EVALUATION OF ITERATIVE ALGORITHMS FOR TOMOGRAPHY IMAGE RECONSTRUCTION

Alexandre F Velo,Alexandre G Alvarez,Carlos Henrique De Mesquita,Margarida Mizue Hamada

doi:10.15392/bjrs.v7i2a.660

Abstract

The greatest impact of the tomography technology currently occurs in medicine. The success is due to the fact that human body presents standardized dimensions with well-established composition. These conditions are not found in industrial objects. In industry, there is a great deal of interest in using the tomography in order to know the inner part of (i) manufactured industrial objects or (ii) the machines and their means of production. In these cases, the purpose of the tomography is: (a) to control the quality of the final product and (b) to optimize the production, contributing to the pilot phase of the projects and analyzing the quality of the means of production. This scan system is a non-destructive, efficient and fast method for providing sectional images of industrial objects and it is able to show the dynamic processes and the dispersion of the materials structures within these objects. In this context, it is important that the reconstructed image may present a great spatial resolution with a satisfactory temporal resolution. Thus, the algorithm to reconstruct the images has to meet these requirements. This work consists in the analysis of three different iterative algorithm methods, namely the Maximum Likelihood Estimation Method (MLEM), the Maximum Likelihood Transmitted Method (MLTR) and the Simultaneous Iterative Reconstruction Method (SIRT. The analyses involved the measurement of the contrast to noise ratio (CNR), the root mean square error (RMSE) and the Modulation Transfer Function (MTF),in order to know which algorithm fits the conditions to optimize the system better. The algorithms and the image quality analyses were performed by Matlab® 2013b.

Highlights

Unlike the standard aspect of the computed tomography (CT) for medical application, industrial tomography systems applications should be adapted to the different size and geometry objects, usually placed in an aggressive environment, which contains flammable superheated or corrosive materials, and, eventually, subject to high internal pressure, all these factors bring in many difficulties for setting CT devices [1, 2]
The reconstructed images of the phantom obtained by the third generation industrial tomography are presented in Fig. 7, where Fig. 7a. represents the reconstruction by Simultaneous Iterative Reconstruction Technique (SIRT) algorithm; Fig. 7b., by Maximum Likelihood Algorithm for Transmission Tomography (MLTR) algorithm; and Fig. 7c., by Maximum Likelihood Expectation Maximization (MLEM) algorithm
It is possible to observe that even MLEM and MLTR algorithms reaching a higher Contrast to Noise Ratio (CNR) value in the first iterations, these values decrease as the number of iterations rises, different from SIRT algorithm, which reaches a lower CNR value, but decreases less than the others, what means that the noise influences less in the images reconstructed by the SIRT algorithm, as the number of iteration increases, when compared to the other two algorithms

Summary

Introduction

Unlike the standard aspect of the computed tomography (CT) for medical application, industrial tomography systems applications should be adapted to the different size and geometry objects, usually placed in an aggressive environment, which contains flammable superheated or corrosive materials, and, eventually, subject to high internal pressure, all these factors bring in many difficulties for setting CT devices [1, 2]. The CT systems based on transmission uses an array of encapsulated radioactive sources and detectors placed in opposite sides of the targeted object [7-9]. In the second generation CT systems, a set of detectors is placed opposite to the radioactive source with fan beam, moving (source and detector) around the object under study (Fig.1b) [10]. The system rotates around the targeted object, obtaining a particular view for an "x" position of the source-detector array. In this type of system, several sources and arrays of multiple detectors may be used [10]. The so-called fourth-generation CT systems use a fixed array of detectors (a large number of detectors mounted on a fixed ring) and a radioactive source that rotates around the object (Fig.1d). All CTs are constituted, basically, of the same parts: radioactive sources; radiation detectors; a data acquisition system and a suitable computer [10]

Methods

Results

Conclusion