Breaking the Speed Limit in MRI

Mark S. Cohen

A Statement of the Problem

Patients want to know. Physicians want to know. Legislators want to know: Why does MRI cost so much? For those of us who operate MRI centers, we might rephrase the question. Why does MRI take so long - and what is to be done about it?

At both public and private imaging centers, the majority of the expenditures represent fixed costs. As a consequence, patient throughput has an enormous effect on the bottom line. Beyond some threshold number of scans per day, additional studies reflect almost pure profit for the imaging center. Optimum patient utilization thus potentially can increase overall revenues, and thereby stabilize the amount of the fixed cost required for individual MR exams. Surely thisd explains some of our emphasis on reducing scan time. But there is a different spin we can put on the problem. With each substantive decrease in imaging time, we open a new range of applications for MRI.

Prior to the introduction, in 1982 [1] , of practical 2D multi-slice imaging, and shortly thereafter of multi-echo multi-slice imaging [2] , MRI scan times were unrealistically long for patient care. Once a brain examination could be completed in a time somewhat shorter than a patient's tolerance for the procedure, in a scan time comparable to that of CT, MR imaging's acceptance began to grow. While scan times decreased steadily with technological improvements, the next major gain in speed came with the introduction, in 1986 [3] , of the FLASH (Fast Low Angle SHot) gradient echo method. By reducing practical scan times to as little as 10 seconds, the FLASH method considerably increased MR applications in body regions, especially the abdomen, where suspended respiration could eliminate most motion related image distortions [4, 5, 6] . In many cases however, the burden of altered contrast (as compared to spin echo) in FLASH imaging has limited its acceptance [7] . Where contrast based on tissue relaxation time differences is not the key clinical issue, the speed gains with FLASH imaging can be used to tremendous advantage. MR angiography, for example, simply would not be practical without gradient echo techniques.

Cardiac MRI, though the subject of intense interest for years, has been plagued by the difficulty of reliably triggering MR scans to the cardiac cycle. Each error in triggering potentially results in tremendous image degradation. When scans could be completed in a time short compared to the human heartbeat, a major physiological threshold was reached. In the past few years, at least two major techniques have cleared the hurdle of sub-second imaging: practical echo planar imaging in 1987, and sub-second FLASH methods three years later. While most gross organ motion can be resolved with these scan times of a few tenths of a second, there remain essential physiological events that require still better temporal resolution. Precise scan synchronization allows visualization of temporal details in, for example, cardiac function, but many physiological processes, such as peristalsis, swallowing, saccadic eye motion, joint clicks, phonation, and so on, are not cyclically repetitive. The most recent threshold of scanner speed vs. physiology was therefore crossed when MRI scan times reached the tenth of a second range. MRI now can examine such details of function as mental processes, rendering brain physiology and pathophysiology for the first time subject to minimally invasive, in vivo imaging. Ultimately, the major impact of high speed imaging may be on a highly expanded range of applications, rather than a simple increase in the number of scans routinely obtained per site.

FIGURE 1. Relationship between physiological processes and imaging speeds of available methods. In order to avoid image artifacts from motion, scan times must be not be longer than the duration of the motion. In order to study that motion, scan times must be five to ten times faster.

What Limits Scan Time?

The nature of the MR signal imposes complex constraints on scanning speed. The free induction decays (FIDs) or spin echoes from biological tissues are very brief, seldom lasting more than a few tenths of a second (limited by "T2*" or "T2"). Because we generally cannot perform the spatial encoding necessary to convert this signal to an image in such short times, current scanning technology is inefficient: the signal must be sampled repeatedly to acquire sufficient data to build a complete image. Unfortunately, when we go back to sample the signal again, its amplitude is reduced. In essence, the act of sampling the signal demagnetizes the patient, and we must wait for remagnetization to occur before there is adequate signal for another sample to be taken. The remagnetization process is quite slow, limited by "T1." If forming an image requires 256 samples, and we must wait 3 seconds for recovery between samples (TR), the imaging time becomes more than 12 minutes.

Speed Enhancement Methods

Myriad methods have been proposed for reducing MR scan times. Before throwing ourselves into the deep waters of unfamiliar technology, it may be useful to get our bearings within the technological mainstream. The most widely used methods of speed enhancement have been applied primarily to reducing table or magnet time for standard clinical applications. The ultra-fast techniques, on the other hand, have provided the means for new applications of MR imaging.

Technical Foundations of Fast Imaging

Anyone who has designed imaging protocosl to complete a clinical examination in a fixed time slot is acutely familiar with the basic imaging time equation:

Scan Time = TR x Number of Phase Encodes x NEX

where TR is the repetition time between successive RF pulses, the Number of Phase Encodes determines image spatial resolution, and NEX is the number of averages of the data required to form a sufficiently noise-free image. The three terms in that equation have resulted in three major directions towards the reduction of scan times: To explain the present state-of-the-art, we will touch on each of these.

How can TR be reduced? What are the costs?

Cutting back on TR is a simple way to reduce scan times, but its implications are complex. When TR is reduced, the magnetization from tissues with long T1's does not recover fully, and the MR signal is diminished. Because the signal strength depends on T1 as TR is shortened, the images become increasingly T1-weighted. More contrast is good, right? In fact, no. As a general rule, in MRI contrast arises from the reduction in signal of some tissues, so that they can be seen as dark against others that remain bright. But signal is a limited currency in MRI. Except where T1-contrast is desired, cutting the TR generally compromises image quality.

One way to overcome this is to use an excitatory RF pulse which disturbs the magnetization to a lesser degree: a so-called shallow RF pulse. The flip angle of the RF pulse describes how much of the magnetization is converted to MR signal with each excitation. A 90 degree; pulse fully converts that magnetization; the signal is, in general, proportional to the sine of the RF flip angle (figure 2).

Figure 2. The strength of the MR signal is shown in [blue] and is equal to the projection of the magnetization on the transverse or "imaging" plane. Large flip angles yield larger MR signals.

Given that shallow flip angles yield less signal, of what use is a shallow flip technique? The answer lies in the realization that, with small flip angles which convert less magnetization, less time is required for the protons to re-magnetize after the excitation pulse (Figure 3).

Figure 3. When small flip angles are used for excitation, there is relatively little loss of longitudinal magnetization (shown vertically). As a result, the tissues will re-magnetize more rapidly following a small excitation pulse.

The first imaging technique to utilize this reduced flip angle advantage was the FLASH method of Frahm and Haase [3] . Though small flip angles cannot yield signals as strong as those of large flip angles, it is possible at any TR to specify a flip angle, less than 90 degree;, which will give the maximum signal possible at that TR. Generally speaking, with FLASH-type scanning it is possible to control the T1 contrast weighting with both flip angle and TR: as the TR is increased, or the flip is angle decreased, the T1 weighting in the image is reduced (figure 4).

Unfortunately, FLASH imaging does not admit the use of 180 degree; RF pulses to form spin echoes. Therefore it is not possible to use TE to control T2-contrast in the same fashion as conventional imaging. The contrast behavior of FLASH scans is therefore unlike that of spin echo. Where contrast based on T1 or proton density is sufficient, FLASH has proven a reliable alternative to spin echo scans [4, 5, 6] with a scan time short enough for suspended respiration. Omission of the 180 degree; pulse makes FLASH images more susceptible to the effects of iron-containing substances that distort the magnetic field. This too, has been used to clinical advantage in, for example, the evaluation of hemorrhage [8, 7, 9] . As we will see below, the relatively high sensitivity of FLASH and other gradient echo imaging techniques to magnetic field distortions has been helped the development of functional brain imaging.

Hardware permitting, the FLASH method may be extended to the domain of extremely short TR's if sufficiently small flip angles are used. This has been the goal in the development of "snapshot" or "turbo" FLASH methods [10, 11, 12, 13] . A TR of 5 msec and flip angle of 6 degree; allows even moving structures, such as the heart, to be scanned in about one-half second (using a 128 line imaging matrix) with minimal motion artifact. Because small flip angles must be used, this can yield only one-twentieth of the signal available with a 90 degree; flip and a long TR. Many of today's commercial scanners now allow TR's of well under 10 msec and have incorporated "RF spoiling" to provide more precise control of image contrast in FLASH scans.

The generalized success of FLASH imaging has resulted in the proliferation of a variety of small flip angle pulse sequences known by the acronyms GRASS, FISP, FAST, CE-FAST, SPGR, SSFP, PFI, PSIF, FFE, and so on. This short article cannot possibly hope to summarize all of these methods, but the interested reader is invited to read a more comprehensive review in chapter 31A of Taveras' Radiology [14] .

k-space strategies

Understandably, many radiologists are wary of the term "k-space." Less mysterious than it sounds, k-space is simply the name given to the representation of the raw data which form the MR image. A process known as the Fourier transform converts the k-space data into an image. Forming a 256 x 256 point MR image requires as many separate points of raw data (k-space). Conventionally the k-space data are acquired a single line at a time, separated by a TR interval, because the gradient and data acquisition subsystems of standard imagers require from four to ten milliseconds to spatially encode and record each data line. Since T2 (or more accurately, T2*) decay is taking place during this time, there is generally not sufficient time to take more than one line before the signal has decayed away.

Changing the Matrix Size

Every additional data line we collect in the MR image requires an additional TR's worth of imaging time. If fewer pixels (lower spatial resolution) can be tolerated in the MR image, then reducing the matrix size is an efficient way to cut scan times. Contemporary scanners allow a range of matrix sizes. Some allow a completely free choice of imaging matrix, so that the physician can effectively balance imaging time against the desired spatial resolution.

Strip Scanning - Rectangular FOV

A straightforward way to reduce scan times is to collect image data from only the designated body regions. As obvious as this would seem, the practical introduction of rectangular field-of-view "strip scans" required substantial technological development. For example, aliasing or "wrap-around" is difficult to control in such scans. When combined with free matrix size selection, the scanning prescription can be devised to minimize scan time for any desired resolution and field-of-view

Conjugate Synthesis - Half-NEX or Half-Fourier Scanning

A certain redundancy exists in the raw data used for MR imaging. It was clear in the mid-80's that a modest amount of processing should make it possible to form a complete MR image by collecting only a little more than half of the MR raw data. This principle underlies the half-NEX, or half-Fourier, scanning method. Use of the method allows retention of the spatial and contrast resolution provided by full data acquisition at the cost of a modest loss in SNR and slightly increased sensitivity to motion artifacts.

RARE or "Fast Spin Echo"

Rather than pausing between each separate excitation (a whole TR period) to between each data line in k-space, it is possible to collect an additional data line in the same image by encoding an additional spin echo. This is the strategy adopted in RARE (sometimes called "Fast Spin Echo" or "FAISE", a sequence first developed some seven years ago [15] , but only recently brought into routine practice [16] . With RARE, the total scan time for an image is reduced in direct proportion to the number of separate echoes. Because the actual echo time is different for different portions of the MR raw data, the contrast behavior of RARE imaging is quite complex. For practical purposes, however, the RARE sequence can emulate standard spin echo contrast behavior rather effectively [16] . Because many such spin echoes can be formed following a single RF pulse, RARE can be very effective in cutting scan times; sixteen-fold reductions are achievable.

Echo Planar Imaging

Even without the intervening 180 degree; pulses used in RARE imaging, it is possible to collect multiple data lines following a single RF excitation. The echo-planar imaging (EPI) method seeks to collect all of the data lines for a complete image following a single excitation [17] . One way to conceive of EPI is to compare it to a RARE sequence in which the intervening 180 degree; pulses between data lines are replaced by gradient echoes. This can work only of the gradient echoes may be formed very rapidly. In practice, therefore an EPI-equipped scanner requires the addition of substantial special purpose hardware, including high power gradients, rapid data collection hardware and modified image processing software (Figure 5, equipment block diagram for Instascan). The results are dramatic, however: with EPI it is routinely possible to acquire complete MR images in as little as 30 msec and as rapidly as 20 frames images per second. Furthermore, because only a single excitation pulse is used, the signal losses associated with reduced TR approaches are eliminated. Compared to an optimally designed FLASH sequence with a 5 msec TR (one-half second scan time) the EPI signal is some twenty times larger. When additional considerations of bandwidth are accounted for, the EPI scan still yields a three to four fold SNR advantage over the much slower FLASH method [18] . In our lab, we utilize the Instascan method of Rzedzian [19] , an echo-planar derivative compatible with high field imaging. Our unit is supplied by Advanced NMR Systems, inc. of Wilmington, Mass., as a retrofit to a General Electric Signa® scanner. Since EPI uses only a single excitation pulse to form an image, it is possible to create "infinite" TR scans with arbitrarily long periods between successive 90 degree; pulses, and therefore with no signal losses from T1 effects. Generally speaking, scans acquired with TR more than 5 times the tissue T1 are denoted as TR =infinity.

Figure 5. Equipment block diagram showing hardware modifications required for the addition of Instascan echo planar imaging to existing hardware.

Applications of Ultra Fast Imaging: How Fast is Fast Enough?

Motion Artifact Reduction

Certainly one of the most straightforward applications of high speed imaging is in the reduction of motion artifacts such as those which occur in the abdomen. There are two general strategies: one can either scan fast enough for breath-holding [4, 6] or scan fast compared to abdominal motion [20, 21].

The development of positive MR contrast agents for use in imaging the human bowel has been hampered by prominent smearing artifacts produced by the agents in the presence of respiratory and peristaltic motion. Since the latter is involuntary, it is essential that the MR images be obtained rapidly compared to normal bowel motion. As demonstrated by Hahn at the MGH, standard CT contrast agents can be used to great advantage in MR imaging of the gastrointestinal system when combined with ultra-fast MRI [22] (figure 6).

Figure 6. T2-weighted Instascan (echo-planar) images of the gut after administration of the readily-available CT contrast agent, Readi-Cat 2 (E-Z-M corporation). The long T2 of the agent results in dramatic signal enhancement in these scans. Matrix 128 x 128, TR = ∞, TE= 75 msec, slice thickness 10 mm.

While the most obvious motion artifact challenges are in the thorax and abdomen, a surprising number of cranial MR images become technical failures due to patient motion. Ultra-fast imaging may be a useful alternative in such cases (Figure 7).

Cardiac Imaging

Though many methods have been demonstrated for cardiac MR imaging, those relying on cardiac synchronization (triggering or gating) for motion artifact control still suffer in dependability both from unreliable ECG lead setups or, more insidiously, from cardiac arrythmias in the typical cardiac patient. This, coupled with the long and therefore costly cardiac MR scanning process, has done much to limit the acceptance of cardiac MRI. On the other hand, with its non-invasiveness, its flexibility of scan plane selection, and its high intrinsic contrast, MRI should be a near ideal cardiac imaging tool. The earliest EPI or Instascan cardiac images [23, 24, 25, 26] were limited to axial views. By modifying both hardware and software on the commercial Instascan device, Weisskoff was able to demonstrate oblique echo -planar imaging capabilities [WEISSKOFF, (Figure 8). At about the same time, snapshot FLASH images of the heart were demonstrated. With these speed advances, high speed cardiac imaging is making significant advances in clinical and scientific applications.

Kinematics and Orthopedic Function

With its extraordinary spatial resolution, NMR would seem to be an ideal tool for the study of joint articulation and muscle involvement in movement. As demonstrated by Shellock and others, NMR may be used to investigate patellar tracking abnormalities [27] . In the original methods, because of even fast gradient echo imaging's long scan times, it was necessary to study joint motion by acquiring a series of static images at a variety of articulatory positions. The addition of true real-time imaging has greatly assisted in the study of active joint motions [28] .

It has been recognized for some time that the NMR signal from skeletal muscle changes dramatically with exercise [29] . By combining ultra-fast T2-weighted imaging with active exercise - within the magnet - it has been possible to elucidate the time course of such signal changes and so to demonstrate recruitment patterns of muscles during complex exercise [30] (Figure 9). Such studies have clear implications for both clinical and basic scientific research.

Figure 9. Transverse sections of the thigh during active knee flexion. Following only thirty seconds of exercise, changes in the signal intensity of the hamstring muscles are clearly visible, as a result of large T2 changes. As exercise progresses, the signal intensity of the various synergists in this region increase (unequally) until fatigue. Such variations in signal changes may reflect both use and recruitment patterns during this active joint motion. TR = ∞, TE = 60 msec, slice thickness 10 mm, 64 x 128 matrix.

Precise Motion Detection and Measurement

As shown more than 25 years ago by Stejskal and Tanner [31] , and later by Le Bihan in the context of MR imaging [32] , adding additional gradient pulses to the MR imaging sequence allows proton diffusion to be quantified by magnetic resonance. Unfortunately, doing so makes the MR images extraordinarily sensitive to patient motion, and unacceptably large motion artifacts result. If ultra-fast imaging techniques are used, however, it becomes possible to measure precisely the remaining water motion, which results in a well-characterized signal loss in the MR image. At least two major applications of such "diffusion" imaging have been identified. Working at UC San Francisco, Moseley and his collaborators have used diffusion imaging to assess the early effects of stroke [33] . The vascular catastrophe of stroke leads to enormous costs in morbidity and mortality. When neurons are deprived of their supply of glucose and oxygen their metabolic activity is compromised. When these metabolic deprivations are prolonged, irreversible cell damage is the natural consequence. When cerebral ischemia is not the result of cardiac dysfunction, it is generally regional, due to obstruction of hemorrhage of one or more intracerebral vessels. The ischemic insult may arise from a variety of factors, including stenosis of the vessel lumen, cardiac insufficiency, decreased blood oxygen, increased blood viscosity, coagulation or embolis [34] . Whatever the cause, it is increasingly clear that there exists cerebrovascular resistance to the initial hemodynamic insult. It is in the post-ischemic phase that intervention could have the most benefit. If intervention is not successful, infarction will spread radially towards the periphery from the initial necrotic center. This penumbral zone of non-functioning, but viable, tissue may be spared if blood flow is restored [35, 36] . Moseley's group has shown that a drop in the apparent diffusion coefficient appears to occur before any other MR-detectable effects and may be visible within fifteen minutes of an occlusive event. Such early detection may prove crucial in stroke management.

Another key application arises from the realization that diffusion is strongly temperature dependent. Our group, working with collaborators at the Brigham and Women's Hospital [37, 18] , and that of Le Bihan at the National Institutes of Health [38] , has shown that these temperature sensitive techniques may be applicable in the real-time monitoring of intraparenchymal surgery, potentially minimizing the surgical invasiveness and morbidity of even intracranial surgery. Since adequately diffusion-weighted imaging can now be performed using suitably modified gradient echo sequences, this is likely to become an area of increasing interest and growth.

Slightly coarser grained motions may also be measured with ultra-fast MRI. In a series of experiments inspired by the work of Feinberg, Poncelet has shown that when combined with velocity encoding gradients, echo planar imaging may be used to quantify precisely the pulsation-induced three-dimensional motion of the human brain within the calvarium [39] . Using similar imaging technology, but a different analysis strategy, Wedeen [40] has shown that velocity-dependent phase changes may be used to generate strain-rate images of the human heart, in which local deformations of the myocardium (as opposed to bulk tissue displacements) can be seen readily. Using this method it is possible to assess accurately the contractile activity of any voxel of myocardial muscle.

Contrast-Agent Dynamics and Cerebral Perfusion

Perhaps the most exciting application of ultra-fast MR imaging has been its use in the quantitative assessment of cerebral perfusion. These developments were assisted both by technological advances in MRI which resulted in subsecond imaging and a growing understanding of exogenous and endogenous contrast agents. In neuroimaging, these achievements primarily were accomplished by exploiting high magnetic susceptibility contrast agents, whether exogenously administered (e.g. Gadolinium-DTPA (Gd-DTPA)) or intrinsic (e.g. deoxyhemoglobin).

As commonly utilized in the CNS, injectable Gd-DTPA provides enhanced T1 relaxation when the blood brain barrier (BBB) is disrupted grossly. Since these T1 effects are local, and because the intravascular space within the brain is small, such agents have little impact in when the BBB is intact. Some tumors, acute infarcts, and most neurodegenerative diseases do not result in BBB destruction, while in others, such as multiple sclerosis, the disruption is variable [41] . As a result, conventional Gd-DTPA-enhanced imaging may not be adequate to characterize these disease states fully.

Susceptibility (T2 or T2*) -based contrast occurs over greater distances. When paramagnetic substances such as Gadolinium or Dysprosium are confined within the vasculature, they induce magnetic field distortions which extend beyond the vascular boundaries and result in transverse relaxation enhancement of the surrounding tissue protons [42] . Since the intact BBB effectively compartmentalizes injected Gd-DTPA, it is possible to use this effect to track signal intensity changes following bolus contrast agent administration. The resulting relaxation enhancement shows a near linear relationship to the intravascular contrast agent concentration, at least within the physiological range of study [43] . When combined with ultra-fast imaging during the passage of Gd-DTPA through the cerebral vasculature, it is thus possible to derive concentration time curves of the cerebral transit of the injected contrast agent [44] and, by using the principles of tracer kinetics, to derive cerebral blood volume (CBV) maps [45, 46] . Interestingly, depending upon the MR imaging sequence used, the signal intensity decreases have a variable dependence on the size of the vessel carrying the paramagnetic agent. Images based on gradient echo contrast demonstrate total blood volume, while spin-echo based contrast, as achievable with Instascan imaging, is biased towards the microvascular volume [43] . At our center we utilize such spin echo scans for their relative insensitivity to artifacts as well as for their optimum tissue sensitivity.

An important application of CBV imaging is in the evaluation of stroke, as demonstrated by Rosen [47] . The absence of perfusion within in farcted tissues is conspicuous in the lack of signal change during contrast agent passage. In animal models the results compare favorably to those obtained with more conventional perfusion measures [45, 48] . Brain tumors are generally accompanied by neovascularization, which supports the increased metabolic requirements of active tumor. It is thus reasonable to expect that the MRI-derived cerebral blood volume maps might find application in the study of cerebral neoplasm. Because each echo-planar image is acquired in about 1/10th of a second, several slices can be acquired at once while still achieving good temporal resolution with this method. Figure 10, from the work of Aronen [49] , shows results from our multi-slice imaging protocol. The multi-slice technique affords clear advantage in that the volume extent of the lesion can be evaluated clearly without the need for repeated contrast agent administration. In a study of thirteen tumor patients, the MRI cerebral blood volume mapping technique was able to demonstrate features of tumor activity not seen by conventional MRI and, when used longitudinally, to provide information about microvascular changes that occurred secondarily to either radio- or chemotherapy [49] . Perhaps more surprising, the correspondence between the MRI CBV maps on the one hand and those derived using positron-emission tomography (PET) with 18F fluorodeoxyglucose (18FDG) on the other, was good to excellent in the majority of the tumor patients studied [44]. these results may imply a linkage between the metabolism, angiogenesis, microvascular density, and tumor grade.

Figure 10. Multi-slice cerebral blood volume (CBV) map of tumor patient. Using a single injection of Gadolinium, this multi-slice CBV map was calculated from the signal intensity time course derived from over 300 separate images. The grey scale represents relative cerebral blood volume.

Functional Brain Imaging

The energetic demands of neuronal activity are considerable, and a coupling between electrical excitation and nutrient delivery was hypothesized over 100 years ago. This metabolic coupling has provided the basis over recent years for an explosive growth in the use of PET for neurophysiological study in humans. Postulating that cerebral blood volumes as shown on MRI should be similarly sensitive, Belliveau demonstrated, with using a dual injection protocol, that blood volume changes on the order of 30% occur between the resting state and conditions of photic (visual) stimulation, and that these hemodynamic changes are well localized to the primary visual cortex [50] . At this juncture, contrast-enhanced high speed MRI made the jump from largely anatomical imaging to nearly direct assessment of neuronal function.

The method developed by Belliveau in 1991 required the use of exogenous contrast agents. Some years before, Thulborn had shown that paramagnetic changes in hemoglobin result in MR-visible decreases in T2 as hemoglobin goes from an oxygenated to a deoxygenated states [51] . Ogawa, using conventional strategies [52] , and Turner using echo-planar methods [53] , have shown that signal intensity changes are apparent in apnea in animal models. Based on these results, Kwong suspected that echo-planar MRI could be used to follow changes in brain oxygenation in humans, first during breath-holding and ultimately during photic stimulation [54] . Similar results now have been achieved using a variety of high speed imaging methods and field strengths, and have resulted in an enormous wealth of data on the use of MRI in functional studies of the human brain [55, 56] . The oxygen contrast methods rely on the hyperperfusion of brain tissue which accompanies neuronal activity. The consequence of this increased blood supply is an increase in the oxygen content of the venous blood, implying that blood flow increases exceed the metabolic requirements of the cells. Kwong's study used both flow-dependent (T1) contrast and O2 dependent (T2*) contrast and showed that the onset of the O2 increases had a rise time of about 4.4 seconds. This suggests that the temporal resolution requirements of the functional MR imaging method must be of about the same order - that is, a few seconds. Because this is long compared to the speed capabilities of the echo-planar experiment, as in the CBV mapping experiments discussed earlier, EPI will provide the advantage of multi-slice or volume imaging for mapping of human brain function. Figure 11.

Figure 11 shows the activation of primary motor cortex during finger tapping moving of either the right or left hand (data from Chantal Stern, Massacusetts General Hospital).

CONCLUSIONS

From its recent position as a scanning technique, suitable only for the study of stationary tissue in immobile patients, MRI has moved to a premiere position in the observation of dynamic physiological processes, as a result of dramatic improvements in its speed. Though driven primarily by the practical desire to lower scan times and thus improve throughput, increases in MR imaging speed have pushed the modality to the forefront of investigations of the human brain and have made possible the advent of functional neuroimaging.

In their book, NMR Imaging in Biomedicine [57] Mansfield and Morris wrote:

"Although reported only at an anecdotal level, it is known that some of the early pioneers of NMR had used themselves as specimens in some early biological experiments. For example, as long ago as thirty years, Edward Purcell is reputed to have placed his head in a suitable NMR coil and magnet in an effort to try to observe differences in NMR signal shape or line shape caused by his thinking intensively or concentrating hard on a specific task as opposed to, as far as possible, making the mind blank. Erwin Hahn also tried experiments of a similar nature on his own head. In all cases there appeared to be no difference between the observed signals when the brain was idling as opposed to when the brain was concentrating. No effects, adverse or otherwise, were reported. Although these experiments could in no way be interpreted as imaging experiments, or indeed, forerunners of NMR imaging experiments, it is clear that the early practitioners of NMR had strongly in their minds the possibility of applying NMR to the study of biological systems."
They could hardly have been more prescient.

REFERENCES

1. Crooks, L., M. Arakawa, J. Hoenninger, J. Watts, R. McRee, L. Kaufman, P. L. Davis, A. R. Margulis and J. DeGroot. "Nuclear magnetic resonance whole-body imager operating at 3.5 KGauss." Radiology. 143(1): 169-74, 1982.

2. Crooks, L. E., D. A. Ortendahl, L. Kaufman, J. Hoenninger, M. Arakawa, J. Watts, C. R. Cannon, Z. M. Brant, P. L. Davis and A. R. Margulis. "Clinical efficiency of nuclear magnetic resonance imaging." Radiology. 146(1): 123-8, 1983.

3. Frahm, J., A. Haase and D. Matthaei. "Rapid NMR imaging of dynamic processes using the FLASH technique." Magn Res Med. 3: 321-327, 1986.

4. Edelman, R., P. Hahn, R. Buxton, J. Wittenberg, J. Ferrucci, S. Saini and T. Brady. "Rapid MR imaging with suspended respiration: clinical application in the liver." Radiology. 161: 125-131, 1986.

5. Unger, E. C., M. S. Cohen, R. A. Gatenby, M. R. Clair, T. R. Brown, S. J. Nelson and J. S. McGlone. "Single breath-holding scans of the abdomen using FISP and FLASH at 1.5 T." J Comput Assist Tomogr. 12(4): 575-83, 1988.

6. Unger, E., A. Darkazanli and M. Cohen. "Fast MR scanning reduces artifacts in the abdomen." Diagnostic Imaging. 11(11): 248-256, 1989.

7. Atlas, S., A. Mark, E. Fram and R. Grossman. "Vascular intracranial lesions: applications of gradient-echo MR imaging." Radiology. 169: 455-461, 1988.

8. Unger, E., M. Cohen and T. Brown. "FISP and FLASH of hemorrhage." Magnetic Resonance Imaging. 5(1): 1987.

9. Unger, E. C., M. S. Cohen and T. R. Brown. "Gradient-echo imaging of hemorrhage at 1.5 Tesla." Magn Reson Imaging. 7(2): 163-72, 1989.

10. Atkinson, D. and R. Edelman. "Ultrafast 2D FT cardiac imaging." Society of Magnetic Resonance in Medicine. : 137, 1988.

11. Atkinson, D., D. Burstein and R. Edelman. "First-pass cardiac perfusion: evaluation with ultrafast MR imaging." Radiology. 174: 757-762, 1990.

12. Frahm, J., K. Merboldt, H. Bruhn, M. Gyngell, W. Hänicke and D. Chien. "0.3-second FLASH MRI of the human heart." Magnetic Resonance in Medicine. 13(1): 150-157, 1990.

13. Haase, A. "Snapshot FLASH MRI. Applications to T1, T2, and chemical shift imaging." Mag Reson Med. 13: 77-89, 1990.

14. Cohen, M. "Rapid MR Imaging: techniques and performance characteristics." Radiology. Taveras and Ferrucci ed. 1992 Lippincott. New York.

15. Hennig, J., A. Nauerth and H. Friedburg. "RARE imaging: a fast method for clinical MR." Magnetic Resonance in Medicine. 3: 823-833, 1986.

16. Mulkern, R. V., S. T. Wong, C. Winalski and F. A. Jolesz. "Contrast manipulation and artifact assessment of 2D and 3D RARE sequences." Magn Reson Imaging. 8(5): 557-66, 1990.

17. Mansfield, P. "Multi-planar image formation using NMR spin echoes." J Phys C. 10: L55-L58, 1977.

18. Cohen, M. S. and R. M. Weisskoff. "Ultra-fast imaging." Magn Reson Imaging. 9(1): 1-37, 1991.

19. Rzedzian, R. "A method for instant whole-body MR imaging at 2.0 Tesla and system design considerations in its implementation." Society for Magnetic Resonance in Medicine. : 229, 1987.

20. Saini, S., D. Stark, R. Rzedzian, I. Pykett, R. E, H. PF, W. J and J. Ferrucci. "Forty-millisecond imaging of the abdomen at 2.0 T." Radiology. 173: 111-116, 1989.

21. Saini, S. and M. Cohen. "Ultrafast Liver Imaging." Liver Imaging: New Techniques. Ferrucci ed. 1990 Andover Medical. Andover.

22. Hahn, P. F., S. Saini, M. S. Cohen, M. Goldberg, P. Reimer and P. R. Mueller. "An aqueous gastrointestinal contrast agent for use in echo-planar MR imaging." Magn Reson Med. 25(2): 380-3, 1992.

23. Doyle, M., B. Chapman, R. Turner, R. J. Ordidge, M. Cawley, R. Coxon, P. Glover, R. E. Coupland, G. K. Morris, B. S. Worthington and e. al. "Real-time cardiac imaging of adults at video frame rates by magnetic resonance imaging [letter]." Lancet. 2(8508): 682, 1986.

24. Chapman, B., R. Turner, R. J. Ordidge, M. Doyle, M. Cawley, R. Coxon, P. Glover and P. Mansfield. "Real-time movie imaging from a single cardiac cycle by NMR." Magn Reson Med. 5(3): 246-54, 1987.

25. Rzedzian, R. and I. Pykett. "Instant images of the human heart using a new, whole-body MR imaging system." American Journal of Roentgenology. 149: 245-250, 1987.

26. Cohen, M., R. Weisskoff and R. Rzedzian. "Clinical Methods for "Single-Shot" Instant MR Imaging of the heart." Radiological Society of North America. : 359, 1989.

27. Shellock, F., J. Mink and J. Fox. "Patellofemoral joint: kinematic MR imaging to assess tracking abnormalities." Radiology. 168: 551-553, 1988.

28. Shellock, F., M. Cohen, T. Brady, J. Mink and M. Pfaff. "Evaluation of patellar alignment and tracking: comparison between kinematic MRI and "true" dynamic imaging by hyperscan MRI." Society for Magnetic Resonance Imaging. : 1991.

29. Fleckenstein, J., R. Canby, R. Parkey and R. Peshock. "Acute effects of exercise on MR imaging of skeletal muscle in normal volunteers." Am J Roent. 151: 231-237, 1988.

30. Cohen, M., F. Shellock, K. Nadeau, J. Oldershaw, J. Boxerman, R. Weisskoff and T. Brady. "Acute muscle T2 changes associated with exercise." Tenth Annual Meeting of the Society of Magnetic Resonance in Medicine. : 107, 1991.

31. Stejskal, E. and J. Tanner. "Spin diffusion measurements: spin echoes in the presence of a time-dependent field gradient." Journal of Chemical Physics. 42(1): 288-292, 1965.

32. Le Bihan, D., E. Breton, D. Lallemand, P. Grenier, E. Cabanis and M. Laval-Jeantet. "MR imaging of intravoxel incoherent motions: application to diffusion and perfusion in neurologic disorders." Radiology. 161: 401-407, 1986.

33. Moseley, M., J. Kucharczyk, J. Mintorovitch, Y. Cohen, J. Kurhanewicz, N. Derugin, H. Asgari and D. Norman. "Diffusion-weighted MR imaging of acute stroke: correlation of T2 weighted and magnetic susceptibility-enhanced MR imaging in cats." AJNR. 11: 423-429, 1990.

34. Brant-Zawadzki, M. and W. Kucharczyk. "Vascular Disease: Ischemia." Magnetic resonance imaging of the central nervous system. Brant-Zawadzki and Norman ed. 1987 Raven Press. New York.

35. Astrup, J., B. K. Siesjo and L. Symon. "Thresholds in cerebral ischemia - the ischemic penumbra." Stroke. 12(6): 723-725, 1981.

36. Hakim, A. M. "The cerebral ischemic penumbra." Can J Neurol Sci. 14: 557-559, 1987.

37. Bleier, A. R., F. A. Jolesz, M. S. Cohen, R. M. Weisskoff, J. J. Dalcanton, N. Higuchi, D. A. Feinberg, B. R. Rosen, R. C. McKinstry and S. G. Hushek. "Real-time magnetic resonance imaging of laser heat deposition in tissue." Magn Reson Med. 21(1): 132-7, 1991.

38. Le Bihan, D., J. Delannoy and R. Levin. "Temperature mapping with MR imaging of molecular diffusion: application to hyperthermia." Radiology. 171: 853-857, 1989.

39. Poncelet, B., V. Wedeen, R. Weisskoff and M. Cohen. "Measurement of brain parenchyma motion with ciné echo planar MRI." Radiology. in press: 1992.

40. Wedeen, V. "Magnetic resonance imaging of myocardial kinematics. Technique to detect, localize and quantify the strain rates of the active human myocardium." Magn Res Med. 27: 52-67, 1992.

41. Kermode, A. G., P. S. Tofts, A. J. Thompson, D. G. MacManus, P. Rudge, B. E. Kendall, D. P. E. Kingsley, I. F. Moseley, E. P. G. H. du Boulay and W. I. McDonald. "Articles: Heterogeneity of blood-brain barrier changes in multiple sclerosis: An MRI study with gadolinium-DTPA enhancement." Neurology. 40(2): 229-235, 1990.

42. Villringer, A., B. R. Rosen, J. W. Belliveau, J. L. Ackerman, R. B. Lauffer, R. B. Buxton, Y. S. Chao, V. J. Wedeen and T. J. Brady. "Dynamic imaging with lanthanide chelates in normal brain: contrast due to magnetic susceptibility effects." Magn Reson Med. 6(2): 164-74, 1988.

43. Fisel, C. R., J. L. Ackerman, R. B. Buxton, L. Garrido, J. W. Belliveau, B. R. Rosen and T. J. Brady. "MR contrast due to microscopically heterogeneous magnetic susceptibility: Numerical simulations and applications to cerebral physiology." Magn Reson Med. 17: 336, 1991.

44. Rosen, B. R., J. W. Belliveau, H. J. Aronen, D. Kennedy, B. R. Buchbinder, A. Fischman, M. Gruber, J. Glas, R. M. Weisskoff, M. S. Cohen and e. al. "Susceptibility contrast imaging of cerebral blood volume: human experience." Magn Reson Med. 22(2): 293-9, 1991.

45. Rosen, B., J. Belliveau and D. Chien. "Perfusion imaging by nuclear magnetic resonance." Magn Res Q. 5(4): 263-281, 1989.

46. Rosen, B., J. Belliveau, B. Buchbinder, K. Kwong, L. Porkka, R. Fisel, R. Weisskoff, M. Neuder, H. Aronen, M. Cohen, A. Hopkins and T. Brady. "Contrast agents and cerebral hemodynamics." Magnetic Resonance in Medicine. 19: 285-292, 1991.

47. Rosen, B. "MR studies of perfusion in the brain." Current Practice in Radiology. Thrall ed. 1992 Mosby Yearbook. Philadelphia, PA.

48. Moseley, M., Y. Cohen, J. Mintorovitch, L. Chileuitt, H. Shimizu, J. Kucharczyk, M. Wendland and P. Weinstein. "Early detection of regional cerebral ischemia in cats: comparison of diffusion and T2 weighted MRI and spectroscopy." Magn Reson Med. 14: 330-346, 1990.

49. Aronen, H., M. Cohen, J. Belliveau, J. Fordham and B. Rosen. "Ultrafast imaging of brain tumors." Topics in Magnetic Resonance Imaging. Yuh ed. 1992 (in press)

50. Belliveau, J. W., D. N. Kennedy Jr., R. C. McKinstry, B. R. Buchbinder, R. M. Weisskoff, M. S. Cohen, J. M. Vevea, T. J. Brady and B. R. Rosen. "Functional mapping of the human visual cortex by magnetic resonance imaging." Science. 254(5032): 716-9, 1991.

51. Thulborn, K. R., J. C. Waterton, P. M. Matthews and G. K. Radda. "Oxygenation dependence of the transverse relaxation time of water protons in whole blood at high field." Biochim Biophys Acta. 714: 265-270, 1982.

52. Ogawa, S., T. M. Lee, A. R. Kay and D. W. Tank. "Brain magnetic resonance imaging with contrast dependent on blood oxygenation." Proc Natl Acad Sci U S A. 87(24): 9868-72, 1990.

53. Turner, R., D. Le Bihan, C. T. Moonen, D. Despres and J. Frank. "Echo-planar time course MRI of cat brain oxygenation changes." Magn Reson Med. 22(1): 159-66, 1991.

54. Kwong, K. K., J. W. Belliveau, D. A. Chesler, I. E. Goldberg, R. M. Weisskoff, B. P. Poncelet, D. N. Kennedy, B. E. Hoppel, M. S. Cohen, R. Turner and e. al. "Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation." Proc Natl Acad Sci U S A. 89(12): 5675-9, 1992.

55. Bandettini, P., E. Wong, R. Hinks, R. Tikofsky and J. Hyde. "Time course EPI of human brain function during task activation." Magn Res Med. 25: 390-397, 1992.

56. Ogawa, S., D. W. Tank, R. Menon, J. M. Ellermann, S. G. Kim, H. Merkle and K. Ugurbil. "Intrinsic signal changes accompanying sensory stimulation: functional brain mapping with magnetic resonance imaging." Proc Natl Acad Sci U S A. 89(13): 5951-5, 1992.

57. Mansfield, P. and P. G. Morris. "NMR Imaging in Biomedicine." Advances in Magnetic Resonance. Waugh ed. 1982 Academic Press. New York.