New Component-Based Normalization Method to Correct PET System Models

Abstract

Normalization correction is necessary to obtain high quality reconstructed images in positron emission tomography (PET). There are two types of normalization methods, the direct method and component-based methods. The former has a problem that a huge count number in the blank scan data is required. Therefore, the latter have been proposed to obtain high statistical accuracy normalization coefficients with a small count number in the blank scan data. In iterative image reconstruction methods, on the other hand, the quality of obtained reconstructed images depends on the system modeling accuracy. Therefore the normalization weighing approach in which normalization coefficients are directly applied to the system matrix instead of a sinogram has been proposed. In this paper, we proposed a new component-based normalization method to correct system model accuracy. In the proposed method, two components are defined which are calculated iteratively in such a way as to minimize errors of system modeling. For comparison of the proposed method and the direct method, we applied both methods to our small OpenPET prototype. We achieved acceptable statistical accuracy of normalization coefficients while reducing the count number of the blank scan data to one-fortieth of the count number of the blank scan data used in the direct method.