Transmission reconstruction using an ordered-subset algorithm for single photon transmission tomography in PET

The authors have implemented the penalized ordered-subset transmission reconstruction algorithm (OSTR) for the ECAT ART PET scanner, which uses single photon transmission sources rather then positron sources. The OSTR algorithm of Erdogan and Fessler (1999) uses separable surrogates for fast convergence and a more accurate system model for transmission tomography than OSEM. The OSTR algorithm is directly applicable to post-injection transmission scanning with a single photon source, as emission contamination from the patient mimics the effect, in the original derivation of OSTR, of random coincidence contamination in a positron source transmission scan. Multiple noise realizations of simulated post-injection transmission data were reconstructed using OSTR, FBP, and OSEM algorithms. The image noise, contrast, uniformity, and signal-to-noise ratios (SNRs) of the reconstructed images were then calculated. The authors show: (1) the use of a 3D regularizing image roughness penalty with OSTR improves SNR relative to the use of a 2D penalty. (2) OSTR with an 3D penalty has higher SNR than FBP or OSEM. (3) The use of image uniformity (std. dev. from a single realization) to estimate the true SNR can be particularly misleading in the case of OSEM. It is concluded that using OSTR with a 3D penalty allows for shorter transmission scans in single photon transmission tomography in PET, relative to FBP or OSEM reconstructed images with the same image SNR