Some physical factors influencing the accuracy of convolution scatter correction in SPECT

Abstract

In these techniques scatter correction in the projection relies on filter functions, Q_F, evaluated by Fourier transforms, from measured scatter functions, Q_p, obtained from point spread functions. The spatial resolution has a marginal effect on Q_p. Thus a single Q_F can be used in the scatter correction of SPECT measurements. However, it is necessary to examine the details of the shape of point spread functions during evaluation of Q_p. Q_F is completely described by scatter amplitude A_F, slope B_F and filter sum S_F. S_F is obtained by summation of the values of Q_F occupying a 31*31 pixels matrix. Regardless of differences in amplitude and slope, two filter functions are shown to be equivalent in terms of scatter correction ability, whenever their sums are equal. On the basis of filter sum, the observed small influence of ellipticity on Q_F implies that an average function can be used in scatter correcting SPECT measurements conducted with elliptic objects. Scatter correction by convolution may be severely hampered by photon statistics when SPECT imaging is done with low-energy photons. Whenever superficial and inner radioactive distributions coexist the observed reduction of S_F close to the phantom surface indicates that scatter correction of such distributions has to rely on two distinct filter functions. Corrections based on a surface function produce accurate results in the superficial region, while the central distributions are substantially overestimated.