Euromedim 2006 :1st European Conference on Molecular Imaging Technology

MR Image Reconstruction from Partial k-Space Using Singularity Function Representation

9 May 2006, 14:00

1h

Palais du Pharo, Marseille

poster • Image reconstruction and processing Poster Session :Simulation, Modeling, Reconstruction

Speaker

Dr Yuemin ZHU (CREATIS, CNRS UMR5515 & INSERM U630, Lyon)

Description

Partial k-space acquisition is usually employed in magnetic resonance imaging (MRI) for reducing imaging time while maintaining image quality. In this field, image reconstruction from the incomplete k-space is an important matter. A number of reconstruction methods have been reported in the literature, but all these methods do not cope with truncation artifacts. We present a method that is particularly suitable for reconstructing magnetic resonance (MR) images from partial k-space by reducing substantially truncation artifacts. The proposed method is based on the use of a so-called singularity function representation. It consists in first representing the image to be reconstructed by the linear combination of singularity functions (step functions), and then estimating the parameters of the representation model through using information derived from the truncated k-space. In contrast to MR reconstruction using, directly or indirectly, inverse Fourier transform, this method does not reconstruct images directly from k-space. The key point of this method using singularity function-based reconstruction is the appropriate determination of singular points and singularity degrees. Therefore, we present a new strategy for estimating these two parameters by imposing constraints. It consists of first extracting singular points using a layer extraction technique, and second using the obtained singular points in the equation system for obtaining singularity degrees. We restrict the singularity degrees within some dynamic range, beyond which the corresponding singular points are set to zero. This new approach allows to reduce or suppress false singular points, and thus improves the reconstruction quality. The proposed MR reconstruction method was evaluated on both simulated and acquired MR data from human brain. For simulated data, a slowly varying phase was introduced to generate a complex image whose k-space does not exhibit Hermitian symmetry. The acquired MR data saved as raw k-space data were obtained using a 1.5 T Siemens Sonata system. The reconstruction quality was assessed visually and using the quantitative criteria such as normalized mean-square error (NMSE). The obtained results showed that the method is particularly efficient for overcoming reconstruction limitations due to truncation artifacts in partial k-space acquisitions. It exhibits a substantially improved performance in comparison with the classical and popular zero-filling technique and more state of the art methods such as the Margosian/Homodyne method. In conclusion, the proposed method allows to obtain images of good quality with a significant reduction of scanning time by maximizing the asymmetry of k-space.