Effects of attenuation in single slow rotation dynamic SPECT

The goal of single slow rotation dynamic SPECT is to reconstruct a dynamic image from data acquired with a single slow rotation of the camera. In this approach, the distribution of activity is assumed to change with every stop made by the camera, and all time frames of the 4D image are simultaneously reconstructed, using a temporal constraint to link them together. The dSPECT algorithm, for instance (Farncombe et al., 2001), restricts the behaviour of the time activity curve (TAC) in every voxel of the image, requiring it either strictly increase, strictly decrease, or increase to a maximum and then decrease. Attenuation correction (AC) is critical to this approach, in order to separate the effects of attenuation on projection data from the actual tracer kinetics. Here we present new analysis indicating that modeling attenuation in the system matrix, although sufficient for AC in conventional (static) SPECT imaging, does not properly account for attenuation effects when reconstructing a dynamic image from data acquired with a single slow rotation. Noticeable artifacts occur in TACs extracted from the reconstructed image as a result. These artifacts correspond to periods of the acquisition where the amount of attenuating material between the camera and dynamic regions changes significantly. We investigate a post-reconstruction approach to correct for these artifacts, in which a template approximating the dynamic image is created, analytically projected, and then reconstructed. This procedure provides an indication of the nature of the artifacts present in the original reconstructed image, allowing for a correction to be applied. An experiment using a digital phantom and the dSPECT reconstruction algorithm indicates that this approach noticeably reduces most attenuation artifacts, although some still persist. Thus, the acquisition protocol in any single slow rotation dynamic SPECT study must be carefully chosen to minimize errors related to attenuation.