Efficient Off-Resonance Correction for Spiral Imaging Krishna S. Nayak, * Chi-Ming Tsai, Craig H. Meyer, and Dwight G. Nishimura A new spiral imaging technique incorporates the acquisition of a field map into imaging interleaves. Variable density spiral trajectories are designed to oversample the central region of k-space, and interleaves are acquired at two different echo times. A field map is extracted from this data and multifre- quency reconstruction is used to form an off-resonance cor- rected image using the entire dataset. Simulation, phantom, and in vivo results indicate that this technique can be used to achieve higher image and/or field map spatial resolution com- pared to conventional techniques. Magn Reson Med 45: 521–524, 2001. © 2001 Wiley-Liss, Inc. Key words: spiral MRI; off-resonance correction; k-space; re- construction Spiral imaging is used for a variety of applications due to its efficient k-space coverage (1) and excellent flow prop- erties (2). In spiral imaging, however, off-resonance can result in image domain blurring (3). Conventional ap- proaches to spiral off-resonance correction involve acquir- ing a field map using extra acquisitions. Typically, two low-resolution single-shot spiral images, taken at different echo times, are used to compute a field map. This map may then be used for linear field map correction or a frequency- sensitive reconstruction (4,5). Disadvantages of this ap- proach are that separate field map acquisitions are re- quired, magnitude information from the field map images are unused, and the field map resolution is not very flex- ible. We present a new technique, termed ORC-VDS (off- resonance correction using variable density spirals) that combines the image and field map acquisitions using a variable density spiral (VDS) trajectory that oversamples the k-space origin. This technique has greater scan time efficiency relative to conventional techniques and enables the flexible tradeoff of field map resolution for image res- olution. A similar technique has been previously applied to pro- jection reconstruction imaging (6) where the central k- space is inherently oversampled and off-resonance blur- ring occurs. In spiral imaging, oversampling the k-space origin requires trajectory modification, but the longer read- outs make off-resonance blurring a more significant obsta- cle that requires accurate correction. METHODS In the proposed technique, variable density spiral (VDS) trajectories (see Fig. 1) are designed such that a small circular region around the k-space origin is oversampled by a factor of 2. Every alternate spiral acquisition is then delayed by a small amount labeled TE. The 2 over- sampled region is used to generate two lower-resolution images with different echo times— one from the early-TE interleaves and one from the late-TE interleaves. Notice that the low-resolution images have the same field of view (FOV) as an image from the full dataset. A field map is then computed in the standard way, using the phase difference between the two low-resolution im- ages. Since these field map estimates are only reliable in areas of sufficient signal, a smooth polynomial approxima- tion (7) over the entire FOV can be computed for use during reconstruction. A multifrequency image reconstruction is then used to simultaneously compensate for off-resonance and correct for the difference in echo times (4 – 6). In multifrequency reconstruction, a small finite set of frequency samples { f n } are selected such that they span the full range of off-resonant frequencies. For each f n , an image is recon- structed based on that frequency of precession. This is done by modulating the raw data of each readout by e +2i f n t and phase-aligning the early and late echo data by multiplying the raw data from late-TE acquisitions by e +2i f n TE . An image is then generated via gridding re- construction (8,9) with the modulated and aligned data from both early-TE and late-TE acquisitions. For the final image, each voxel is estimated by interpolating between the images based on the closest { f n } frequency samples to f ( x, y) (4) or by taking a linear combination of all the { f n } images (5). Resulting images are formed without separate field map acquisitions and with only a moderate increase in recon- struction time. This combination of image and field map acquisitions provides for greater scan efficiency, while the ability to specify the oversampled region provides flexibil- ity in trading off field map resolution for image resolution. RESULTS Performance comparisons and results are presented based on our scanner, a GE Signa 1.5T CV/i scanner (General Electric, Milwaukee, WI). This scanner is equipped with gradients capable of 40 mT/m magnitude and 150 T/m/sec slew rate and a receiver capable of 4 sec sampling (125 kHz). For phantom studies a head coil was used and for in vivo studies a body coil was used for RF transmission and 5-inch surface coil used for signal reception. Unless spec- ified, all ORC-VDS images use a TE of 1 msec. Performance Evaluation A first comparison of ORC-VDS against the conventional (separate field map) technique is illustrated in Fig. 2. Op- Department of Electrical Engineering, Stanford University, Stanford, Califor- nia. A preliminary account of this work was presented at the 8th Annual Scientific Meeting of the ISMRM, Denver, 2000 (abstract 116). Grant sponsors: National Institutes of Health; GE Medical Systems. *Correspondence to: Krishna S. Nayak, Packard 211, ISL, 350 Serra Mall, Stanford University, Stanford, CA 94305-9510. E-mail: nayak@lad.stanford.edu Received 16 May 2000; revised 31 August 2000; accepted 14 September 2000. Magnetic Resonance in Medicine 45:521–524 (2001) © 2001 Wiley-Liss, Inc. 521