Frequency analysis of SPECT images

The characteristics of the relationship between image and frequency was discussed on images in the frequency space, which was converted from images in the actual space, from 2-dimensional power spectral distribution. Its application for evaluating the optimal filter for the actual SPECT images was also discussed by quantitative analysis of frequency distribution of the images. Input wave signals can be Fourier transformed to signals in the frequency field, which make it possible to see from what spacial frequency components the input images are composed or to make some treatment on them. The 2-dimensional image function f(x, y) is Fourier transformed to the function F(u, v). The 2-dimensional power spectrum can be described by /F(u, v)/2 and from which images can be evaluated by the following two parameters: Pr(n) and Pθ(n). As the application, frequency analysis of nuclear medicine images and collimator systems was performed by using the parameters: Optimal shielding frequency in the Butterworth filter was examined in the 201Tl and 99mTc myocardial and cerebral blood flow SPECT images with phantoms (Kyoto Kagaku Co.). Further, the examination on the image frequency properties revealed the dependence on the target organ. In conclusion, if the problems of scattering, absorption and partial volume effect are solved, the quantification of SPECT images will further make a progress. (K.H.)