COMPARISON OF MAGNETIC RESONANCE ELECTRICAL IMPEDANCE TOMOGRAPHY (MREIT) RECONSTRUCTION ALGORITHMS B. Murat Eyübo÷lu, V. Emre Arpﾕnar, Rasim Boyacﾕo÷lu, Evren De÷irmenci, Gökhan Eker Department of Electrical and Electronics Engineering, Faculty of Engineering, Middle East Technical University, 06531 Ankara, Turkey ABSTRACT Several algorithms have been proposed for image reconstruction in MREIT. These algorithms reconstruct conductivity distribution either directly from magnetic flux density measurements or from reconstructed current density distribution. In this study, performance of all major algorithms are evaluated and compared on a common platform, in terms of their reconstruction error, reconstruction time, perceptual image quality, immunity against measurement noise, required electrode size. J- Substitution (JS) and Hybrid J-Substitution algorithms have the best reconstruction accuracy but they are among the slowest. Another current density based algorithm, Equipotential Projection (EPP) algorithm along with magnetic flux density based B z Sensitivity (BzS) algorithm has moderate reconstruction accuracy. BzS algorithm is the fastest. Index Terms— Magnetic Resonance, Imaging, Tomography, Electrical Impedance, Conductivity 1. INTRODUCTION High resolution images of electrical conductivity distribution can be obtained by Magnetic Resonance Electrical Impedance Tomography (MREIT) [1]. In MREIT, magnetic flux density created by an applied current to a volume conductor, with magnetic resonance (MR) active nuclei, acts like a local gradient in MR imaging field. An additional phase, which is proportional to strength of the current induced magnetic flux density in the direction of main MR imaging field, duration of the applied current and gyromagnetic constant, accumulates on the acquired Free Induction Decay (FID) signal. Therefore, distribution of magnetic flux density in the MR main field direction can be determined from phase of the acquired FID signal. Relationship between current density and magnetic flux density in a volume conductor is described by Biot-Savart law. After obtaining magnetic flux density from MREIT phase image, current density and conductivity can be reconstructed. Many reconstruction algorithms have been proposed for MREIT reconstruction. Some of these algorithms (J-based) calculates current density utilizing all components of magnetic flux density by means of Biot- Savart law, then determines conductivity distribution. This requires measurement of three orthogonal components of magnetic flux density [1], which is not practical. Another group of algorithms (B-based) can reconstruct conductivity distribution directly utilizing only one component of the magnetic flux density data in the direction orthogonal to the reconstruction plane. In this study, reconstruction error, reconstruction time, perceptual image quality, immunity against measurement noise, required electrode size of the algorithms are compared. Due to space limitation in this manuscript, the readers are referred to the original manuscripts for underlying mathematical description of the algorithms. These algorithms are J-Substitution (JS) [2], Hybrid J- Substitution (HJS) [3], Equipotential Projection (EPP) [4], Integration Along Cartesian Grid Lines (IACGL) [5], Integration Along Equipotential Lines (IAEPL) [5], Solution as a Linear Equation System (SLES) [5], Harmonic B z (HBz) [6], Variational Gradient B z (VGBz) [7], B z Sensitivity (BzS) [8] and Algebraic Reconstruction (AR) [9] algorithms. The first six algorithms are J-based and the last four algorithms are B-based algorithms. 2. METHODS Performances of all algorithms implemented in this study are evaluated using simulated MREIT measurements for a thorax phantom. Simplified geometry of the phantom and the conductivity values used for different tissues in the phantom are given in Figure 1 and Table 1, respectively. Current injection electrode pairs are located at the opposite sides of the phantom. The size of each electrode is selected as equal to 1/5 of the phantom side length for all algorithms except for SLES and IACGL algorithms. These two algorithms work only when the electrode size is large. Therefore, for these two algorithms, electrodes covering the entire sides of the phantom are used. A finite element (FE) model is constructed based on this geometry and conductivity values. Then, magnetic flux densities for given injection currents are calculated. Noise model proposed by Scott et al [10] is adopted to simulate measurement noise. 700 978-1-4244-4126-6/10/$25.00 ©2010 IEEE ISBI 2010