k-TE GRAPPA for Rapid Abdominal R2* Mapping

Introduction R2* measurements have proven useful for a number of abdominal imaging applications [1, 2]. The commonly used multiple-GRE (MGRE) sequence collects a train of echoes after each excitation thereby permitting the reconstruction of separate images at each TE and subsequent derivation of R2* maps. However, spatial resolution and overall volumetric coverage often remains limited due to breath-hold requirements. Based upon an extension of previously proposed k-t GRAPPA parallel imaging methods [3] we recently developed a k-TE GRAPPA approach to reduce MGRE scan times. The purpose of our study was to compare k-TE GRAPPA to conventional GRAPPA and viewsharing approaches for MGRE R2*-mapping in phantoms and healthy volunteers. Methods k-TE GRAPPA k-t GRAPPA exploits correlations in k-space and time to reduce residual parallel imaging artifacts [3]. Reconstruction is similar to GRAPPA but involves additional contributions from earlier and later sampled k-space lines to the coefficient fitting parameters. For MGRE we sought to exploit correlations between k-space and echo-time dimensions therefore dubbing the technique ‘k-TE’ GRAPPA. Data Acquisition Imaging was performed using a 1.5T clinical MRI scanner (Siemens Magnetom Sonata) with 6 channel anterior and posterior array coils. MGRE sequence was modified to sub-sample k-space along the ky dimension while alternating sub-sampling positions along the TE dimension (i.e. for acceleration rate = 2, odd ky sampled for odd echo numbers and even ky for even echo numbers). The latter alternation was accomplished with the application of phase-encoding gradients between adjacent read-outs. MGRE parameters: TR/TE=60/3-40ms, 30 flipangle, 630Hz/px BW, 5mm slices, ETL=16. We performed additional acquisitions with spatial-spectral water excitations (WE) to investigate the impact of inand out-of-phase fat signals upon view-sharing and k-TE methods. Acquisitions were repeated at k-TE GRAPPA acceleration rates 1, 2, 3, 4, and 5. We separately reconstructed images at each TE using view-sharing, GRAPPA, and k-TE GRAPPA approaches. Voxelwise linear fitting to produced R2* maps from each acquisition. Phantom studies: MnCl2 –doped saline bottles were suspended in a peanut oil bath providing phantoms with wide ranging R2* and exterior ring representative of fatty tissues. In vivo studies: Axial and coronal abdominal MGRE scans were performed in 7 healthy volunteers according to IRB-approved study protocol. Data Analysis For both phantom and volunteer studies, we calculated the artifact power (AP) at each TE for view-sharing (VS), GRAPPA, and k-TE GRAPPA approaches by comparison to fully sampled MGRE datasets. We measured mean R2* measurement error at each rate (medulla and liver ROIs). AP and R2* measurement error from the in vivo studies were compared using ANOVA with Tukey post-hoc correction (α=0.05). Results Phantom Initial comparisons revealed that VS was significantly more sensitive to fat in/out-of-phase effects (>AP at each TE) than kTE and WE yielded reduced AP for both VS and k-TE approaches (Fig. 1A). As expected for each of the three approaches AP increased with increasing R2* and TE due to more rapid signal decay (Fig. 1B). At each of the acceleration rates k-TE R2* measurements were more accurate (lower absolute error) than VS and conventional GRAPPA approaches (Fig. 1C). In vivo Similar to our previous phantom studies, we found that AP was significantly reduced using water-excitation as opposed to conventional RF excitation methods. At rate=2 VS, GRAPPA, and k-TE GRAPPA offered similar performance (no significant difference in AP) while at all higher rates (i.e. rates>2) k-TE GRAPPA reduced AP and improved overall image quality compared to both alternative approaches (p<0.05). In vivo R2* maps for a representative volunteer study are shown in Fig.2. At acceleration rate=2 there was generally no discernable differences between the three approaches. However, when proceeding to higher rates, we note increased artifact levels for conventional GRAPPA R2*maps and significant spatial blurring within VS maps compared to k-TE GRAPPA. Conclusion k-TE GRAPPA is an effective parallel imaging method to accelerate MGRE R2*-mapping. k-TE GRAPPA reduced artifact levels and provided more accurate R2* measurements compared to VS and GRAPPA approaches. k-TE GRAPPA offers the potential to reduce breath-hold times and/or increase spatial resolution or coverage. Future studies will focus on 3D volumetric and clinical applications.