Impact of variability in respiration on estimation and correction of respiratory motion of the heart in SPECT

Respiratory motion combined with the "respiratory creep" of the heart can impact the diagnostic accuracy of cardiac PET or SPECT perfusion images. Several groups have investigated the use of registration to estimate respiratory motion between 3D reconstructions of respiratory binned projection data in PET, and recently in SPECT. Motion estimation can be done by first binning the list-mode data into different respiratory gates using an external monitor of respiratory monitor. Amplitude binning has been determined to be more accurate than phase-binning thus in general it is favored. However one issue with amplitude binning in SPECT is that, if the patients' respiratory motion varies greatly with time, it may result in low or no counts in some amplitude bins at some angles. In previous studies of respiratory motion correction in SPECT we investigated 3D registration with phase-binned datasets. With phase-binning no gates at any projection angle would be count-deficit on account of the motion. In this work we investigate the limits on the accuracy of respiratory motion estimation with amplitude binned data when one tries to estimate motion from 3D datasets with reconstruction artifacts due to some of the projection angles having no or few counts. We also investigate correction for respiratory motion within reconstruction when for some of the amplitude binned projection sets have few or no counts. For the purpose we derive and implement a motion-gated MLEM algorithm, where at each iteration we go through all the bins, reconstructing intermediate transformation states of the object. In a future work we will present another derivation where-in we use all the information from the different bins at once and calculate the update equation for a single object.