Identification of patterns of tumour change measured on CBCT images in NSCLC patients during radiotherapy
Identification of patterns of tumour change measured on CBCT images in NSCLC patients during radiotherapy
复制标题
DOI：
10.1088/1361-6560/aba7d3
复制
发表时间：
2020-11-07
影响因子：
3.5
通讯作者：
McWilliam, Alan
中科院分区：
文献类型：
--
作者：
Amugongo, Lameck Mbangula;Osorio, Eliana Vasquez;McWilliam, Alan
在这项研究中，我们提出了一种新的方法来研究可见肿瘤和周围组织的变化，目的是在不分割随访图像的情况下识别放射治疗（RT）期间肿瘤变化的模式。240例接受55 Gy放射治疗的非小细胞肺癌（NSCLC）患者的锥形束计算机断层扫描（CBCT）图像被纳入研究。使用两步刚性配准过程将cbct自动对齐到计划计算机断层扫描（计划CT）上。为了探索肺肿瘤边界的密度变化，创建了八个局限于总肿瘤体积（GTV）形状的贝壳。外壳在GTV边框内外延伸6mm，每个外壳厚1.5 mm。在对cbct进行强度校正后，从所有cbct的每个壳中提取平均强度。然后，建立线性拟合，表明在处理期间每个壳的密度随时间的变化。所有八个外壳的斜坡被聚集在一起，以探索斜坡上的模式，显示肿瘤是如何变化的。共分为7组，97%的患者可分为3组。目视检查后，我们发现这些聚集代表患者很少或没有密度变化，进展和退化。三组间生存曲线差异无统计学意义，p值= 0.51。然而，结果表明肿瘤变化的明确模式存在，这表明可能有可能从治疗时的CBCT图像中识别肿瘤变化的模式。
In this study, we propose a novel approach to investigate changes in the visible tumour and surrounding tissues with the aim of identifying patterns of tumour change during radiotherapy (RT) without segmentation on the follow-up images. On-treatment cone-beam computed tomography (CBCT) images of 240 non-small cell lung cancer (NSCLC) patients who received 55 Gy of RT were included. CBCTs were automatically aligned onto planning computed tomography (planning CT) scan using a two-step rigid registration process. To explore density changes across the lung-tumour boundary, eight shells confined to the shape of the gross tumour volume (GTV) were created. The shells extended 6 mm inside and outside of the GTV border, and each shell is 1.5 mm thick. After applying intensity correction on CBCTs, the mean intensity was extracted from each shell across all CBCTs. Thereafter, linear fits were created, indicating density change over time in each shell during treatment. The slopes of all eight shells were clustered to explore patterns in the slopes that show how tumours change. Seven clusters were obtained, 97% of the patients were clustered into three groups. After visual inspection, we found that these clusters represented patients with little or no density change, progression and regression. For the three groups, the survival curves were not significantly different between the groups, p-value = 0.51. However, the results show that definite patterns of tumour change exist, suggesting that it may be possible to identify patterns of tumour changes from on-treatment CBCT images.