Deep Learning-based Prescription of Cardiac MRI Planes
Deep Learning-based Prescription of Cardiac MRI Planes
复制标题
DOI：
10.1148/ryai.2019180069
复制
发表时间：
2019-11-01
期刊：
影响因子：
--
通讯作者：
Hsiao, Albert
中科院分区：
文献类型：
--
作者：
Blansit, Kevin;Retson, Tara;Hsiao, Albert
目的：为了开发和评估一个系统，以规定成像平面的心脏MRI基于深度学习（DL）的关键解剖landmarks.Materials and Methods定位：892长轴（LAX）和493短轴（SAX）电影稳态自由旋进序列从心脏MR图像的注释标志回顾性收集2012年2月至2017年6月。基于U-Net的热图回归用于心脏标志的定位，心脏标志用于计算心脏MRI平面。通过比较DL预测和地面实况之间的定位距离和平面角度差异来评估性能。将DL的平面成角与技术人员在原始采集时规定的平面成角进行比较。数据分为80%的训练和20%的测试，结果证实了五重cross-validation.Results：在LAX图像，DL本地化的平均值12.56毫米+/- 19.11（标准差）和二尖瓣（MV）内的顶点7.68毫米+/- 6.91。在SAX图像上，DL将主动脉瓣定位在5.78 mm +/- 5.68 mm内，MV定位在5.90 mm +/- 5.24 mm内，肺动脉瓣定位在6.55 mm +/- 6.39 mm内，三尖瓣定位在6.39 mm +/- 5.89 mm内。基于这些定位，DL预测的成像平面相对于地面实况注释的平均角度偏差和平均误差如下：SAX，-1.27度+/- 6.81和4.93度+/- 4.86;四腔，0.38度+/- 6.45和5.16度+/- 3.80;结论：基于DL的解剖定位是一种可行的心脏MRI平面设计策略。这种方法可以产生与由地面实况界标定义的那些成像平面相当的成像平面。本文有补充材料。(c)RSNA，2019
Purpose: To develop and evaluate a system to prescribe imaging planes for cardiac MRI based on deep learning (DL)-based localization of key anatomic landmarks.Materials and Methods: Annotated landmarks on 892 long-axis (LAX) and 493 short-axis (SAX) cine steady-state free precession series from cardiac MR images were retrospectively collected between February 2012 and June 2017. U-Net-based heatmap regression was used for localization of cardiac landmarks, which were used to compute cardiac MRI planes. Performance was evaluated by comparing localization distances and plane angle differences between DL predictions and ground truth. The plane angulations from DL were compared with those prescribed by the technologist at the original time of acquisition. Data were split into 80% for training and 20% for testing, and results confirmed with fivefold cross-validation.Results: On LAX images, DL localized the apex within mean 12.56 mm +/- 19.11 (standard deviation) and the mitral valve (MV) within 7.68 mm +/- 6.91. On SAX images, DL localized the aortic valve within 5.78 mm +/- 5.68, MV within 5.90 mm +/- 5.24, pulmonary valve within 6.55 mm +/- 6.39, and tricuspid valve within 6.39 mm +/- 5.89. On the basis of these localizations, average angle bias and mean error of DL-predicted imaging planes relative to ground truth annotations were as follows: SAX, -1.27 degrees +/- 6.81 and 4.93 degrees +/- 4.86; four chambers, 0.38 degrees +/- 6.45 and 5.16 degrees +/- 3.80; three chambers, 0.13 degrees +/- 12.70 and 9.02 degrees +/- 8.83; and two chamber, 0.25 degrees +/- 9.08 and 6.53 degrees +/- 6.28, respectively.Conclusion: DL-based anatomic localization is a feasible strategy for planning cardiac MRI planes. This approach can produce imaging planes comparable to those defined by ground truth landmarks. Supplemental material is available for this article. (c) RSNA, 2019