A Probabilistic Patch Based Hybrid Framework for CT/PET Image Reconstruction

Abstract

Statistical image reconstruction for computed tomography and positron emission tomography (CT/PET) play a significant role in the image quality by using spatial regularization that penalizes image intensity difference between neighboring pixels. The most commonly used quadratic membrane (QM) prior, which smooth’s both high frequency noise and edge details, tends to produce an unfavourable result while edge-preserving non-quadratic priors tend to produce blocky piecewise regions. However, these edge-preserving priors mostly depend on local smoothness or edges. It does not consider the basic fine structure information of the desired image, such as the gray levels, edge indicator, dominant direction and frequency. To address the aforementioned issues of the conventional regularizations/priors, this paper introduces and evaluates a hybrid approach to regularized ordered subset expectation maximization (OSEM) iterative reconstruction technique, which is an accelerated version of EM, with Poisson variability. Regularization is achieved by penalizing OSEM with probabilistic patch-based regularization (PPB) filter to form hybrid method (OSEM+PPB) for CT/PET image reconstruction that uses neighborhood patches instead of individual pixels in computing the non-quadratic penalty. The aim of this paper is to impose an effective edge preserving and noise removing framework to optimize the quality of CT/PET reconstructed images. A comparative analysis of the proposed model with some other existing standard methods in literature is presented both qualitatively and quantitatively using simulated test phantom and standard digital image. An experimental result indicates that the proposed method yields significantly improvements in quality of reconstructed images from the projection data. The obtained results justify the applicability of the proposed method.