The influence of different ROI delineation strategies for relaxation measurements in nasopharyngeal carcinoma using Synthetic MR imaging
Liangru Ke1, Tie-bao MENG1, Hui-ming LIU1, Long Qian2, Bing Wu2, Hui Li1, Yun He1, Hao-qiang HE1, and Chuan-miao XIE1

1Department of Medical Imaging, Sun Yat-sen University Cancer Center, Guangzhou, China, 2GE Healthcare, MR Research China, Beijing, China


Recently, a novel quantification method named synthetic MRI have attracted more and more attention in the field of clinical research, such as neural disorders and tumor since the first study in 2008. However, this quantitative assessment of diseases based on relaxation times requires regions of interest (ROI), the delineation of which can impact the accuracy of estimated values. To evaluate how the distinct methods of ROI delineation would impact the relaxation value estimation, in current study, 30 patients with nasopharyngeal carcinoma (NPC) were acquired using synthetic MRI.


MRI plays an important role in the diagnosis and treatment of nasopharyngeal carcinoma (NPC). However, its current applications are limited to the recognition of focal size of the lesion and invaded extension to local normal structure, whereas the composition and biological properties of NPC are not often assessed. Synthetic MRI is a novel imaging method that may offer simultaneous relaxation weighted imaging as well as quantitative relaxation mapping. Cases in neural disorders and tumors 1-3 have shown the clinical value of quantitative relaxmeotry measures, however its applications in nasopharyngeal carcinoma (NPC) have rarely been reported. Quantitative assessment usually involves the definition of regions of interest (ROI), whose delineation may impact the accuracy of estimated values. In this study, the effects of different ROI delineation methods in nasopharyngeal carcinoma (NPC) are studied.


A total of 30 participants with histopathology confirmed undifferentiated nonkeratinizing nasopharyngeal carcinoma and nasopharyngeal chronic inflammation were enrolled in this study, ethical approval and consent forms were obtained prior to the study. All of the participants underwent MR exams consisting of both conventional MR sequences as synthetic MRI on a 3.0T whole body scanner (Signa Pioneer, GE, Milwaukee, WI, USA) at the initial diagnosis in both snuff and neck. Three methods of ROI delineation were used and compared in this study, including the whole coverage of the lesion (whole lesion, w), and the maximal lay of the lesion (Single Lay, L), and the rectangular region outlined with the largest left and right diameters as the long sides, and the anterior and posterior diameters as the short side within the Single Lay (Single Voxel, V). All the ROI delineation were performed by an experienced radiologist and the necrosis regions were avoided. Then the mean relaxation values (T1, T2, PD) within the ROIs were obtained. The comparison of all above metrics among the three kinds of ROI delineation were performed using student’s t-test or Wilcoxon Test if the values in different ROI delineation methods showed skewed distribution.


A total of 90 ROIs in 30 patients were obtained in current study (Table 1). Representative images of a nasopharyngeal carcinoma patient with different ROIs are shown in Figure 1. The T1 (1390.6 ± 371.7ms vs 1439.3 ± 386.2ms, P = 0.008), and PD (91.5 ± 4.9ms vs 93.7 ± 4.9ms,P < 0.001) between the Single Lay and Single Voxel, as well as the T2 (93.3 ± 10.7ms vs 95.0 ± 12.4ms, P = 0.082) and PD (91.8 ± 4.3ms vs 93.7 ± 4.9ms, P = 0.007) between the whole lesion and Single Voxel showed significant difference in retropharyngeal lymph nodes. Similarly, The T1 (1288.1 ± 221.4ms vs 1360.9 ± 269.0ms, P = 0.001), T2 (99.8 ± 33.6ms vs 93.90.9ymp, P = 0.011) and PD (84.9 ± 5.8ms vs 87.2 ± 6.7ms, P = 0.002) between the whole lesion and Single Voxel, as well as T1 (1360.9 ± 269.0ms vs 1302.3 245.7ms, P < 0.001) and PD (84.9 ± 5.8ms vs 84.9 ± 6.1ms, P < 0.001) between the Single Lay and Single Voxel were significant in cervical lymph nodes. No significant differences were observed in the measurements with different types of ROI (Table 2).


The differing outcomes observed with using different ROI delineations strategies in lymph nodes and snuff lesion indicate the underlying biological properties are different: region of snuff lesion might be homogeneous that alternating the ROIs did not lead to measurement differences; whereas the different measurements obtained in lymph nodes using different ROIs indicate the complexities and heterogeneities of the inner structures of lymph nodes. Based on the results, whole lesion method might be a more suitable strategy for lymph nodes.


For the use of relaxation maps offered by synthetic MRI in nasopharyngeal carcinoma, suitable ROI delineation strategies need to be selected for ensuring measurements.


No acknowledgement found.


1. Kang, Koung Mi, et al. "T1 Shortening in the Globus Pallidus after Multiple Administrations of Gadobutrol: Assessment with a Multidynamic Multiecho Sequence." Radiology 287.1 (2017): 258-266.

2. Jung, Yongsik, et al. "The feasibility of synthetic MRI in breast cancer patients: comparison of T 2 relaxation time with multiecho spin echo T 2 mapping method." The British journal of radiology 91.xxxx (2018): 20180479.

3. Lee, Seung Hyun, et al. "Quantitative T2 Mapping of Knee Cartilage: Comparison between the Synthetic MR Imaging and the CPMG Sequence." Magnetic Resonance in Medical Sciences (2018): tn-2017.


Table 1: Quantitative values of different lesions in synthetic MRI.

Table 2: Comparison of quantitative values using different methods in contouring ROI.

Figure 1: A to E show the ROI delineation of whole lesion with ROI placed in the consecutive axial sections of the lesion. F and C are the same slice corresponding to the maximal slice of the lesions. C and F represent the ROI delineation of Single Lay and Single Voxel, respectively.

Proc. Intl. Soc. Mag. Reson. Med. 27 (2019)