观察者在彼此评估结果未知的情况下对射血分数（EF）、舒张末期容积（EDV）、收缩末期容积（ESV）、每搏输出量（SV）、心输出量（CO）和心肌质量进行评估。计算观察者间的一致性以及不同分割方法的可靠性。同时记录手动轮廓追踪和半自动轮廓追踪所需的时间。
两名独立观察者在量化左心室功能和质量方面具有极好的相关性（r>0.94；p<0.05）。半自动轮廓软件得出的左心室功能参数与传统手动追踪方法无显著差异（p>0.05）。与手动轮廓追踪方法相比，半自动轮廓检测算法显著高估了左心室质量（平均差异29.45±1.64g；p<0.05）。使用半自动工具计算这些参数所需的时间（248.85±99.8秒）明显低于手动轮廓追踪（452.7±73.92秒）（p<0.05）。
使用DSCT量化左心室功能和质量时观察者间的一致性极佳。尽管高估了左心室质量，但半自动软件工具能够在一半的时间内以与传统手动方法相同的可靠性来量化心脏参数。
