RIT - a new robust iterative technique for image reconstruction in emission tomography

Emission tomography is a reliable tool for testing nuclear fuel elements. The conventional algebraic reconstruction methods of computer tomography are non-robust and can be fatally affected by outliers in the input data, i.e. by data having unexpectedly high errors due to some unpredictable effects. In the present paper, a robust iterative technique (RIT) for emission tomography is described. RIT is based on robust M-estimation methods and on a new algorithm for computing the M-estimates. RIT needs no filtering of the input data. It's computational expenses do not exceed those of SIRT method. RIT is included in the TOMODAT program and is being used for testing the fuel elements after irradiation in the reactors. The fuel distributions reconstructed by RIT are practically not affected by outliers in the input data, while the ART, SIRT and MENT give quite unstable results for the same spoiled data. (orig.)