Phantom study on three-dimensional target volume delineation by PET/CT-based auto-contouring.

The aim of this study was to determine an appropriate threshold value for delineation of the target volume in PET/CT and to investigate whether we could delineate a target volume by phantom studies. A phantom consisted of six spheres (phi 10-37 mm) filled with 18F solution. Data acquisition was performed PET/CT in non-motion and motion status with high 18F solution and in non-motion status with low 18F solution. In non-motion phantom experiments, we determined two types of threshold value, an absolute SUV (T(SUV)) and a percentage of the maximum SUV (T%). Delineation using threshold values was applied for all spheres and for selected large spheres (a diameter of 22 mm or larger). In motion phantom experiments, data acquisition was performed in a static mode (sPET) and a gated mode (gPET). CT scanning was performed with helical CT (HCT) and 4-dimensional CT (4DCT). The appropriate threshold values were aT% = 27% and aT(SUV) = 2.4 for all spheres, and sT% = 30% and sT(SUV) = 4.3 for selected spheres. For all spheres in sPET/HCT in motion, the delineated volumes were 84%-129% by the aT% and 34%-127% by the aT(SUV). In gPET/4DCT in motion, the delineated volumes were 94-103% by the aT% and 51-131% by the aT(SUV). For low radioactivity spheres, the delineated volumes were all underestimated. A threshold value of T% = 27% was proposed for auto-contouring of lung tumors. Our results also suggested that the respiratory gated data acquisition should be performed in both PET and CT for target volume delineation.