ABSTRACT
Objective To explore an algorithm to define the threshold value for tumor contouring on 18F-fluorodexyglucose (FDG) PET imaging. Methods A National Electrical Manufacturing Association (NEMA)NU 2 1994 PET phantom with 5 spheres of different diameters were filled with 18F-FDG. Seven different sphere-to-background ratios were obtained and the phantom was scanned by Discovery LS 4. For each sphere-to-background ratio, the maximum standardized uptake value ( SUVmax ) of each sphere, the SUV of the border of each sphere ( SUVborder ), the mean SUV of a 1 cm region of background (SUVbg) and the diameter (D) of each sphere were measured. SPSS 13.0 software was used for curve fitting and regression analysis to obtain the threshold algorithm. The calculated thresholds were applied to delineate 29 pathologically confirmed lung cancer lesions on PET images and the obtained volumes were compared with the volumes contoured on CT images in lung window. Results The algorithm for defining contour threshold is TH% = 33.1% + 46.8% SUVbg/SUVmax + 13.9%/D ( r = 0.994) by phantom studies. For 29 lung cancer lesions, the average gross tumor volumes ( GTV ) delineated on PET and CT are ( 7.36 ± 1.62 ) ml and (8.31 ±2.05) ml, respectively (t = -1.26, P＞0.05). Conclusion The proposed threshold algorithm for tumor contouring on PET image could provide comparable GTV with CT.