Co-registration and application of Positron Emission Tomography (PET), Computed Tomography (CT) as sompared with Magnetic Resonance Imaging (MRI) in target localization for undifferentiated Carcinoma of the Nasopharynx (NPC)

Abstract

Purpose/Objective(s): Accurate target localization is always challenging in Intensity Modulated Radiation Therapy (IMRT). The use of multi-modality imaging tools is believed to improve the accuracy of target localization. The application of PET images for target localization is still preliminary and lacking consensus. In this study, we compare the tumor volumes of the primary site and neck nodes of NPC contoured on PET, CT and MRI images independently. Materials/Methods: 32 patients with newly diagnosed NPC treated with IMRT underwent incorporated PET/CT scan in the treatment position in a single session for planning purposes. 18F-fluorodeoxyglucose (FDG) and intravenous contrast were injected during PET and CT scans respectively. MRI scan was also performed without cast but with the position of the head and neck simulating the actual treatment position. The MRI images were then co-registered with the PET/CT images. Gross tumor volumes (GTV) of both the primary site (GTV-P) and the neck nodes (GTV-N) were contoured on the images of PET, CT and MRI independently by the same oncologist without cross-referencing. The window settings of PET images were fixed at 10000Bq/ml (width) and 0Bq/ml (length). The GTVs contoured on PET, CT, PET combined with CT (PET+CT) and MRI were analyzed and compared. Results: The mean volumes of GTV-P contoured on CT, PET, PET+CT and MRI were 22.81cm3, 24.49cm3, 30.30cm3 and 26.81cm3 respectively. The mean volumes of GTV-N on CT, PET, PET+CT and MRI were 20.37cm3, 22.50cm3, 29.09cm3 and 26.22cm3 respectively. Taking MRI as the gold standard, Pearson correlation revealed a strong correlation in the localization of GTV-P by different imaging tools: (1) MRI vs CT (r=0.990, p=0.000), (2) MRI vs PET (r=0.964, p=0.000) and (3) MRI vs PET+CT (r=0.979, p=0.000). Subgroup analysis stratified into advanced (T3 & T4, AJCC 2002) and early T-stage (T1, T2a & T2b, AJCC 2002) showed that high degrees of correlation for GTV-P were still maintained (MR vs CT, r=0.992, p=0.000; MR vs PET, r=0.969, p=0.000; and MR vs PET+CT, r=0.979, p=0.000). A high degree of correlation was also noted in GTV-N (MRI vs CT, r=0.976, p=0.000), (MRI vs PET, r=0.909, p=0.000) and (MR vs PET+CT, r=0.975, p=0.000). The minimum concentration of FDG in localizing both GTV-P (from 6800 to 7300 Bq/ml) and GTV-N (from 6800 to 7500 Bq/ml) were highly consistent and reproducible. Conclusions: The incorporation of PET scan into the standard MRI and CT co-registration is definitely feasible and helpful in tumor localization. Future efforts should be made to derive the most appropriate algorithm in contouring the GTV on PET images.