Abstract: To optimize phantom and reconstruction strategy for the best activity quantitative performance of clinical SPECT/CT, three kinds of phantoms (point, small cylinder and big cylinder) and four methods (FBP, OSEM, OSEM+AC and OSEM+AC+SC) were employed to calculate calibration factor respectively, which was then used to reconstructed activity of a small cylinder. The relative errors were calculated by comparing these reconstructed activities with the measured ones from a dose calibrator. The relative quantification errors of OSEM+AC and OSEM+AC+SC are both less than 10%, while those of FBP and OSEM are greater than 20%. When using the big cylinder or the point phantom as calibration phantom, OSEM+AC outperforms OSEM+AC+SC on quantification accuracy of about 3.38% or 6.48%（P<0.05） respectively. The small cylinder phantom, with the smallest quantification error of less than 2%, was found to be the best option. Specifically in using OSEM+AC+SC, difference of mean relative quantification errors between the small cylinder phantom and the big one (or the point one) is 3.92% (or 6.42%, P<0.05). The activity quantification accuracy can be greatly improved using OSEM+AC. However, this is not the case for scatter correction, which may even lead to larger errors. The calibration phantom should be as similar as the object to be quantified for better results. Finally, the proper selection of calibration phantoms and reconstruction methods can effectively reduce the activity quantification error of SPECT/CT.