The use of maximum a posteriori and maximum likelihood transmission images for attenuation correction in PET

An algorithm for approximating the maximum a posteriori (MAP) estimate of transmission images is proposed. Attenuation correction factors can be formed from these images by reprojecting, sign-inverting and exponentiating them. The authors evaluate the quality of emission images of the myocardium as quantified by the average bias and variance for attenuation correction factors formed from unprocessed transmission scans, smoothed transmission scans. MAP images, and maximum-likelihood (ML) images. The MAP and ML methods yield emission images with significantly lower bias and variance than the other two methods. Moreover, they are relatively insensitive to the correction for accidental coincidences.<<ETX>>