Background: Previous studies have shown that SPECT camera coupled with Wire Mesh Collimator (WMC) has good sensitivity in detecting breast cancer. In this paper, we intend to improve the breast cancer detection through enhancing images by characterized Butterworth filter. The characterization was based on trade-off between noise suppression and spatial resolution degradation. Methods: Monte Carlo simulation studies were performed to obtain projections for a half-ellipsoidal breast phantom in the prone position with five different tumor sizes, and four projection sets with acquisitions of 20, 40, 80, 160 seconds were simulated which gave different count density in the range of 28 to 229 counts/cm 2 for each tumor size. Butterworth filter was used to filter the sinograms with 13 different cutoff frequency (Fc), ranging from 0.2 to 0.8 Nyquist frequency (Nq) with 0.05 steps. The relationship between optimum Fc, count density and tumor size were revealed. Results:The results showed that the optimum Fc not only depended on count density but also linked to the change of tumor size. For the case of tumor size, it is suggested that the Fc can be relocated to a higher spatial frequency when tumor size was close to the spatial resolution of SPECT system in order to preserve tumor signal. Conclusion:It may be possible to use a priori knowledge of tumor size and count density as guideline for choosing the optimal parameters of Butterworth filter in Tc-99m breast SPECT imaging.