Objective: In the process of precise treatment of nasopharyngeal carcinoma,the requirements for tumor control and reduction of organ damage are higher and higher.Therefore,the outline of gross tumor area combined with multimodal images(CT,MRI,PET-CT)is very important.Multimodal images can improve the precision of target delineation,and the application of radiomics can further analyze the target.This paper aims to:1.To compare the difference of gross tumor volume(GTV)and lymph node of nasopharyngeal carcinoma(NPC)on computerized tomography(CT),magnetic resonance imaging(MRI)and 18F-fluorodeoxygiucose positron emission tomography(18F-FDG PET-CT),and to investigate the optimal standard uptake value(SUV)for gross tumor target on PET referring to MRI.2.To compare the imaging features of radiomics of nasopharyngeal carcinoma(NPC)gross tumor volume in CT,MRI and PET-CT imaging,and analyze the differences of different imaging features in multimodal images,so as to provide the basis for multimodal imaging guided accurate radiotherapy feature selection of NPC.Methods:1.Totally Fifty-three patients with NPC who accepted radiation therapy were selected.The CT simulation,MRI simulation and PET images of each patient were obtained before radiotherapy.GTV and positive lymph nodes,named GTVMRI,GTVCT,GTVPET2.5(SUV=2.5),LymphMRI,LymphCT,LymphPET2.5,were contoured on CT,MRI,PET-CT respectively,and the differences of GTVs were compared.The GTV∩2.5(intersecting region)was obtained on MRI and PET-CT registration.The GTVPET4.0,GTVPET4.5,GTVPET5.0,GTVPET5.6 were obtained on PET with SUV = 4.0,4.5,5.0,5.6.The volume and Dice similarity coefficient(DSC)were compared.2.Fifty-three patients with nasopharyngeal carcinoma were selected.The gross tumor volumes were delineated in plain CT scan,enhanced CT scan,MR-T1 weighted image,MR-T2 weighted image and 18F-FDG PET-CT image,which were named GTVCT1,GTVCT2,GTVMR1,GTVMR2,GTVPET4.0 respectively.The features First Order,GLCM(Gray Level Co-occurrence Matrix),GLDM(Gray Level Dependence Matrix),GLRLM(Gray Level Run Length Matrix),GLSZM(Gray Level Size Zone Matrix),NGTDM(Neighbouring Gray Tone Difference Matrix)were extracted by using 3D slicer,and the maximum,minimum values and line chart were obtained after statistical analysis.Results:1.Compared to GTVMRI,GTVCT reduced by 1.73%(t=0.26,P > 0.05),while GTVPET2.5 increased by 21.34%(t=-3.52,P < 0.05).LymphPET2.5 was 1.61 and 1.87 times to LymphMRI and LymphCT respectively(t=-4.12、-5.18,P < 0.05).The volume of high SUV lymph nodes was 4.07 times to non-or low SUV lymph nodes(t=5.50,P < 0.05)on PET.The DSC of GTVPET4.0 and GTVMRI was 0.78 ± 0.27,lower than GTVPET2.5 and GTVMRI(0.84 ± 0.18).GTVPET4.0 and GTV∩2.5 were similar(t=1.42,P > 0.05).2.According to the sequence of CT plain scan,CT enhancement,MR-T1 weighted image,MR-T2 weighted image and PET image,four kinds of curve types were obtained(gradually rising type,gradually falling type,first falling and then rising type,first rising and then falling type).Gradually rising type contains the most features(72)and first falling and then rising typecontains the least features(7).In GTVCT1,GTVCT2,and GTVPET4.0,the largest feature value was First Order Total Energy.The largest feature value of GTVMR1 and GTVMR2 was First Order Energy.And those maximum values were 3.6(9)105,4.5(9)105,8.01(9)1011,3.53(9)1011 and 2.95(9)1011 times as big as the minimum values,respectively;The curve trends of CT,MRI and PET are different,but the trends of different sequences in same imaging are similar.The largest values in Firstorder,GLCM,GLDM,GLRLM,GLSZM and NGTDM were found in Total Energy,Cluster Prominence,Large Dependence High Gray Level Emphasis,Long Run High Gray Level Emphasis,Large Area High Gray Level Emphasis,Complexity respectively.Conclusion:1.Compared to CT and 18F-FDG PET-CT,MRI holds favorable boundary display for GTV and lymph node.When GTV was automatically delineated on 18F-FDG PET-CT with SUV=4.0,the GTV was more closer on GTVMRI.2.Feature value of radiomics are significant differences in different imaging,so the specific features should be selected based on specific images in precise radiotherapy of nasopharyngeal carcinoma. |