In order to determine the optimized cut-off frequency of the Butterworth filter for cerebral perfusion single photon emission computed tomography (SPECT) images, we compared optimized cut-off frequencies assessed by both frequency and real space domains. SPECT images of a brain phantom were obtained using a SPECT 2000H-4 (Hitachi, Tokyo). The optimized cut-off frequencies were determined for nuclides of 99m-Tc and 123-I, and for LEGP and LEHR collimators, and compared. The radius direction distribution function Pr(n) (power spectrum method) was used to evaluate the SPECT images for frequency space analysis, and the normalized mean squared error (NMSE) method was used for the assessment of images in real space. The optimal cut-off frequencies using the power spectrum method for 99m-Tc were 0.656 cycle/cm for LEGP and 0.802 cycle/cm for LEHR. However, those for 123-I were 0.583 cycle/cm for LEGP and 0.656 cycle/cm for LEHR. In both analyses, the optimal cut-off frequencies varied depending on the nuclides and collimators. The optimal cut-off frequencies assessed by the NMSE method were similar to those assessed by the power spectrum method. In conclusion, the cut-off frequency of the Butterworth filter for cerebral perfusion SPECT images should be changed according to collimators and nuclides.
