Department Electronics and Computer Science, University of Santiago de Compostela, Santiago de Compostela, Spain.
CTIM, DIS, Universidad de Las Palmas de Gran Canaria, Las Palmas de Gran Canaria, Spain.
Med Biol Eng Comput. 2020 May;58(5):903-919. doi: 10.1007/s11517-019-02110-x. Epub 2020 Feb 19.
Computed tomography angiography (CTA) is one of the most common vascular imaging modalities. However, for clinical use, it still requires laborious manual analysis. This study demonstrates the feasibility of a fully automated technology for the accurate detection and identification of several anatomical reference points (landmarks), commonly used in intravascular imaging. This technology uses two different approaches, specially designed for the detection of aortic root and supra-aortic and visceral branches. In order to adjust the parameters of the developed algorithms, a total of 33 computed tomography scans with different types of pathologies were selected. Furthermore, a total of 30 independently selected computed tomography scans were used to assess their performance. Accuracy was evaluated by comparing the locations of reference points manually marked by human experts with those that were automatically detected. For supra-aortic and visceral branches detection, average values of 91.8 % for recall and 98.8 % for precision were obtained. For aortic root detection, the average difference between the positions marked by the experts and those detected by the computer was 5.7 ± 7.3 mm. Finally, diameters and lengths of the aorta were measured at different locations related to the extracted landmarks. Those measurements agreed with the values reported by the literature. Graphical abstract Schematic description of the proposed algorithm. The input includes an already segmented aorta (left), there are two main sub-processes related to the detection of branches and roots (center), and the output includes the segmented original aorta with the branches and the detected landmarks superimposed (right).
计算机断层血管造影 (CTA) 是最常见的血管成像方式之一。然而，在临床应用中，它仍然需要繁琐的手动分析。本研究展示了一种全自动技术用于准确检测和识别几种常用于血管内成像的解剖参考点（标志物）的可行性。该技术使用两种不同的方法，专门用于检测主动脉根部和主动脉上及内脏分支。为了调整开发算法的参数，总共选择了 33 个具有不同类型病变的计算机断层扫描。此外，总共选择了 30 个独立的计算机断层扫描来评估它们的性能。通过将人工专家手动标记的参考点的位置与自动检测到的位置进行比较来评估准确性。对于主动脉上及内脏分支的检测，召回率的平均值为 91.8%，精确度的平均值为 98.8%。对于主动脉根部的检测，专家标记的位置与计算机检测到的位置之间的平均差异为 5.7±7.3mm。最后，在与提取的标志物相关的不同位置测量了主动脉的直径和长度。这些测量值与文献中报道的值一致。
