【作者】 王孝深； 胡超苏； 邱健健； 陆惠忠； 李龙根； 冯炎；

【Author】 Wang Xiao-Shen Hu Chao-Su Qiu Jian-Jian Lu Hui-Zhong Tang Feng Li Long-Gen Feng Yan Departmentof Radiation Oncology, Cancer Hospital, Fudan University, Shanghai 200032

【机构】 复旦大学附属肿瘤医院放射治疗科；

【摘要】 目的:探讨三种CT／MRI融合方法在鼻咽癌中的应用,并比较三种方法的融合效果。方法:10例鼻咽癌患者模拟放射治疗的体位、采用面罩固定分别进行CT增强扫描和MRI扫描,CT扫描范围:颅顶到锁骨头下缘,鼻咽和颅底区域层厚3mm,其他区域层厚5mm;MRI横断面扫描范围中脑到第四颈椎,进行T1加权、T2加权、T1加权增强脂肪抑制序列扫描,层厚5mm。全部图像经网络传送到治疗计划工作站,分别采用标记点法、交互法和归一化互信息法来进行CT／MRI融合。记录达到满意融合效果时花费的时间;测量X轴和Y轴方向上CT和MRI图像上相同两个解剖点的距离差异作为融合误差;分别根据CT和MRI图像勾画上颌窦的轮廓,记录上颌窦中心坐标的差异。采用方差分析来比较三种融合方法花费的时间,X、Y坐标方向上的融合误差,上颌窦中心坐标的差异。结果:三种融合方法花费的平均时间分别为7．15±1．47min、5．30±1．08min和5．15±0．91min, 标记点法花费的时间最长(P=0．001),交互法和归一化互信息法花费的时间没有统计学差异(P=0．78)。三种方法在x轴方向的平均融合精度均为0．64±0．25mm,(P=0．97),三种方法 Y轴方向的平均融合精度分别为1．16±0．13mm,1．14±0． 14mm,1．15±0．13mm(P=0．87)。上颌窦中心坐标在x轴方向的差异分别为0．39±0．14mm,0．38±0．15mm,0．37± 0．14mm(P=0．97);Y轴方向的差异分别为0．55±0．12mm, 0．54±0．13mm,0．54±0．12mm(p=0．93);Z轴方向的差异分别为0．75±0．79mm,0．60±0．77mm,0．45±0．72mm(P =0．68)。结论:标记点法、交互法和归一化互信息法都能达到满意的CT／MRI融合效果,标记点法花费的时间长。更多还原

【Abstract】 Objective: To explore and evaluate the applicability of three methods for CT and MRI registration in nasopharyngeal carcinoma (NPC), and to compare the registration effects of the three. Methods: Contrast enhanced CT and MRI scans of 10 NPC patients were carried out in treatment position with immobilized mask. The range of CT scans extended from the scalp to the inferior border of the clavicle, the thickness was 3mm per slice at the area of skull base and nasopharynx, and 5mm per slice at the rest area. And the range of MRI scans extended from the midbrain to the fourth cervical vertebra, the thickness was 5mm per slice. T1 weighted, T2 weighted, and T1 weighted contrast enhanced fat suppression imaging was performed. All images were transformed to the treatment planning system, and undergone CT/MRI registration using the following three methods, respectively: ① Links and Marks, ② Interactive, ③ Normalized Mutual Information. The time needed to complete CT/MRI registration was recorded. Distances between two anatomically identical points on CT and MRI were measured both at the X axis and Y axis, and differences between them was taken as registration error. Contours of the maxillary sinus was delineated both on CT and MRI, differences between coordinates of the maxillary center was recorded. One-way ANOVA was used to test the differences among the time, registration errors at both the X axis and Y axis, as well as coordinates of the center of the maxillary sinus. Results: The mean time needed for the three methods to complete CT/MRI registration was 7.15 ± 1.47min, 5.30 ± 1.08min, and 5.15 ± 0.91min, respectively. Time needed for Links and Marks method was the longest(P =0.001), while there was no difference between the time needed for Interactive and Normalized Mutual Information method(P=0.78). Mean registration error of the three methods at X axis was all 0.64 ± 0.25mm ( P=0.97), and 1.16 ± 0.13mm, 1.14 ± 0.14mm, 1.15 ± 0.13mm, respectively at Y axis( P=0.87 ). Differences between coordinates of the center of the maxillary sinus at X axis was 0.39 ± 0.14mm, 0.38 ± 0.15mm, and 0.37 ± 0.14mm, respectively ( P=0.97 ), and 0.55 ± 0.12mm, 0.54 ± 0.13mm, 0.54 ± 0.12mm at Y axis ( P= 0.93 ), and 0.75 ± 0.79mm, 0.60 ± 0.77mm, 0.45 ± 0.72mm at Z axis (P=0.68 ). Conclusion: All the three registration methods-Marks and Links, Interactive, Normalized Mutual Information-can achieve good CT/MRI registration result But more time is needed for the Marks and Links method.更多还原