Although intervertebral disc degeneration is a part of natural aging process, there are factors that influence the rate and nature of degeneration. One of the proposed mechanisms is disruption of nutrient delivery through the disc endplates, which could lead to disc degeneration¹. We are currently exploring a novel approach to study such endplate changes in vivo using Dynamic Contrast Enhanced MRI (DCE-MRI). Although our initial results demonstrated profound changes in the endplates of degenerating discs, accurate quantification of such changes was hindered by low temporal resolution. Despite using the fastest protocol available from the manufacturer, the temporal resolution was sacrificed (30s) to image the thin endplates with high spatial resolution. With higher temporal resolutions, it would be feasible to quantify changes in the vascular and extravascular space using pharmacokinetic models. This could help us study the pathological changes in endplates noninvasively and explore associations between endplate changes and low back pain. Another problem with the existing protocol was motion artifacts because it is based on 3D SPGR, which is sensitive to motion. Here, we implemented a novel acquisition technique to obtain DCE-MRI data at high spatial and temporal resolution, which is also more tolerant to motion.Radial imaging has several advantages over Cartesian sampling. Radial scans are less prone to subject motion, allows higher acceleration rates with less artifacts and it covers the center of k-space with each spoke. In conventional radial sampling, k-space data is first interpolated onto a rectangular grid followed by Fourier Transform. However, if data is acquired on a concentric-squares grid (aka Linogram), there exists a direct, exact and fast transformation². Fig.1 shows the k-space sampling trajectories for the two techniques. Although Linogram method has been implemented in MRI before³, acceleration with parallel imaging was not studied to this date. Furthermore, its performance for high spatial and temporal resolution DCE-MRI was not explored. Here we developed a 3D-Linogram acquisition with GRAPPA acceleration and acquired DCE-MRI data from the lumbar spine. The 3D-Linogram was implemented using stack-of-stars sampling technique. Only in-plane acceleration was used and images were reconstructed using a modified radial-GRAPPA technique⁴. In this approach, k-space is divided into smaller segments, and the weight matrix for each segment is calibrated using the same segment from a fully sampled calibration data acquired separately. More than one kernel is required for radial or linogram sampling schemes because undersampling direction and spacing are different in different regions of the k-space. A 2×5 kernel size was used in this study. Data were acquired on a 3T GE-MR750 MRI scanner.Images were acquired from the lumbar spine of a volunteer using the 3D-Linogram. The study was approved by the IRB and written consent was obtained. We first collected single-volume images with and without acceleration. Fig.2 shows 3D-Linogram images acquired with full k-space scan and also with 2-fold acceleration using two channels of the CTL coil. For reference, a comparable full k-space acquisition with conventional 3D Cartesian scan was also shown. Linogram data were acquired with FOV=240mm (actual FOV=480mm with 512 samples to avoid fold-over artifacts, then cropped to 240mm), TR/TE=5ms/2.2ms, 12 slices with 6mm thickness. 384 radial spokes for full scans and 128 radial spokes for 2-fold acceleration were acquired to reconstruct 256×256 images (acceleration rate is given with respect to the Nyquist rate of Cartesian sampling). Total acquisition time of 3D-Linogram with 2-fold acceleration was 7.7s. DCE-MRI was acquired with 2-fold acceleration and 130 frames were collected. Gd-DTPA was administered as a bolus after the 10th dynamic frame. Enhancement plots from vertebral endplates are shown in Fig.3. Tofts’ pharmacokinetic model parameters were estimated from these curves and plotted (a group-averaged arterial input function was used). Typical Ktrans values were between 0.067/min and 0.107/min.We demonstrated the feasibility of a new acquisition technique for high spatial and temporal resolution DCE-MRI to investigate disc endplate changes. In this study we had to use only two channels of a CTL coil, which limited the rate of acceleration. One could obtain higher acceleration rates, hence better temporal resolution, using RF coil arrays with more elements. Note that 2-fold acceleration with the product 3D Cartesian would still take 15s and the image quality was inferior. In addition to poorer parallel imaging performance, Cartesian scans in the spine with sagittal orientation takes longer because one has to choose A-P direction for frequency encoding to reduce motion artifacts, which requires phase oversampling to minimize fold-over artifacts. The 3D-Linogram technique can also be used in other dynamic imaging studies where high spatial and temporal resolution is required.