Abstract
106
Objectives The objectives are to develop a four-dimensional (4D) PET image reconstruction method with respiratory and cardiac motion compensation and evaluate its performance with patient list-mode data from cardiac PET scans.
Methods The respiratory motion (RM) amplitude signal was first extracted from the list-mode data and used to rebin the data into 6 frames of amplitude-based respiratory-gated sinograms. A maximum likelihood estimation method was applied to the respiratory-gated sinograms to estimate a PET image at a pre-selected reference frame and patient RM from the reference frame to the other respiratory-gated frames. We then utilized the data-driven RM signal and ECG triggers embedded in the list-mode data for simultaneous respiratory and cardiac gating to bin the data into 48 sinograms corresponding to 8 cardiac phases and 6 respiratory amplitudes respectively. For each cardiac phase, we reconstructed a PET image with RM compensation from the 6 respiratory-gated sinograms using a 4D ML-EM image reconstruction algorithm with modeling the previously estimated RM. Then we registered the 8 RM-compensated cardiac-gated PET images to every cardiac phase by a non-rigid image registration method for cardiac motion (CM) compensation. The final outputs of our 4D image reconstruction method were 8 cardiac-gated PET images, each was reconstructed using the entire PET dataset and with both RM and CM compensation. This approach was evaluated with both FDG(n=1) and NH3(n=1) stress studies.
Results The myocardium-to-chamber contrast was improved by 15% using our method compared to a conventional method without any motion compensation. The noise level of each 4D PET image by our method was 60% lower than cardiac-gated images reconstructed by the conventional method. As a result, the cardiac motion defect became more prominent with our method.
Conclusions Our 4D PET image reconstruction method with RM and CM compensation has a great potential to improve the image quaity for cardiac-gated PET imaging .
Research Support NIH research grant R01 EB16