Novel Approaches to Low-Cost MRI zyx Alhert Macovski, Steven Conolly zyxwvu This paper presentsa combination of speculativeapproaches, some related to earlier work and some apparently novel, which show great promise in providing a new class of MRI machines that would be considerably less expensive. This class would have advantages and disadvantages as com- pa& to existing MRI, over and above that of low cost. The disadvantages include the apparent inability to perform clas- sic slpectroscopy, and limited flexibility in the area of selective excitation. The advantages include a fundamental immunity to inhoinogeneity and susceptibility problems, the ability to cre- ate zyxwvutsrqp 81 wide class of machines that are designed for specific anatomy-related applications, the ability to design open ma- chinos for physician access, and Improved capability for high speed Imaging. Generlc to all of the methods presented are a pulstd polarizing field and an oscillatory readout bias field. The pulsed field initlally polarizes the magnetic moments. Sinat it is not zyxwvuts on during the readout operation it has negligible homogeneity requirements since changes in the field ampli- tude will merely shade the image Intensity. During readout a relatively low bias field is used. To enable the use of a reia- tively inhomogeneous bias field, an osciliatory field is used that lhas a zero average value. This prevents any long-term buildup of phase errors due to a frequency error associated with inhomogeneity. Thus the average bias frequency will be determined solely by the frequency rather than the amplitude of the bias field. Three methods are described, all including the above features. The first two Involve imaging in the labo- ratory frame, while the third involves Imaging in the rotating framo. The second approach requires no RF excitation and the third approach uses RF bias and gradient signals. Some ap- proaches to slice selection are described. Key words: magnetic resonance imaging; prepoiarization pulsed magnet; oscillating bias field. INTRODUCTION The dramatic success of clinical MRI presents an appar- ent paradox. On the one hand, the exquisite image qual- ity and lack of any toxicity or ionizing radiation would suggest that MRI is an ideal candidate for general screen- ing at an asymptomatic level. However, the present costs make this prohibitive. It is hoped that, using techniques of the type outlined in this paper, a new class of ma- chines will be built providing high performance at a small fraction of present costs. The primary basis for the reduced cost is a system that can tolerate a very high degree of inhomogeneity. This feature, of itself, leads to a MRM 305?21-230 (1993) From tihe Department of Electrical Engineering, Stanford University, Durand 120, Stanford, California. Address correspondence to: Albert Macovski, Ph.D., Information Systems Laboratory, Durand Building. Room 109, Stanford, CA 94305-4055. Received January 11, 1993; revised March 25,1993; accepted March 25, 1993. This wgork was suppo zyxwvutsrqp ed by a Whitaker Foundation Biomedical Research Copyright Q 1993 by blliarns zyxwvutsrqp 8 Wilkins All rights of reproduction in any form reserved. Grant to SC. f 0740-31W93 $3.00 wide variety of cost savings and performance advantages. It enables smaller magnets to be used with selected re- gions of the anatomy. These have significantly reduced stored energy. More important, they can be built with very wide tolerances, drastically reducing their cost. We must caution, however, that at this juncture the full re- alization remains speculative. For conventional imaging, susceptibility variations within the body cause distortion and signal losses. The accompanying zyxwv T: reduces the available readout time, thus lowering the SNR. In some regions, such as the lungs and bones, susceptibility makes imaging almost impos- sible. Hence, the proposed system’s insensitivity to inho- rnogeneity represents major performance advantages. This insensitivity should enable the imaging of regions with various metallic clips and pins. These desirable qualities are achieved primarily through the use of two features: a pulsed “prepolarizing” field, BJt), and an oscillating bias field, Bo(t). These techniques, called prepolarized magnetic resonance im- aging (PMRI), will be described using two generic read- out systems, precessing spins and spins driven in the rotating frame. In addition we will describe a variety of associated considerations including slice selection and 3D imaging. Previous Work M. Packard and R. Varian (1) first applied a prepolariza- tion pulse in a nuclear magnetic resonance experiment in 1954. They used a pulsed field of 100 gauss to polarize spins. After turning off the polarizing field nonadiabati- cally, they recorded the FID in the Earth’s magnetic field. They obtained a SNR of 20, which allowed them to esti- mate the FIDs frequency accurately enough to compute the local magnetic field to one part in 15,000. They also noted that the FID from a sample of HF showed a beat frequency of 120 Hz, consistent with the difference in zy y for hydrogen and fluorine. A. Bloom and D. Mansir in 1954 devised an experiment for measuring T, relaxation times at very low fields (2). After prepolarizing at 100 gauss, they “depolarize” the sample for a variable time at a low field of about 2 gauss. The received signal varies as e-T’T1(lOw) where T,(low) is the longitudinal relaxation time at 2 gauss. We expect that the freedom to customize polarization waveforms will allow one to create images weighted with a rich spectrum of T, dispersion contrast. Since Packard and Varian’s innovation, much basic re- search has been done using the Earth’s magnetic field as the bias field. BBnB of Geneva has worked in this area since 1949 (3). In 1977, his group published in situ T, dispersion results that discriminated between normal and pathological amniotic fluid. In 1980, he published an exhaustive overview paper (4). Borcard of the Geneva 221