Prospective motion correction for intra-cardiac 3D delayed enhancement MRI using an MR-Tracking Tetrahedron

Introduction Myocardial Delayed Enhancement (MDE) MR imaging has been used for visualization of myocardial scar tissue created due to radio-frequency ablation (RFA) of the left atrium for the treatment of atrial fibrillation [1,2]. After administration of gadolinium-DTPA contrast, ECG-gated 3D-MDE is performed with an inversion-recovery preparatory pulse, followed by a segmented k-space gradient-echo (GRE) acquisition, which requires >10 minutes to acquire an image of sufficient resolution to discriminate ablation gaps. Prior work showed an intra-cardiac MR coil offers 8-10 times higher signal-to-noise ratio (SNR) than surface cardiac-arrays, allowing acquisition of high-resolution images in shorter scan times [3]. However, since intra-cardiac coils move with the anatomy, motion artifacts are more severe. Integrating MR tracking coils and intra-cavitary coils, combined with prospective motion correction, has demonstrated improved image quality in 2D-GRE [4]. In this study, we propose to extend prospective motion correction techniques using MR tracking coils to 3D-MDE imaging. Methods Four tracking coils were mounted on an expandable intra-cardiac imaging catheter, which when expanded, assumed a tetrahedral shape [4]. Each of the 4 coils was connected to an individual receiver channel, allowing simultaneous motion detection. Readouts in three orthogonal axes detected 3-D translational and rotational motion. Phase dithering [5] was used to eliminate background signal coupled from the adjacent imaging coil. A 3D-MDE sequence was modified, adding 2 tracking segments in each RR, one before and one after the imaging segment (Fig.1). At the start of the scan, the position of the tracking coils was set to be the baseline. During imaging, the position detected by the first tracking segment in each RR was compared with the baseline. When the detected motion was bigger than a preset threshold (1 mm), the acquired imaging data was rejected and the 2 tracking segment was omitted. This process assured the same volume was imaged during the entire scan. If data was rejected too many (>16) times continuously, a dynamic baseline update was implemented using the mean of the prior 16 positions, which accommodated gradual baseline changes. The motion detected from the 2 tracking segment was compared with the 1 segment. If the difference was bigger than a threshold, the previously acquired data was also rejected, since it was assumed to be contaminated with motion which occurred during acquisition. Experiments were performed on a GE (Milwaukee, WI) 1.5T. An ex-vivo swine heart was imaged using the modified sequence with the table cyclically displaced by 10mm in the Superior/Inferior direction, interspersed by motionless periods. A 5” imaging coil was used with tetrahedron tracking array.