Reconstruction of PET images using anatomical adaptive parameters and hybrid regularization

Abstract

Positron Emission Tomography (PET) is a nuclear medicine technique used to obtain metabolic images of the body. PET scanners used in the research, treatment, and monitoring of several diseases provide images of metabolic activity associated with the ailments. However, the data produced by PET are heavily corrupted by noise and other errors, thereby causing degradation in the quality of the final reconstructed images. In order to improve the image reconstruction process, this paper presents a new algorithm that addresses the problem from a variational perspective. We propose the use of a modified version of total variation regularization by including a second term in order to better deal with noise; in the proposed version, both regularizing terms are balanced by calculating weights adapted to the PET images through the use of anatomical information from another medical modality, such as computer tomography (CT) or magnetic resonance imaging (MRI). Simulated image results show that our proposed method is more effective in dealing with heavy noise and in preserving small structures (e.g., possible lesions) than the expectation maximization method that is commonly used with commercial scanners