Deschamps Frederic, Solomon Stephen B, Thornton Raymond H, Rao Pramod, Hakime Antoine, Kuoch Viseth, de Baere Thierry
Service de Radiologie Interventionnelle, Institut Gustave Roussy, 94805, Villejuif, France,
Cardiovasc Intervent Radiol. 2010 Dec;33(6):1235-42. doi: 10.1007/s00270-010-9846-6.
The purpose of this study was to evaluate computed analysis of three-dimensional (3D) cone-beam computed tomography angiography (CTA) of the liver for determination of subsegmental tumor-feeding vessels (FVs). Eighteen consecutive patients underwent transarterial chemoembolization (TACE) from January to October 2008 for 25 liver tumors (15 hepatocellular carcinomas [HCCs] and 10 neuroendocrine metastases). Anteroposterior projection angiogram (two-dimensional [2D]) and 3D cone-beam CTA images were acquired by injection of the common hepatic artery. Retrospectively, FVs were independently identified by three radiology technologists using a software package (S) that automatically determines FVs by analysis of 3D images. Subsequently, three interventional radiologists (IRs) independently identified FVs by reviewing the 2D images followed by examination of the 3D images. Finally, the "ground truth" for the number and location of FVs was obtained by consensus among the IRs, who were allowed to use any imaging-including 2D, 3D, and all oblique or selective angiograms-for such determination. Sensitivities, durations, and degrees of agreement for review of 2D, 3D, and S results were evaluated. Sensitivity of 3D (73%) was higher than 2D (64%) images for identification of FVs (P = 0.036). The sensitivity of S (93%) was higher than 2D (P = 0.02) and 3D (P = 0.005) imaging. The duration for review of 3D imaging was longer than that for 2D imaging (187 vs. 94 s, P = 0.0001) or for S (135 s, P = 0.0001). The degree of agreement between the IRs using 2D and 3D imaging were 54% and 62%, respectively, whereas it was 82% between the three radiology technologists using S. These preliminary data show that computed determination of FVs is both accurate and sensitive.
本研究的目的是评估肝脏三维（3D）锥形束计算机断层扫描血管造影（CTA）的计算机分析，以确定肝段下肿瘤供血血管（FVs）。2008年1月至10月，连续18例患者因25个肝脏肿瘤（15个肝细胞癌[HCCs]和10个神经内分泌转移瘤）接受了经动脉化疗栓塞术（TACE）。通过注入肝总动脉获取前后位投影血管造影（二维[2D]）和3D锥形束CTA图像。回顾性地，三名放射技术人员使用一个软件包（S）独立识别FVs，该软件包通过分析3D图像自动确定FVs。随后，三名介入放射科医生（IRs）通过查看2D图像并随后检查3D图像来独立识别FVs。最后，通过IRs之间的共识获得FVs数量和位置的“真实情况”，允许他们使用任何成像方式——包括2D、3D以及所有斜位或选择性血管造影——进行此类确定。评估了2D、3D和S结果审查的敏感性、持续时间和一致程度。在识别FVs方面，3D（73%）的敏感性高于2D（64%）图像（P = 0.036）。S（93%）的敏感性高于2D（P = 0.02）和3D（P = 0.005）成像。3D成像审查的持续时间比2D成像（187秒对94秒，P = 0.0001）或S（135秒，P = 0.0001）更长。使用2D和3D成像的IRs之间的一致程度分别为54%和62%，而使用S的三名放射技术人员之间的一致程度为82%。这些初步数据表明，FVs的计算机确定既准确又敏感。
