Evaluation of Fourier rebinning of time-of-flight PET data to non time-of-flight

Objectives We previously developed exact and approximate Fourier rebinning methods for TOF PET that substantially reduce the dimensions of the data [1] and result in non-TOF sinograms instead of TOF sinograms in alternative rebinning methods [2,3]. Here we evaluate the performance of these rebinning methods and specifically investigate whether the SNR advantages of the TOF data are retained after rebinning.

Methods : We extended the projection slice theorem to include TOF information and developed exact and approximate rebinnings to map TOF data to lower dimensional non-TOF formats [1]. To evaluate the performance , we simulated a Siemens Biograph PET/CT True Point TrueV scanner with TOF capability, and generated noisy TOF sinograms with 13 TOF bins and 500ps TOF resolution. The sinograms were rebinned by the proposed methods and the approximation errors of the rebinnings quantified. We also conducted a Monte Carlo simulation to characterize the resolution loss through the rebinning process. For the real data, we acquired a 60min torso phantom scan and rebinned the TOF data into 3D non-TOF sinograms.Then we reconstructed the resulting non-TOF sinograms using 3D MAP and compared the results with the images reconstructed from sinograms that are simply summarized over all TOF bins (denoted as snon-TOF).

Results Simulation results showed significant variance reduction using TOF rebinning; Monte Carlo simulation results indicatedimproved resolution vs noise trade-off compared to the snon-TOF case. For the phantom data, 3D MAP reconstruction with approximate Fourier rebinning produced improved SNR compared to the snon-TOF 3D case.

Conclusions Fourier rebinning methods are able to reduce the dimensions of 3D TOF PET data in the TOF variable while still retaining the SNR advantages of TOF. Our imulation and phantom studies support this conclusion.