Perhaps the simplest approach to scatter correction is to fit an analytical function to the scatter tails outside the object in projection space. For example, a second order polynomial  and a 1D Gaussian  have both been used to fit the scatter tails. This approach is based on the observations that coincidences recorded outside the object boundary are entirely due to scatter (assuming that randoms have previously been subtracted) and the scatter distribution contains mainly low spatial frequencies.
The method is effective for neurological PET studies because it guarantees that the scatter recorded outside the object is reduced to approximately zero and it inherently corrects for scatter arising from activity outside the axial field of view, something some of the more complex methods are unable to do. It also has the advantages of being simple to implement and computationally very efficient. The main drawback of this approach is that the scatter distribution is not always well approximated by a smooth analytical function, particularly in the thorax where tissue density is heteroge-nous, which may lead to over- or under-subtraction. A further problem in the thorax is that the body occupies a large portion of the field of view leaving relatively
small scatter tails to fit. This reduces the accuracy of the fit and may lead to over- or under-subtraction of scatter in the centre of the body.
Was this article helpful?