Shaomin Zhang1, Jun Xin, Hongzan Sun, Jietao Ma, Quanmei Ma, Qiyong Guo, Xuna Zhao. 1. aSchool of Computer Science and Engineering, Beifang University of Nationalities, Yinchuan Departments of bRadiology cOncology, Shengjing Hospital of China Medical University, Shenyang dPhilips Healthcare, Beijing, China.
Abstract
PURPOSE: This study aimed to evaluate the accuracy of sequential whole-body PET/MR image coregistration of cervical lesions. MATERIALS AND METHODS: Twenty-five patients with cervical carcinomas underwent fluorine-18 fluorodeoxyglucose PET/MR before radiotherapy. MR-volumes of interest (VOIs), PET-VOIs, and apparent diffusion coefficient (ADC)-VOIs were outlined manually on T2-weighted MR images, PET images, and ADC maps. The difference between the lesion centers on PET and MR was determined by calculating the distance of the respective geometric center of gravity. In addition, the tumor volume contour differences were assessed using the dice similarity coefficient for PET and ADC. Results were analyzed by mean±SD and a two-sample t-test. RESULTS: The mean values of the center of gravity mismatch were relatively higher with lesions between PET and ADC (5.79±1.70 mm) than that between PET and MR-T2 (5.22±1.97 mm; P=0.304). Tumor location overlap difference between MR-T2 and PET images (0.64±0.13) was larger than that between ADC and PET (0.56±0.14; P=0.054). The average differences between the centers of lesions on PET and T2-weighted images were 6.25±1.91, 5.24±2.17, and 4.30±1.30 mm for MR-VOI less than 14, 14-62, and at least 62 ml. The average differences between the center of lesions on PET and ADC were 5.97±1.48, 5.43±1.40, and 5.78±2.75 mm, respectively. Image registration tended to be slightly less accurate in the smaller lesions than in the larger lesions (P>0.05). The average overlaps were 0.51±0.13, 0.63±0.10, and 0.76±0.03 between the T2-weighted image and PET, respectively. The average overlaps were 0.44±0.14, 0.58±0.11, and 0.66±0.04 between the ADC and PET, respectively. Larger tumors had a higher degree of overlap compared with small tumors (P<0.05). CONCLUSION: Image coregistration of cervical lesions is usually accurate in sequential whole-body PET/MR. The accuracy of image registration between MR-T2 and PET was larger than that between ADC and PET. Image registration tended to be more accurate in the larger lesions.
PURPOSE: This study aimed to evaluate the accuracy of sequential whole-body PET/MR image coregistration of cervical lesions. MATERIALS AND METHODS: Twenty-five patients with cervical carcinomas underwent fluorine-18 fluorodeoxyglucose PET/MR before radiotherapy. MR-volumes of interest (VOIs), PET-VOIs, and apparent diffusion coefficient (ADC)-VOIs were outlined manually on T2-weighted MR images, PET images, and ADC maps. The difference between the lesion centers on PET and MR was determined by calculating the distance of the respective geometric center of gravity. In addition, the tumor volume contour differences were assessed using the dice similarity coefficient for PET and ADC. Results were analyzed by mean±SD and a two-sample t-test. RESULTS: The mean values of the center of gravity mismatch were relatively higher with lesions between PET and ADC (5.79±1.70 mm) than that between PET and MR-T2 (5.22±1.97 mm; P=0.304). Tumor location overlap difference between MR-T2 and PET images (0.64±0.13) was larger than that between ADC and PET (0.56±0.14; P=0.054). The average differences between the centers of lesions on PET and T2-weighted images were 6.25±1.91, 5.24±2.17, and 4.30±1.30 mm for MR-VOI less than 14, 14-62, and at least 62 ml. The average differences between the center of lesions on PET and ADC were 5.97±1.48, 5.43±1.40, and 5.78±2.75 mm, respectively. Image registration tended to be slightly less accurate in the smaller lesions than in the larger lesions (P>0.05). The average overlaps were 0.51±0.13, 0.63±0.10, and 0.76±0.03 between the T2-weighted image and PET, respectively. The average overlaps were 0.44±0.14, 0.58±0.11, and 0.66±0.04 between the ADC and PET, respectively. Larger tumors had a higher degree of overlap compared with small tumors (P<0.05). CONCLUSION: Image coregistration of cervical lesions is usually accurate in sequential whole-body PET/MR. The accuracy of image registration between MR-T2 and PET was larger than that between ADC and PET. Image registration tended to be more accurate in the larger lesions.