Atlas-based delineation of lymph node levels in head and neck computed tomography images
Atlas-based delineation of lymph node levels in head and neck computed tomography images
复制标题
DOI：
10.1016/j.radonc.2008.01.018
复制
发表时间：
2008-05-01
影响因子：
5.7
通讯作者：
Malandain, Gregoire
中科院分区：
文献类型：
--
作者：
Commowick, Olivier;Gregoire, Vincent;Malandain, Gregoire
目的：放射治疗计划需要准确描绘肿瘤和关键结构。基于图谱的分割已被证明能够非常有效地自动描绘大脑关键结构。因此，我们建议构建头部和颈部区域的解剖图谱。方法和材料：由于该区域的解剖学变异性很高，从单个图像构建的大脑图谱是不够的。我们通过从手动分割图像的数据库构建对称图集来解决这个问题。首先，我们开发了一种图谱构建方法，并将其应用于包含 45 张计算机断层扫描 (CT) 图像的数据库，这些图像来自手动描绘用于放射治疗的淋巴结阴性咽喉鳞状细胞癌患者。然后，我们在数据库上对基于留一法框架构建的图集生成的结果进行定性和定量评估。结果：我们使用该图集构建方法给出了定性和定量结果。该评估是对 45 名患者的原始 CT 数据库中的 12 名患者进行的。定性结果描绘了视觉上清晰的结构。定量结果也很好，相对于最佳可实现结果的误差范围为 0.196 至 0.404，平均值为 0.253。结论：这些结果表明使用此类图集进行放射治疗计划的可行性。这项工作提出了许多观点，从广泛验证到构建代表亚群的几个图集，以解释患者之间的巨大差异以及淋巴结阳性肿瘤的人群。 (C) 2008 Elsevier Ireland Ltd. 保留所有权利。
Purpose: Radiotherapy planning requires accurate delineations of the tumor and of the critical structures. Atlas-based segmentation has been shown to be very efficient to automatically delineate brain critical structures. We therefore propose to construct an anatomical atlas of the head and neck region.Methods and materials: Due to the high anatomical variability of this region, an atlas built from a single image as for the brain is not adequate. We address this issue by building a symmetric atlas from a database of manually segmented images. First, we develop an atlas construction method and apply it to a database of 45 Computed Tomography (CT) images from patients with node-negative pharyngo-laryngeal squamous cell carcinoma manually delineated for radiotherapy. Then, we qualitatively and quantitatively evaluate the results generated by the built atlas based on Leave-One-Out framework on the database.Results: We present qualitative and quantitative results using this atlas construction method. The evaluation was performed on a subset of 12 patients among the original CT database of 45 patients. Qualitative results depict visually well delineated structures. The quantitative results are also good, with an error with respect to the best achievable results ranging from 0.196 to 0.404 with a mean of 0.253.Conclusions: These results show the feasibility of using such an atlas for radiotherapy planning. Many perspectives are raised from this work ranging from extensive validation to the construction of several atlases representing sub-populations, to account for large inter-patient variabilities, and populations with node-positive tumors. (C) 2008 Elsevier Ireland Ltd. All rights reserved.