MAP transmission image reconstruction via mean field annealing for segmented attenuation correction of PET imaging

In PET studies, accurate attenuation correction is crucial in quantitative studies. We propose to use a MAP reconstruction algorithm based on a mean field annealing technique to reconstruct the transmission image for segmented attenuation correction. This reconstruction method combines both the measured projection data and the a-priori knowledge of the property of the image to be reconstructed. A piecewise-constant prior is assumed such that the reconstructed transmission image consists of smooth homogeneous regions separated by sharp edges. By using this approach, the reconstructed image is less noisy and has sharper edges preserved in comparison with the filtered backprojection (FBP) method. As a result, segmentation of the transmission image is more accurate. By using this calculated attenuation correction approach, the final reconstructed emission image has less statistical noise than the image corrected by the measured attenuation data.