文献速递胸部肿瘤放疗放疗物理与技术

新型数据驱动四维CT的靶区与伪影评估

Target volume and artifact evaluation of a new data-driven 4D CT.
PubMed 文献 · Pract Radiat Oncol · 2017-02-02

本研究纳入35例接受立体定向体部放疗的患者,将电影CT数据分别用标准四维CT和数据驱动四维CT重建,比较两种方法生成的内大体肿瘤体积(IGTV)。结果显示两种方法IGTV体积差异无统计学意义(数据驱动组小2.1±1.0%,P=.08),Dice相似系数0.949±0.006,数据驱动法伪影评分显著更优(平均+0.37,P=1.5×10⁻⁵)。

要点

  • 数据驱动四维CT与标准四维CT生成的IGTV体积差异无统计学意义(P=.08)。
  • 两种方法轮廓Dice相似系数0.949±0.006,平均表面间距0.4±0.1 mm。
  • 数据驱动法伪影显著更少或更轻(平均评分+0.37,P=1.5×10⁻⁵)。

核心结论

  • 数据驱动四维CT与标准四维CT所得IGTV差异无临床相关性且无统计学意义,可替代标准四维CT而无需调整IGTV。
  • 部分患者差异较大,多归因于自动勾画局限或伪影差异。
  • 伪影减少和摆位简便提示数据驱动四维CT具有临床优势。

研究设计与样本

  • 收集35例接受立体定向体部放疗患者的电影CT数据。
  • 分别用标准四维CT和数据驱动四维CT进行相位排序。
  • 在两种方法的最大密度投影图像上以半自动方法勾画IGTV轮廓。
  • 表征方法重复性所致误差,并用归一化互相关法对相位图像伪影评分(+1为数据驱动更优,-1为标准更优)。

主要发现

  • 数据驱动与标准IGTV体积差异无统计学意义(数据驱动组小2.1±1.0%,P=.08)。
  • Dice相似系数0.949±0.006,提示轮廓相似度良好。
  • 平均表面间距0.4±0.1 mm,Hausdorff距离3.1±0.4 mm。
  • 平均伪影评分+0.37,数据驱动法伪影显著更少和/或更轻(与0比较P=1.5×10⁻⁵)。

关键数值与推荐

  • IGTV体积差异:数据驱动组较标准组小2.1±1.0%(P=.08)。
  • 轮廓相似性:Dice相似系数0.949±0.006,平均表面间距0.4±0.1 mm,Hausdorff距离3.1±0.4 mm。
  • 伪影评分:平均+0.37,与0差异P=1.5×10⁻⁵。

临床怎么用

  • 数据驱动四维CT可替代标准四维CT用于肺癌放疗IGTV定义,无需调整IGTV。
  • 伪影减少和摆位简便支持其临床应用优势。
  • 对个别差异较大患者需注意自动勾画局限或伪影影响。

局限性

  • 部分患者差异较大,归因于自动勾画局限或伪影差异。
  • 方法重复性误差已被表征,但原文未报告具体数值。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

PURPOSE: Four-dimensional computed tomography (4D CT) is often used to define the internal gross target volume (IGTV) for radiation therapy of lung cancer. Traditionally, this technique requires the use of an external motion surrogate; however, a new image, data-driven 4D CT, has become available. This study aims to describe this data-driven 4D CT and compare target contours created with it to those created using standard 4D CT. METHODS AND MATERIALS: Cine CT data of 35 patients undergoing stereotactic body radiation therapy were collected and sorted into phases using standard and data-driven 4D CT. IGTV contours were drawn using a semiautomated method on maximum intensity projection images of both 4D CT methods. Errors resulting from reproducibility of the method were characterized. A comparison of phase image artifacts was made using a normalized cross-correlation method that assigned a score from +1 (data-driven "better") to -1 (standard "better"). RESULTS: The volume difference between the data-driven and standard IGTVs was not significant (data driven was 2.1 ± 1.0% smaller, P = .08). The Dice similarity coefficient showed good similarity between the contours (0.949 ± 0.006). The mean surface separation was 0.4 ± 0.1 mm and the Hausdorff distance was 3.1 ± 0.4 mm. An average artifact score of +0.37 indicated that the data-driven method had significantly fewer and/or less severe artifacts than the standard method (P = 1.5 × 10-5 for difference from 0). CONCLUSIONS: On average, the difference between IGTVs derived from data-driven and standard 4D CT was not clinically relevant or statistically significant, suggesting data-driven 4D CT can be used in place of standard 4D CT without adjustments to IGTVs. The relatively large differences in some patients were usually attributed to limitations in automatic contouring or differences in artifacts. Artifact reduction and setup simplicity suggest a clinical advantage to data-driven 4D CT.

来源:Pract Radiat Oncol,查看原文。版权归原作者所有。

引用

  1. Martin R, Pan T. Target volume and artifact evaluation of a new data-driven 4D CT. Pract Radiat Oncol. 2017. PMID: 28341317. doi:10.1016/j.prro.2017.01.014

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。