Median root prior and ordered subsets in Bayesian image reconstruction of single-photon emission tomography

Median root prior allows Bayesian image reconstruction without any a priori knowledge of the final solution. It limits the noise generated by maximum likelihood-expectation maximization, including when the ordered subsets accelerating procedure is used. Therefore the number of iterations can be optimized to obtain the best resolution for cold lesions. Moreover, the higher the number of subsets, the better the contrast, with optimal results for subsets containing between four and eight projections.