Defining PET tumor volume in cervical cancer with hybrid PET/MRI

Abstract

The aim of the study was to determine the standardized uptake value (SUV) threshold on gross tumor volume from PET images (PET-GTV) that generates the maximal spatial overlap with MRI-defined GTV (MR-GTV) in locally advanced cervical cancer with the aid of hybrid PET/MRI. Twenty-seven patients with cervical carcinomas underwent F-FDG PET/MRI before radiotherapy. MR-GTV was manually outlined on T2-weighted MRI. PET-GTVs were contoured on PET images using visual inspection (PETvis-GTV), 2.5 SUV (PET2.5-GTV) and 40% SUVmax threshold (PET40-GTV), and compared with MR-GTV. All PET data were reviewed again. The PET-GTV was determined, with the SUVmax threshold defined as the value at which the Dice similarity coefficient (DSC) between PET-GTV and MR-GTV was maximal (PETdsc-GTV). Results were analyzed by Pearson's analysis, the t-test, and one-way analysis of variance. The mean PETvis-GTV, PET40-GTV, and PET2.5-GTV were significantly different from the mean MR-GTV. The average DSCs were 0.64, 0.65, and 0.68 for PETvis-GTV, PET40-GTV, and PET2.5-GTV with MR-GTV, respectively. The mean threshold of PETdsc-GTV was 29.4% SUVmax, with a maximal DSC value of 0.72 in all patients. PET-GTVs as defined by thresholds of 25 and 30% of SUVmax were not statistically different from the MR-GTV. The mean thresholds of PETdsc-GTV were 43.3±8.8, 27.9±5.8, and 22.2±2.6% for MR-GTV<14, 14-62, and ≥62 cm, respectively. PETvis-GTV, PET2.5-GTV, and PET40-GTV do not appear to be suitable for target volume delineation in locally advanced cervical cancer. The PETdsc-GTV may increase the accuracy in target volume delineation. Furthermore, a high level of tumor overlap existed between MR-GTV and PETdsc-GTV in hybrid PET/MRI.