文献速递骨肌技术与安全CT

利用植入物计算机辅助设计数据作为先验信息的X射线计算机断层扫描金属伪影减少

Metal Artifact Reduction in X-ray Computed Tomography Using Computer-Aided Design Data of Implants as Prior Information.
PubMed 文献 · Invest Radiol · 2017-01-01

本研究探讨以植入物几何与材料的先验信息改进金属伪影减少(MAR)方法,比较基于先验知识的分割(PS)与阈值分割(TS)对线性插值(LI)和归一化MAR(NORMAR)两种方法的影响。体模实验与人体尸体小腿扫描显示,PS使植入物形状估计更准确且不依赖阈值,结构可见性改善、CT值精度提高、图像噪声降低。作者认为基于植入物先验信息的PS图像质量优于基于TS的MAR,尤其当金属植入物形状复杂时。

要点

  • PS利用三维配准的CAD数据估计金属位置与方向,较TS提高LI与NORMAR的MAR效果。
  • PS使植入物形状估计更准确且不依赖阈值,结构可见性改善、CT值精度提高、噪声降低。
  • 作者认为PS图像质量优于TS-based MAR,尤其金属植入物形状复杂时。

核心结论

  • 基于植入物先验信息的PS方法提供的图像质量优于基于TS的MAR,尤其当金属植入物形状复杂时;该新方法可能有助于改进MAR方法及基于MAR校正CT图像的放射治疗剂量计算。

研究设计与样本

  • 体模研究结合人体尸体小腿扫描;用临床CT系统在植入人工膝关节前后扫描;比较PS与TS两种分割对LI和NORMAR两种MAR方法的影响。

关键影像表现/征象

  • PS使LI和NORMAR的效能较TS提高,因分割不足所致的伪影减少。
  • PS使投影数据中获得额外信息,植入物形状估计更准确且不依赖阈值。
  • 与标准方法相比,PS使结构可见性改善,并表现为CT值精度提高和图像噪声降低。

局限性

  • 原文未报告局限性。

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)。

英文原文(摘要)· 点击展开

OBJECTIVES: The performance of metal artifact reduction (MAR) methods in x-ray computed tomography (CT) suffers from incorrect identification of metallic implants in the artifact-affected volumetric images. The aim of this study was to investigate potential improvements of state-of-the-art MAR methods by using prior information on geometry and material of the implant.

MATERIALS AND METHODS: The influence of a novel prior knowledge-based segmentation (PS) compared with threshold-based segmentation (TS) on 2 MAR methods (linear interpolation [LI] and normalized-MAR [NORMAR]) was investigated. The segmentation is the initial step of both MAR methods. Prior knowledge-based segmentation uses 3-dimensional registered computer-aided design (CAD) data as prior knowledge to estimate the correct position and orientation of the metallic objects. Threshold-based segmentation uses an adaptive threshold to identify metal. Subsequently, for LI and NORMAR, the selected voxels are projected into the raw data domain to mark metal areas. Attenuation values in these areas are replaced by different interpolation schemes followed by a second reconstruction. Finally, the previously selected metal voxels are replaced by the metal voxels determined by PS or TS in the initial reconstruction. First, we investigated in an elaborate phantom study if the knowledge of the exact implant shape extracted from the CAD data provided by the manufacturer of the implant can improve the MAR result. Second, the leg of a human cadaver was scanned using a clinical CT system before and after the implantation of an artificial knee joint. The results were compared regarding segmentation accuracy, CT number accuracy, and the restoration of distorted structures.

RESULTS: The use of PS improved the efficacy of LI and NORMAR compared with TS. Artifacts caused by insufficient segmentation were reduced, and additional information was made available within the projection data. The estimation of the implant shape was more exact and not dependent on a threshold value. Consequently, the visibility of structures was improved when comparing the new approach to the standard method. This was further confirmed by improved CT value accuracy and reduced image noise.

CONCLUSIONS: The PS approach based on prior implant information provides image quality which is superior to TS-based MAR, especially when the shape of the metallic implant is complex. The new approach can be useful for improving MAR methods and dose calculations within radiation therapy based on the MAR corrected CT images.

来源:Invest Radiol,查看原文。版权归原作者所有。

引用

  1. Ruth V, Kolditz D, Steiding C, et al. Metal Artifact Reduction in X-ray Computed Tomography Using Computer-Aided Design Data of Implants as Prior Information. Invest Radiol. 2017. PMID: 28106615. doi:10.1097/RLI.0000000000000345

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。