Objective To compare the differences in 4 sets of gross tumor volumes (GTVs) generated from FDG PET/CT in cervical cancer,and determine the optimal method for target volume delineation of cervical cancer.Methods Sixteen cervical cancer patients with 28 primary or metastatic lesions underwent FDG PET/CT.CT and PET images were coregistered,and transferred to Pinnacle therapy planning workstation.Four sets of GTVs were defined.The first set ( GTVvis ) was manually contoured using a visual method on PET images.The second set ( GTV40 ) was autocontoured using a threshold of 40％ of the maximum intensity level for PET images.The third set ( GTV2.5 ) used an autocontour of standardized uptake value (SUV) of 2.5 around the tumor.By phantom measurements we determined an algorithm ( threshold =(mean target concentration + 2.623)/1.975),GTVfunction was defined using this method as the fourth approach.The volumes of 4 sets of GTVs were compared with group t-test.ResultsThe average volumes of GTVvis,GTV2.5,GTV40,and GTVfunction were 63.41,53.20,41.33,and 61.84 cm3. There was no significant difference between GTVfunctoon,GTV2.5 and GTVvis ( t =1.05,0.91,P =0.305,0.37 ),but GTV40were smaller than GTVvis.The SUVmax and target to background value had no significant influence on the differences between GTV40 and GTVfunction or GTVvis ( t =0.00,- 0.34,0.92,0.35,P =1.000,0.746,0.374,0.737),but they had significant influence on the difference between GTV2.5 and GTVvis (t =- 3.87,3.16,P =0.002,0.016).ConclusionsGTVvis,GTV2.5 and GTVfunction all could be used for target delineation if the method can define the GTV.The difference between the GTVfunction and GTVvis was the smallest;GTV40 was smaller than GTVvis. GTV2.5 was significantly influenced by SUVmax and target to background value of the legions.