Wavelet domain ML reconstruction in positive emission tomography

A classic technique for reconstruction of Positive Emission Tomography (PET) images from measured projections is based on the maximum likelihood (ML) parameter estimation along with the Expectation Maximization (EM) algorithm. The authors incorporate the Wavelet transform (WT) into the ML framework, and obtain a new iterative algorithm that incorporates local and multiresolution properties of the WT within the structure of the EM. Using the WT allows one to embed regularization procedures (filtering) into the iterative process, by imposing a new set of parameters on a subset of wavelet coefficients with a desired resolution. Properties of the proposed algorithm are demonstrated on reconstructions of a synthetic brain phantom.