A recursive algorithm for quantification of brain perfusion SPECT images

The accurate quantification of emission computed tomography data (PET-SPECT) is limited by partial volume effects (PVE). The influence of these effects on quantification depends on the size of the structure in relation to the spatial resolution and may lead to an over- and/or underestimation of the true tracer concentration. This particularly holds for brain perfusion SPECT imaging, where white matter and grey matter are two entangled tissue structures with different tracer uptake. This study proposes a recursive algorithm, which uses anatomical side information, to estimate the true tracer activity within the white matter tissue compartment which is needed to correct for partial volume effects. The algorithm first estimates the activity within the white matter and grey matter tissue compartment globally, taking into account the whole SPECT-image. The fitted activity values are then compared with the reconstructed SPELT activity values. For over- and underestimated regions, the activities of the two tissue compartments are estimated again locally, where only the reconstructed activity values within these regions are considered. Next, these fitted activity values are again compared with the reconstructed SPECT activity values. If necessary these regions are again divided in an over- and/or underestimated region, for which the estimation is redone and so on. Simulation studies show that this algorithm is capable of addressing non-homogenous activity distribution within the grey matter tissue compartment for brain perfusion SPECT-images.