Study of the quantification of FBP SPECT images with a correction for partial volume effects

Accurate quantification of emission computed tomography data (PET-SPECT) is limited by partial volume effects. This paper compares two approaches, described in the literature, to measure accurately the true tissue tracer activity within a tissue compartment, defined by anatomical side information. The first approach is based on the selection of an appropriate number of regions. The second method is based on a minimum norm least square solution, taking into account the whole image. These methods assume a constant activity within one tissue compartment and have similar performance when this assumption is correct. We demonstrate the equivalence of both methodologies in the case of an appropriate region selection for the first technique. We also propose two new methods allowing activity fluctuations within the same anatomical tissue compartment. The first technique uses the minimum norm least square solution to estimate a higher number of activities within one tissue compartment while taking into account the whole reconstructed emission computed tomography image. The second method estimates the activity of a tissue compartment locally by linear regression analysis within a sliding window. A simple simulation study shows that these techniques yield a more accurate quantification in the case of a nonhomogeneous activity distribution within one tissue compartment.