Abstract：
We hypothesized that if computed tomography (CT) images were used as learning data, we could overcome volume underestimation by echocardiography, improving the accuracy of left ventricular (LV) volume measurements.
We utilized a fusion imaging modality consisting of echocardiography with superimposed CT images for 37 consecutive patients to identify the endocardial boundary. We compared LV volumes obtained with and without CT learning trace-lines (TLs). Furthermore, 3D echocardiography was used to compare LV volumes obtained with and without CT learning for endocardial identification. The mean difference between the echocardiography and CT-derived LV volumes and the coefficient of variation were compared pre- and post-learning. Bland-Altman analysis was used to assess the differences in LV volume (mL) obtained from the 2D pre-learning TL and 3D post-learning TL.
The post-learning TL was located closer to the epicardium than the pre-learning TL. This trend was particularly pronounced in the lateral and the anterior wall. The post-learning TL was along the inner side of the high echoic layer in the basal-lateral wall in the four-chamber view. CT fusion imaging determined that the difference in LV volume between 2D echocardiography and CT was small (-25.6 ± 14.4 mL before learning, -6.9 ± 11.5 mL after learning) and that CT learning improved the coefficient of variation (10.9 % before learning, 7.8 % after learning). Significant improvements were observed during 3D echocardiography; the difference in LV volume between 3D echocardiography and CT was slight (-20.5 ± 15.1 mL before learning, 3.8 ± 15.7 mL after learning), and the coefficient of variation improved (11.5 % before learning, 9.3 % after learning).
Differences between the LV volumes obtained using CT and echocardiography either disappeared or were reduced after CT fusion imaging. Fusion imaging is useful in training regimens for accurate LV volume quantification using echocardiography and may contribute to quality control.
We utilized a fusion imaging modality consisting of echocardiography with superimposed CT images for 37 consecutive patients to identify the endocardial boundary. We compared LV volumes obtained with and without CT learning trace-lines (TLs). Furthermore, 3D echocardiography was used to compare LV volumes obtained with and without CT learning for endocardial identification. The mean difference between the echocardiography and CT-derived LV volumes and the coefficient of variation were compared pre- and post-learning. Bland-Altman analysis was used to assess the differences in LV volume (mL) obtained from the 2D pre-learning TL and 3D post-learning TL.
The post-learning TL was located closer to the epicardium than the pre-learning TL. This trend was particularly pronounced in the lateral and the anterior wall. The post-learning TL was along the inner side of the high echoic layer in the basal-lateral wall in the four-chamber view. CT fusion imaging determined that the difference in LV volume between 2D echocardiography and CT was small (-25.6 ± 14.4 mL before learning, -6.9 ± 11.5 mL after learning) and that CT learning improved the coefficient of variation (10.9 % before learning, 7.8 % after learning). Significant improvements were observed during 3D echocardiography; the difference in LV volume between 3D echocardiography and CT was slight (-20.5 ± 15.1 mL before learning, 3.8 ± 15.7 mL after learning), and the coefficient of variation improved (11.5 % before learning, 9.3 % after learning).
Differences between the LV volumes obtained using CT and echocardiography either disappeared or were reduced after CT fusion imaging. Fusion imaging is useful in training regimens for accurate LV volume quantification using echocardiography and may contribute to quality control.
摘要：
背景：我们假设如果使用计算机断层扫描（CT）图像作为学习数据，我们可以通过超声心动图克服体积低估，提高左心室（LV）容积测量的准确性。
方法：我们对37例连续患者使用了一种由超声心动图和叠加CT图像组成的融合成像模式来识别心内膜边界。我们比较了使用和不使用CT学习迹线（TLs）获得的LV体积。此外，3D超声心动图用于比较在有和没有CT学习的情况下获得的LV体积以进行心内膜识别。比较学习前后超声心动图和CT衍生的LV体积之间的平均差异以及变异系数。Bland-Altman分析用于评估从2D学习前TL和3D学习后TL获得的LV体积(mL)的差异。
结果：学习后TL比学习前TL更靠近心外膜。这种趋势在侧壁和前壁中尤其明显。在四腔视图中，学习后TL沿着基底侧壁中高回声层的内侧。CT融合成像确定2D超声心动图与CT的左心室容积差异较小（学习前-25.6±14.4mL，学习后-6.9±11.5mL），并且CT学习提高了变异系数（学习前10.9％，学习后的7.8%）。在3D超声心动图中观察到显着改善；3D超声心动图与CT之间的LV体积差异很小（学习前-20.5±15.1mL，学习后3.8±15.7mL），变异系数提高(学习前11.5%，学习后9.3%)。
结论：使用CT和超声心动图获得的LV体积之间的差异在CT融合成像后消失或缩小。融合成像在使用超声心动图进行准确的LV体积定量的训练方案中很有用，并且可能有助于质量控制。
方法：我们对37例连续患者使用了一种由超声心动图和叠加CT图像组成的融合成像模式来识别心内膜边界。我们比较了使用和不使用CT学习迹线（TLs）获得的LV体积。此外，3D超声心动图用于比较在有和没有CT学习的情况下获得的LV体积以进行心内膜识别。比较学习前后超声心动图和CT衍生的LV体积之间的平均差异以及变异系数。Bland-Altman分析用于评估从2D学习前TL和3D学习后TL获得的LV体积(mL)的差异。
结果：学习后TL比学习前TL更靠近心外膜。这种趋势在侧壁和前壁中尤其明显。在四腔视图中，学习后TL沿着基底侧壁中高回声层的内侧。CT融合成像确定2D超声心动图与CT的左心室容积差异较小（学习前-25.6±14.4mL，学习后-6.9±11.5mL），并且CT学习提高了变异系数（学习前10.9％，学习后的7.8%）。在3D超声心动图中观察到显着改善；3D超声心动图与CT之间的LV体积差异很小（学习前-20.5±15.1mL，学习后3.8±15.7mL），变异系数提高(学习前11.5%，学习后9.3%)。
结论：使用CT和超声心动图获得的LV体积之间的差异在CT融合成像后消失或缩小。融合成像在使用超声心动图进行准确的LV体积定量的训练方案中很有用，并且可能有助于质量控制。
