Maspero Matteo, Seevinck Peter R, Schubert Gerald, Hoesl Michaela A U, van Asselen Bram, Viergever Max A, Lagendijk Jan J W, Meijer Gert J, van den Berg Cornelis A T
Universitair Medisch Centrum Utrecht, Center for Image Sciences, Heidelberglaan 100, 3584 CX, Utrecht, The Netherlands.
Phys Med Biol. 2017 Feb 7;62(3):948-965. doi: 10.1088/1361-6560/aa4fe7. Epub 2017 Jan 11.
Magnetic resonance (MR)-only radiotherapy treatment planning requires pseudo-CT (pCT) images to enable MR-based dose calculations. To verify the accuracy of MR-based dose calculations, institutions interested in introducing MR-only planning will have to compare pCT-based and computer tomography (CT)-based dose calculations. However, interpreting such comparison studies may be challenging, since potential differences arise from a range of confounding factors which are not necessarily specific to MR-only planning. Therefore, the aim of this study is to identify and quantify the contribution of factors confounding dosimetric accuracy estimation in comparison studies between CT and pCT. The following factors were distinguished: set-up and positioning differences between imaging sessions, MR-related geometric inaccuracy, pCT generation, use of specific calibration curves to convert pCT into electron density information, and registration errors. The study comprised fourteen prostate cancer patients who underwent CT/MRI-based treatment planning. To enable pCT generation, a commercial solution (MRCAT, Philips Healthcare, Vantaa, Finland) was adopted. IMRT plans were calculated on CT (gold standard) and pCTs. Dose difference maps in a high dose region (CTV) and in the body volume were evaluated, and the contribution to dose errors of possible confounding factors was individually quantified. We found that the largest confounding factor leading to dose difference was the use of different calibration curves to convert pCT and CT into electron density (0.7%). The second largest factor was the pCT generation which resulted in pCT stratified into a fixed number of tissue classes (0.16%). Inter-scan differences due to patient repositioning, MR-related geometric inaccuracy, and registration errors did not significantly contribute to dose differences (0.01%). The proposed approach successfully identified and quantified the factors confounding accurate MRI-based dose calculation in the prostate. This study will be valuable for institutions interested in introducing MR-only dose planning in their clinical practice.
仅基于磁共振（MR）的放射治疗治疗计划需要伪CT（pCT）图像来进行基于MR的剂量计算。为了验证基于MR的剂量计算的准确性，有意引入仅基于MR的计划的机构将不得不比较基于pCT和基于计算机断层扫描（CT）的剂量计算。然而，解释此类比较研究可能具有挑战性，因为潜在差异源于一系列混杂因素，这些因素不一定是仅基于MR的计划所特有的。因此，本研究的目的是识别和量化在CT和pCT之间的比较研究中影响剂量准确性估计的因素的贡献。区分了以下因素：成像时段之间的设置和定位差异、与MR相关的几何不准确性、pCT生成、使用特定校准曲线将pCT转换为电子密度信息以及配准误差。该研究包括14名接受基于CT/MRI的治疗计划的前列腺癌患者。为了生成pCT，采用了一种商业解决方案（MRCAT，飞利浦医疗保健公司，芬兰万塔）。在CT（金标准）和pCT上计算调强放疗（IMRT）计划。评估了高剂量区域（临床靶体积）和身体体积中的剂量差异图，并分别量化了可能的混杂因素对剂量误差的贡献。我们发现导致剂量差异的最大混杂因素是使用不同的校准曲线将pCT和CT转换为电子密度（0.7%）。第二大因素是pCT生成，这导致pCT被分层为固定数量的组织类别（0.16%）。由于患者重新定位、与MR相关的几何不准确性和配准误差导致的扫描间差异对剂量差异的贡献不显著（0.01%）。所提出的方法成功地识别和量化了影响前列腺中基于MRI的准确剂量计算的因素。这项研究对于有兴趣在其临床实践中引入仅基于MR的剂量计划的机构将是有价值的。
