Portal de Periódicos da CAPES

In memoriam: William A. Edelstein, 1944–2014

2014; Wiley; Volume: 72; Issue: 2 Linguagem: Inglês

10.1002/mrm.25296

1522-2594

Paul A. Bottomley,

Advanced NMR Techniques and Applications

So what counts is the “dash” that joins the two dates … Bill Edelstein died suddenly at home in Baltimore in February, 2014. He had been dealing with cancer for the past two years, but given his vigor—working up to his final hours on a project to make quiet MRI gradients—the news still came as a shock. Bill's “dash” was indeed large. His impact on the field of MR in medicine was immense and fundamental to the technology. Originally from Gloversville in upstate New York 1, Bill received a BS in physics from the University of Illinois in 1965 and a PhD from Harvard on “time-dependent directional correlation in 199mHg” in 1974. Robert V. Pound, who had earlier codiscovered NMR, was Bill's thesis adviser at Harvard. From 1974 to 1977, Bill went looking for gravitational waves as a postdoctoral fellow at the University of Glasgow in Scotland. He did not find any, but he did meet and marry his future wife Fiona while in Glasgow. In 1977, he moved further north to the University of Aberdeen to accept a fellowship with John Mallard in the medical physics department and work on a then-new project to build a whole-body NMR imaging scanner 2, 3. A large, open, 4-coil air-cooled magnet was purchased from Oxford Instruments (Abingdon, England). It had a vertical bore, which required that subjects squeeze horizontally between its two central hoop coils. Although its main magnetic field strength (B0) was a measly 0.04 Tesla (T), the coils heated to ≈100°C, so the first order of the day was to fabricate a water-cooling system to avoid frying anyone being scanned 2. It is often forgotten that before 1980, although the four or so groups in the world that were actively working on NMR imaging all “knew” they could make images, in fact, no one really “knew how” to make them. After Paul Lauterbur's seminal idea of using magnetic field gradients to spatially distinguish NMR signals with projections, the problem of how best to deploy the gradients to make images became the central focus of MRIs development in the late 1970s. The overriding question of how to spatially encode NMR signals in all three dimensions so that useful images could be acquired in tolerable scan times had to be solved. Whether or not success could ultimately be achieved was unknown. In particular, due to practical problems with long scan times, low signal-to-noise ratios (SNRs), nonuniform magnets, and the generation of magnetic field gradients, most groups—including Aberdeen—were exploring methods that did not employ projections for encoding in other spatial dimensions, which in fact is what is practiced today. To this end, the Aberdeen group first tried a line-scan method employing two dimensions of spatially selective excitation with a readout (projection) gradient in the third dimension 4-6. This did not work very well. Bill, in reviewing alternative two dimensional (2D) Fourier transform methods, including Peter Mansfield's echo-planar imaging method, came up with the idea of blipping the orthogonal gradient to phase encode the object in the third dimension 2. This was combined with Jim Hutchison's sequence of gradient-refocused slice-selective excitation in one dimension, plus a readout gradient with balancing lobes in the other 5, to yield the fully 3D-resolved spin-warp MRI method 7. Today, over 30 years later, phase-encoding (spin-warp) localization remains one of the three central pillars of MRI technology—the other two being the projection or readout gradient method and spatially selective excitation. Bill and the Aberdeen team's spin-warp method is routinely used in MRI today, and this will likely continue in the foreseeable future. Bill's fellowship at Aberdeen ended in 1980. He was recruited to work as a staff scientist at the General Electric (GE) Schenectady research laboratories in upstate New York, not far from his origins in Gloversville. Our manager there, the late Red Redington, was planning to build a whole-body system to perform localized NMR spectroscopy (MRS) to measure metabolism in the body. At that time, GE had decided not to pursue MRI because it could not possibly compete with X-ray CT, for which GE already had a highly successful product. In order to do MRS, a superconducting magnet with the highest possible field strength (which turned out to be 1.5 T) and uniformity was ordered from Oxford Instruments. That is where Bill and I first met as GE “newbies” arriving within the same month. The high-field magnet would not show up until 1982. To keep us out of trouble while we waited, management agreed to provide the resources to build a whole-body resistive magnet-based (0.12 T) MRI system, figuring that the technology might be useful later for performing localized MRS, although at the time no one “knew how” to do that either. Bill and I wanted to do MRI, and were surprised that it was written off by GE. We both expressed skepticism about MRS's future. In any case, despite Red Redington's best intentions, building an MRI scanner was not a total success at keeping us out of strife. We managed to get into a fair amount, mostly over the right to publish and various upsets to the GE pecking order. The upshot was a comrades-at-arms bonding and collaboration between Bill and I, which was cemented by what happened with the 1.5 Tesla project. By the time the 1.5T magnet arrived, GE's opinion on MRI had done a 180° flip. As a consequence, after accepting the magnet we were supposed to turn it down to 0.3T or maybe down to 0.5T to do MRI. We resisted. To keep it at 1.5T required that we make MRI detector coils that tuned at 64MHz. There were rules that coil wire had to be less than a fraction of the wavelength (λ), say λ/20, and concerns about radio frequency (RF) penetration. To make a 64MHz coil, we had taken to distributing the tuning impedances 8: the only rule that mattered to us was whether the thing could be tuned. The “birdcage coil” was as an extreme example, with one full λ of conductor around it 9. Around late 1983, when we were making distributed capacitance coils for 1.5T MRI, Bill made an open-ended, distributed capacitance “ladder coil” and wrapped it around a cylindrical head-coil former. It sat on our desks for months while the question of how to terminate it was pondered. Finally, Bill got motivated to “connect it to itself” after a phone conversation with Cecil Hayes at GE in Milwaukee; and together the two of them developed it, with the help from several others at the lab. Like many great ideas, the bird-cage coil turned out to be even better than first thought; it enabled true quadrature excitation and detection that saved RF power and increased the SNR, as compared to the existing linear designs 9. In addition, it improved the RF-field uniformity and reduced RF penetration artifacts in the body at 1.5T 10. The bird-cage coil went straight into the new GE Signa 1.5T MRI scanner product, as did a bunch of other inventions that made MRI work but did not otherwise appear in papers, including spin echo 11, T2 relaxation time 12 and slab 13 MRI pulse sequences; gradient “crushers” 14; and GE's distributed transverse gradient windings 15, for which Bill also played a key role 16. These technologies are still standard in MRI as practiced today. When GE announced the first 1.5T head MRI (and MRS) results at the Radiological Society of North America meeting in November 1982, Bill and I thought only that we had done some new science and just wanted to publish it. But what followed over the next few years, including plenary sessions at both August 1983 and 1984 SMRM (Society of Magnetic Resonance in Medicine; predecessor to International Society for Magnetic Resonance in Medicine) meetings in San Francisco and New York, was what I have called the MRI “field wars.” Bill and I were up there defending our work at 1.5T, and everyone else, well … not at all. Whereas the other imaging companies had committed to low field MRI systems (≤ 0.4T), GE threw massive resources and factories at building 1.5T Signa MRI scanners. This was an incredible period on many fronts. Some competitors responded with false statements about our data. These were often dished up by GE management or by potential customers for Bill and I to resolve. Bill took the attitude that the science would speak for itself and the doubters would ultimately be forced to eat their words. When it was said that we could image heads at 1.5T but not bodies, Bill built a distributed capacitance body coil, and body images were obtained and reported. Eventually, the work got published 8, and we resolved not to respond to misinformation in the trade press. After a few years, the 1.5T product took off, other companies made 1.5T systems of their own, and the fuss went away. Necessity is the mother of invention, and Bill had a habit of coming up with solutions just when they were needed. In 1983, dismayed by the poor T1 and T2 contrast in the images from our 0.12 T scanner, Bill introduced the concept of the “contrast-to-noise ratio” 17. The Appendix of that paper had a version of the fast low-angle shot MRI sequence, which created some patent issues later on. In 1985, Bill came up with the intrinsic signal-to-noise ratio (SNR) 18. This arose as we were being pressured by GE to show that the field we happened to be working at, 1.5T, was coincidentally the optimum field strength for MRI—a claim that had already been made by others for lower B0 values. Bill was charged with measuring the MRI SNR at different (B0) field strengths. A central problem was that the MRI performance at different fields could not be compared unless differences in detector performance at each B0 were properly accounted for. The intrinsic SNR does just this, yielding quantitative SNR measurements for comparing B0, scanner, and coil performance 18, 19. This work confirmed the linear dependence of SNR on B0 for external coils with sample-dominated noise, as theorized by Hoult and Lauterbur in 1979, and claims about an “optimum field” for MRI subsided. As Bill famously said at the plenary session of the 1983 SMRM meeting: “If you can't have the field you love, love the field you have.” And then there is the MRI “phased array” project for which Bill collaborated with Pete Roemer in 1989 20. The paper for this might be the fattest published in Magnetic Resonance in Medicine; it certainly is highly cited. The phased array was prompted by the quantitative SNR measurements, showing much better performance from surface coils compared to volume coils and the desire to obtain the SNR advantage without giving up the higher field-of-view provided by volume coils. To build the four-channel prototype 20, four whole quadrature “Signa” receiver chassis were expropriated, each the size of a large refrigerator. Today, phased arrays have been extended not only to 64 or more channels employing miniaturized receivers and for both transmission and reception at B0 > 1.5T, but also for spatially encoding the MRI signals directly. This has enabled some blipped phase-encoding gradients to be eliminated altogether for really fast MRI. Indeed, the MRI phased array is another example of an idea that has turned out even better than the great idea it was when first published. Bill retired from GE in 2001 and was a consultant for some years. In 2007, he moved to Johns Hopkins University in Baltimore as Visiting Distinguished Professor in the Division of MR Research of the Department of Radiology. At Johns Hopkins, he worked on RF dosimetry and fast MRI. However, his main passion was making quiet MRI gradients, a project he began while at GE and for which he had started a small company in Baltimore to pursue. Bill was a former trustee, gold medal winner, and fellow of the SMRM. He also was a GE Coolidge Fellow and a fellow of the American Institute of Physics (AIP). He received the AIP prize for Industrial Applications of Physics in 2005, an honorary DSc from the University of Aberdeen in 2007, and the University of Illinois Alumni Achievement Award in 2013 for contributions to the field of MRI. Bill is survived by his wife, Fiona; his 2 daughters, Jean and Elspeth; his son, Arthur; and his 2 grandchildren. He also was “best man” at my wedding in 1986, although I will remember him most as the “best man” to stand with during the stormy times of the development of high-field MRI in the early 1980s. And as a friend. As long as the spin-warp gradients keep blipping, along with all of his other contributions that lie at the heart of MRI technology, I doubt that Bill will be forgotten. In that sense, his “dash” is not yet ended. I mean, you must take living so seriously that even at seventy, for example, you'll plant olive trees— and not for your children, either, but because although you fear death you don't believe it, because living, I mean, weighs heavier.