Continuously Moving Table Axial Imaging with Radial Acquisitions, JP Ajit Shankaranarayanan, B Her&ens, J Brittain

Tags: SNR, steady state, image contrast, projection, prime importance, spatial resolution, 3Stanford University, Jean Brittain, Ajit Shankaranarayanan, GE Medical Systems, Menlo Park, CA, Jason Polzin
Content: Continuously Moving Table Axial Imaging with Radial Acquisitions
Ajit Shankaranarayanan', Jason Polzin', Bob Her&ens3, Jean Brittain'. GE Medical Systems, `ASL West, Menlo Park, CA, 2Milwaukee, Wisconsin, 3Stanford University, Dept. of Radiology, Stanford, C A
Abstract A technique for whole-body axial magnetic resonance (MR) imaging using a continuously moving table and a projection reconstruction acquisition is presented. An undersampled projection reconstruction acquisition enables good spatial resolution and short scan times. The reconstructed image data set can be reformatted in any desired plane. Results from volunteer experiments illustrating the feasibility of the technique are presented. Introduction Previous work has demonstrated the feasibility of whole body axial imaging with a continuously moving table [ I , 21. Radial sampling, compared to the 2DFT approaches offers a different set of tradeoffs; undersampling gives rise to primarily reduced signal-to-noise ratio (SNR) and, to a lesser extent, decreased resolution. The feasibility of performing whole body imaging with the moving table technique using radial sampling, similar to spiral computed tomography (CT), is assessed here. Methods Figure 1 shows the concept of whole body axial imaging with radial sampling. The projections are acquired as the table is moved continuously. Neglecting the movement of the table during a single readout, the projections are equally spaced with the center of rotation being the axis of a cylinder. Acauisition: The pulse sequences using radial trajectories were implemented on a 1.5T GE Signa Lx Twinspeed (GE Medical Systems, Milwaukee, Wisconsin) with a high performance gradient system (maximum gradient strengths of 4OmTim and maximum slew rate of 150mT/m/msec), with the "zoom" mode chosen. The pulse sequence used here was a 2D Fast Imaging with Steady State Acquisition (FIESTA) (a.k.a fully balanced SSFP) technique. The FIESTA acquisition, with its high signal at short TR, can yield higher resolution in the table motion direction while giving good contrast in the axial images. A regular 2DFT FIESTA sequence was modified for radial acquisition. The table motion software was incorporated into the sequence to acquire the projections as shown in Figure 1. The acquisition parameters were TWTE = 3.32ms/1.6ms. The other parameters for the acquisition were a=50", FOV=400mm x 400mm, xres = 256, no. of projections = 192, reconstruction matrix size = 256x256 and Nex = 1. Table Motion: The table motion software allowed the operator to choose table velocities between 0.5 to l0cmisec. The table velocities were chosen to ensure complete sampling before the table moved a slice thickness. Volunteer exueriments: Following informed consent, healthy volunteers were placed in the scanner and radial acquisitions were performed while the table moved continuously. Post Processing: Reconstruction of the images and implementation of the post processing was done offline, in MATLAB (Natick, MA). The acquired data was first spline interpolated in the direction of table motion. Each plane was then gridded on to a 2x Cartesian grid and reconstructed. The image data set was then reformatted in other planes. Results and Discussions Figure 2 shows axial images of the brain(b), abdomen(c) and thigh(d) and a reformatted image in the sagittal direction(a) of a normal volunteer. A table speed of 2 c d s was used here to acquire 100 revolutions ( lrevolution = 192 projections) in little more than lminute. The raw data was then spline interpolated in the table motion direction to generate 200 raw data sets. Note that small details can be discerned in the images shown in Figure 2. One of the advantages of this technique is that the center of the slice can be chosen arbitrarily or, similarly, any point in the heart cycle can be chosen. In radial sampling techniques, the image resolution is primarily dependent on the number of samples in the read direction. This results in a higher resolution per unit time than Fourier imaging, at the expense of SNR. This can be advantageous
in fast screen for metastases where scan time along with spatial resolution and image contrast are of prime importance. The decreased SNR can be compensated to some extent by using a FIESTA acquisition which inherently has higher SNR. The number of projections can thus be optimized for SNR and minimization of the streak artifacts. 192 projections were used here to avoid the streaking artifacts.. A 1Omm slice thickness allowed most of the projections to be acquired in steady state. The ffaction of spins not in steady state for each projection can be determined by the product of the table velocity and the TR of the acquisition. Since the fraction of spins not in steady state is very low in our protocol, the artifacts are not evident. Preparation pulses were also used for achieving steady state for the first data set. As mentioned earlier the movement of the table during a single readout was ignored during reconstruction. The resultant blurring in the images was found to be very small. However there is opportunity to exploit spiral CT algorithms to improve the Image quality. Conclusion In summary, this work proposes an alternate technique to perform whole body imaging in a short scan time and shows its effectiveness in obtaining good in plane spatial resolution by exploiting the advantages of undersampled projection reconstruction. However the table velocities and number of projections can also be optimized for improving the resolution in the table motion direction. It also provides opportunities, in the future, for more involved processing (e.g. half k-space acquisition) and exploiting reconstruction algorithms developed for spiral CT. Figure 1: Whole body axial MR imaging. This figure shows the acquisiton of the projections as the table moves. (a) shows the 3d represntation of the trajectory and (b) shows the projection of the rrajector). from top. Figure 2: The images from a 2D FIESTA sequence a) reformatted image in the sagittal direction ,b)head, c) abdomen, and c) upper thigh. The total imaging time was 65sec for covering 130 cm. Acknowledeement I would like to acknowledge my former colleagues Dr Jeff Duerk and Dr Claudia Hillenbrand for useful discussions on helical scanning methods involving changing the excitation frequency. References [1] Johnson KMR, Leavitt GD, Kayser HWM. Radiology 1991;202:262-267. [2] Barkhausen J, Quick HH, Lauenstein Y et al. Radiology 2001 ;220:252-256.
© Proc. Intl. Soc. Mag. Reson. Med. 10 (2002)

JP Ajit Shankaranarayanan, B Her&ens, J Brittain

File: continuously-moving-table-axial-imaging-with-radial-acquisitions.pdf
Title: Moving Table Axial Imaging with Radial Acquisitions
Author: JP Ajit Shankaranarayanan, B Her&ens, J Brittain
Author: Ajit Shankaranarayanan, Jason Polzin, Bob Herfkens, Jean Brittain
Subject: 712 Other MRI Sequences
Keywords: Sequences: Fast Imaging, Abstract not focused on specific pathology, Abstract not focused on specific organ/tissue, Feasibility Studies of MR Methods, Data Processing: MRI
Published: Wed Apr 24 12:21:04 2002
Pages: 1
File size: 0.24 Mb

, pages, 0 Mb

Modern control systems, 7 pages, 0.43 Mb

Volume 1 Issue, 12 pages, 0.91 Mb

, pages, 0 Mb
Copyright © 2018