Preprocessing of Spect Data as a Precursor for Attenuation Correction

Abstract

Almost all systems available for performing single photon emission computed tomography (SPECT) reconstruct images by filtered backprojection, and perform some kind of attenuation correction. These systems give qualitatively reasonable images. However, for the purposes of quantitation, the results are far from satisfactory. Specifically, it may be observed that, often, when procedures designed to perform attenuation correct are tested and evaluated, usually simulated data is employed (as in [1] for example). The attenuation at the centre of such simulated objects is much greater than that observed in reality (see Fig 16). Alternatively, the results of correcting for attenuation in a real uniform phantom are studied [2]. Such tests are inadequate since a variety of physical effects such as the presence of scatter, and the effect of the collimator response function, have been ignored. It may also be observed that, firstly, such procedures do not work if the known linear attenuation coefficient μ for the photon energy of the isotope used (e.g 0.15cm^-1 for 140Kev for Tc99M) is employed, and, secondly, tend to over- or under-correct depending on the form of the activity distribution.