Item

Abstract

Summary Through-Time 3D Radial GRAPPA can be used to reconstruct 3D CINE images covering the whole heart in a single breathhold. Background Through-Time Non-Cartesian GRAPPA has been previously demonstrated for real-time 2D cardiac imaging (Seiberlich, et al. Magn Reson Med. 2011 Feb;65(2):492-505.). This parallel imaging method works by acquiring several fully-sampled non-Cartesian datasets with a low temporal resolution, and using the coil sensitivity information from these datasets to reconstruct highly undersampled non-Cartesian data acquired in real-time. By modifying this through-time non-Cartesian GRAPPA method to reconstruct highly undersampled 3D data, whole heart 3D CINE images can be generated using data acquired in a single breathhold. Methods A total of 20 fully-sampled 3D stack-of-stars radial datasets were acquired during free-breathing with no EKG gating using a 1.5T Siemens Espree and the following parameters: bSSFP sequence, TE=1.52ms, TR=3.04ms, matrix size = 128x128x20, projections/partition=128, FOV=300x300x90mm3, Flip Angle=45°, 5/8 Partial Fourier, 18 receiver channels. Segmented undersampled data (using only 16 projections/partition, an acceleration factor of R=8) were acquired with EKG gating and the above parameters during a breathhold for 15 heartbeats, resulting in 15 CINE frames. In order to perform the calibration, each of the time frames and partitions were employed as separate sources of calibration information; thus, a total of 300 repetitions could be used to generate the through-time GRAPPA weight sets. After reconstruction, the undersampled data yielded fully-sampled 3D CINE images, each with a temporal footprint of 48ms, an in-plane resolution of 2.3mm2, and a through-plane resolution of 6mm. The total acquisition time was 116s for the calibration and approximately 15 s for the breathhold CINE acquisition. Results Example images from diastole and systole of one healthy volunteer are shown in Figures 1 and 2. It is important to note that these represent just two of the 15 CINE frames acquired in this dataset. Despite the high acceleration factor (R=8 in comparison to the fully-sampled calibration data), the images demonstrate only minor residual aliasing artifacts. Because a 3D dataset is acquired, the images from each partition can be shown in the same cardiac phase, which is challenging when using multiple breathholds to acquire several 2D CINE slices.