The optimal injection dose for imaging of the pelvic region in 3D FDG PET tests was investigated based on the noise-equivalent count (NEC) rate with use of an anthropomorphic pelvis phantom. Count rates obtained from an anthropomorphic pelvis phantom were compared with those of pelvic images of 60 patients. The correlation between single photon count rates obtained from the pelvic regions of patients and the doses per body weight was also evaluated. The radioactivity at the maximum NEC rate was defined as an optimal injection dose, and the optimal injection dose for the body weight was evaluated. The image noise of a phantom was also investigated. Count rates obtained from an anthropomorphic pelvis phantom corresponded well with those from the human pelvis. The single photon count rate obtained from the phantom was 9.9 Mcps at the peak NEC rate. The coefficient of correlation between the single photon count rate and the dose per weight obtained from patient data was 0.830. The optimal injection doses for a patient with weighing 60 kg were estimated to be 375 MBq (6.25 MBq/kg) and 435 MBq (7.25 MBq/kg) for uptake periods of 60 and 90 min, respectively. The image noise was minimal at the peak NEC rate. We successfully estimated the optimal injection dose based on the NEC rate in the pelvic region on 3D FDG PET tests using an anthropomorphic pelvis phantom.