Joint maximum likelihood estimation of emission and attenuation densities in PET

Accurate attenuation correction can be performed in PET (positron emission tomography) using transmission scanning to estimate the survival probabilities along each coincidence line. However, since these measurements are typically corrupted by Poisson counting noise, they propagate additional uncertainty into reconstructed images and kinetic parameter estimates. This can be especially true in the thorax where the attenuating medium is heterogeneous and the statistical precision of the transmission scan may be approximately the same as that of the emission data. To account for the Poisson noise in the transmission measurement, the authors have developed a sieve-constrained maximum likelihood algorithm that jointly estimates both the survival probability and the emission intensity. They present some of their initial experiences in using the joint alternate and maximize algorithm with simulated PET data.<>