3D image reconstruction of CT-data using the ITEM algorithm

Abstract

ITEM (Imaginary Time Expectation Maximization) is a general minimization technique based on quantum-mechanical (QM)-methods. It has already proven to be a useful method for image reconstruction . ITEM is a fast (compared to other statistical methods) and intrinsically 3D-friendly (for its low demands on memory) algorithm. Therefore it is a good candidate for a reconstruction method for computed-tomography (CT)-data. Being a statistical method, it is possible with ITEM to model the underlying physics of the imaging system better than with an analytical backprojection method like filtered backprojection (FB) and so that better images with the same data set can be obtained. Two implementations of ITEM for CT-data and the resulting images achieved with both methods are presented. Finally the results are compared with standard (FB) techniques. ITEM (algorithm as well as all of its implementations) is published under the terms of GNU/GPL.