University of Rochester MRI Research:
Development of Intermolecular Multiple-Quantum Coherence (iMQC) MR Imaging Technique

1. Intermolecular multiple-quantum coherence (iMQC)
2. iZQC, iDQC and other iMQC imaging
3. Some important results from our preliminary works in iMQC

1.  Intermolecular multiple-quantum coherence (iMQC)

The perturbation of a nuclear spin system by RF pulses allows the excitation of coherence superpositions of energy eigenstates and their transformation into each other. A particular series of such coherence transfers forms a coherence transfer pathway. Multiple quantum coherences are one class of coherence pathways that have found extensive use in high-resolution, multidimensional NMR spectroscopy. MQCs exist among similar spins (intramolecular homonuclear MQCs), different kind of spins (intramolecular heteronuclear MQCs), or when the spin number I > 1/2 for a nuclear spin system. These coherences have been widely used for spectral editing and signal enhancement in high-resolution NMR. Noting that water dominates the signal in the body, the intramolecular DQC experiment also serves as a powerful natural water suppression technique. These types of experiments are useful both in MR spectroscopy and imaging when the tissue metabolites of low concentration, such as lactate, glutaraldehyde, or alanine are to be examined.

On the other hand, based on the traditional view of MQCs, intermolecular MQCs (iMQCs) among spins on different molecules, such as proton-proton iMQCs in water, was considered to be impossible for many years. Beginning in the early 1990s, the discovery of multiple spin echoes (MSEs) and iMQCs in water and other highly polarized systems generated tremendous interest but also controversy in the NMR community. These phenomena have been described using either the classical theory for the demagnetizing field or quantum-mechanical density matrix treatments. To date, both treatments have led to predictions of the signals for simple sequences, such as COrrelated 2D SpectroscopY (COSY) or COSY revamped by asymmetric Z gradient echo detection (CRAZED) experiments. Warren et al have established the basic theoretical background for the phenomenon of iMQC, and have demonstrated its existence in 2D spectroscopy. Additionally, Warren et al have established the connection between the demagnetizing field and intermolecular dipolar coupling. The residual dipolar couplings between distant spins are responsible for the dipolar demagnetizing field, and give rise to the iMQCs. From the classical description, these phenomena are due to the demagnetizing field produced by the spatial modulation of the nuclear magnetization arising in the sample following the second pulse in the CRAZED sequence.

Two important points are made by Warren et al of Princeton. First, despite that the perturbation of the equilibrium energy distribution is small (on the order of 10-4 for protons at 9T), the large value for the number of nuclear spins, N, ensures that higher order terms do not vanish in the thermal equilibrium density matrix, whether they become directly observable or not. Second, in order to become directly observable, there must exist in the spin Hamiltonian terms bilinear in the spin operator (i.e. of the form IziIzj) over all density matrix pairs i, j which are to be refocused. The dipolar field Hamiltonian, which arises from the integration of all pairwise dipolar interactions between distant spins meet the criterion. In particular, Warren noted that while molecular self-diffusion in liquids causes dipolar interactions to be averaged to zero for spins on nearby molecules, beyond the diffusion length corresponding to the characteristic NMR timescale (rmin ~ 10 mm), such averaging does not apply. The volume integral of r2dr over the r3 dependence of the dipolar interaction then ensures a logarithmic contribution whose magnitude depends on the ratio of the sample dimension to the diffusion length. In a spherical sample the angular dependence of the dipolar interaction ensures that the dipolar field vanishes. However, if the spherical symmetry is broken, for example, by application of a magnetic field gradient, G, then the effects of the dipolar field will be felt by the spins and the conversion of higher order coherences into observable magnetization becomes possible.

In the CRAZED pulse sequence, background gradients are minimized while a pair of gradient pulses is applied in order to select desired coherences. Coherence selection relies on relative Larmor precession frequency. For example, when the area under the second gradient pulse is twice that under the first, the subsequent signal (and hence transverse magnetization) with Larmor frequency, w, in the detection period t2 must have originated from coherences whose Larmor frequency were 2w during the evolution period t1. This strongly indicates that the iDQC signal is highly sensitive to local magnetic susceptibility variations.

[chapter top] [return]

2.  iZQC, iDQC and other iMQC imaging

There has recently been great interest in the potential of iMQC contrast mechanisms for MR imaging. Intermolecular MQC theories of Warren et al suggest some very intriguing properties of iMQC useful for medical imaging. They include a contrast based on dipolar interaction correlation distances, and potentials for more sensitive imaging of magnetic susceptibility distributions, which are essential for tumor oxygenation and brain functional studies. Images based on these novel contrast mechanisms may provide improved detection of hypoxic tumors. But to date the success in implementing iMQC imaging is quite limited. Warren et al first proposed intermolecular zero-quantum coherence (iZQC) imaging, which is insensitive to the magnetic field inhomogeneity and was believed to possess a relatively higher signal-to-noise ratio (SNR) than other iMQCs. They have obtained iZQC images of rat brains with varying contrast revealing structural features not seen in conventional MR images (Warren 1998; Rizi 1999). However, so far no results for human iMQC imaging have appeared in publications, and only iZQC was used in the imaging of animals. It is our opinion that this was partially due to some confusion in the theoretical basis about iMQC imaging formation. Meaningful signals derived from the iDQCs were believed to be impossible, utilizing the prototype sequence , with a long detection time t2 and a short evolution t1, which are the preferred conditions for imaging. Based on the derivation of image locations for conventional multiple echoes, van Zijl and coworkers attempted to form an image from the second spin echo after iDQC selection, but found that the image had a very low SNR and no detectable contrast even at the high field strength of 4.7 T. Recently, Bifone et al showed that MSE spectroscopic signals in a localized volume could be observed in vivo with a 1.5 T clinical MR scanner. However, the sensitivity of the detected signal was determined to be too low for MR Imaging acquisition.

In the past year or so, we have performed a series of preliminary studies on iMQC imaging, and have established an important basis and understanding for further development of this potentially very useful technique. We utilized a combination of quantum mechanical and classical formalisms to describe the behavior of the evolution of nuclear spins under long-range dipolar interactions, in the presence of relaxation and molecular diffusion. Theoretical analysis was applied to aid in the design of iDQC imaging sequences with conventional or echo planar imaging (EPI) acquisitions. We are the first research group to obtain iDQC images of the human brains using a whole-body 1.5 T scanner. We also have made preliminary quantitative comparison of iMQCs of different orders at 1.5T and higher field strengths, and have begun to explore applications of iMQC images in animal models and normal human volunteers. A series of publications have resulted from these studies.

[chapter top] [return]

3.  Some important results from our preliminary works in iMQC

1. Characterization of iMQC signals using both quantum mechanical and classical treatments
2. Demonstration of the feasibility of human iDQC imaging at 1.5T
3. Theoretical predictions and experimental confirmation of relative sensitivity of iDQC compared to SQC, and compared to iZQC
4. Theoretical predictions and experimental confirmation of field dependence of iMQC signals.
5. Parameter optimization for iDQC image acquisition.
6. Contrast characteristics of iMQC imaging.
7. Preliminary experiments at high field in tumor models and with higher order iMQCs.
8. Brain activation in human volunteers with iMQC and BOLD imaging at 1.5T
9. iMQC diffusion-weighted imaging of animals at high fields and in human volunteers at 1.5T

A.  Characterization of iMQC signals using both quantum mechanical and classical treatments 

We have studied intermolecular dipolar effects in highly polarized spin systems such as water. Expressions for iMQCs under the influence of magnetic field gradients of varying magnitudes, directions, and durations were explicitly derived with the density matrix formalism and demagnetizing field theory. We demonstrated that the time-averaged, not instantaneous, orientation of the applied gradients determines the contributions of long-range intermolecular dipole effects to multiple-quantum coherences. The time-averaged gradients served as a general model for gradients used in iMQC selection in imaging sequences. Theoretical and experimental results demonstrate, for the first time, that when the time-averaged orientation of a series of gradient pulses during the evolution period is set to the magic angle, intermolecular dipolar effects are suppressed. The experimental evidence presented strongly supports our theoretical predictions, and provides a basis for coherence selection gradient designs for iMQC imaging.

With the apparent diffusion rates for iMQC diffusion, , defined as the slope of the iMQC signal intensity vs. diffusion weighting factor, we have found that of intermolecular MQCs are different from that of the intra-molecular MQCs; Apparent diffusion coefficient for iMQCs follow the relationship: , where is the translational molecular diffusion coefficient. These results do not coincide with as previously suggested, or for the simple classical demagnetizing field model.

We have designed pulse sequences which allow the detection of signal decay solely dependent upon either iMQC relaxation or diffusion during the evolution period, and pulse sequences with selective excitation of specific nuclear spins in a heteronuclear spin system. They allow us to probe iMQCs during different time periods and will potentially provide answers to fundamental questions concerning iMQCs.

[topic 3] [chapter top]

B.  Demonstration of the feasibility of human iDQC imaging at 1.5T

Implementation of iDQC imaging with conventional SE and EPI acquisition schemes was achieved at 1.5T. Images in phantoms and in the human brain and leg demonstrate that the imaging sequences have excellent selection of the signal derived from iDQCs. These images demonstrate contrast of various regions different from that seen in conventional images. When the pulse sequence was implemented with EPI acquisition, whole brain DQC images with signal-to-noise ratios comparable to conventional EPI can be obtained in less than 60 seconds.

Fig. 3. Images of human leg. iDQC EPI image of calf muscle with iDQC-encode gradients along the B0 direction (a), along the magic angle (b) and perpendicular to B0 (c) are shown. In (d) DQC EPI image of fat and bone marrow with iDQC-encode gradients along the B0 direction is shown; (e) is T2-weighted conventional SQC image. For iDQC images, TR = 2500 ms, TE = 66 ms, matrix size is 64x64 and 128 NEX was used. These images unequivocally demonstrate that iDQC images can also be obtained in muscle and subcutaneous fat.

C.  Theoretical predictions and experimental confirmation of relative sensitivity of iDQC compared to SQC, and compared to iZQC

It was found that in general the intensity of iMQC of different orders signals are only a few percent of the intensity of the SQC, but iDQCs have a sensitivity about 30% higher than iZQC for human imaging. This was confirmed with experiments in uniform phantoms and in human brain tissue at 1.5T. The duration of iMQC signals is found to be much greater than that of SQCs with the same T2. Also the maximum of the signal is centered at a later time.

D.  Theoretical predictions and experimental confirmation of field dependence of iMQC signals.

It is demonstrated that both the iZQC and iDQC signal intensities increase with increased static field strengths. However, the sensitivity of the iDQC signal is always greater than that of the iZQC signal at varying static magnetic field.

E.  Parameter optimization for iDQC image acquisition.

We incorporated several measures in the imaging pulse sequence design to achieve optimal sensitivity for iDQC signal detection: (a) The b = pulse was replaced with a pulse, for n = -2 quantum transition, in the standard CRAZED sequence. The maximum signal derived from iDQCs is increased by a factor of . (b) The position of the acquisition window was adjusted based on the property of the iMQC signals. Our theoretical analysis suggests that the maximum signal intensity from tissues with varying occurs at different times, and have a variable duration. Therefore a large acquisition window (small bandwidth) is used to sample a broad range of time-domain signals. (c) Receiver dynamic range was optimized. In our implementation of iDQC imaging at 1.5T, the default coil configuration file for the GE quadrature head coil was modified to increase the reconstruction scale 18dB during the acquisition of iDQC images. (d) A two-step phase cycle scheme for iDQC-encode gradients was designed to remove additional undesired coherence pathways. A pair of gradients of small amplitude, was oriented along the y-direction to eliminate the residual contamination from coherences other than iDQCs.

F.  Contrast characteristics of iMQC imaging.

When the signal intensity from iDQC image was measured as a function of TE in the human brain, a maximum occur when TE ~ T2. The signal increases initially, peaks and then declines with increasing TE. There is good agreement between detected signal changes and our theoretical predictions for gray matter (GM) and white matter (WM) of the brain.

 

Fig.7. Brain iDQC images acquired with the EPI sequence with varying TE values (in ms as labeled). The iDQC evolution time was 9.5 ms. TR = 5000 ms, matrix size was 64x64 and 16 NEX was used for all images. The image time for each TE was 1:20. Images are displayed with the same window setting. The initial increase, peaking and then decline of the signal intensity with TE are apparent, particularly in the cerebrospinal fluid (CSF). This is quite different from conventional SQC imaging where signal intensity decreases monotonically with increasing TE.

 

When the iMQC signals are detected with varying evolution time t, the signal attenuation depends on , which is a sum of the contributions of the intermolecular MQC and the magnetic field inhomogeneities. The signal variation with t may provide a more sensitive means to detect magnetic susceptibility variation caused by oxygenation or brain activation.

[topic 3] [chapter top]

G.  Preliminary experiments at high field in tumor models and with higher order iMQCs

iMQC images of tap water were also acquired with the selected coherence, n =1 to 5. An optimal TE was determined for each order. Their relative SNR compared to SQC was found to be 0.25(n=2), 0.12(n =3), 0.06(n=4), and 0.03(n=5), which agree well with our predictions from theoretical analysis. An increase of iMQC sensitivity with increase in field strength is demonstrated.

H.  Brain activation in human volunteers with iMQC and BOLD imaging at 1.5T

Previous works by Warren et al have shown that iZQC is insensitive to the inhomogeneity of local fields. The iDQCs, on the other hand, are highly sensitive and the apparent diffusion rates for iDQCs are twice that of SQCs. The enhanced diffusion effect may amplify the spin dephasing in the local field due to susceptibility distribution, and further enhance the BOLD signal detected in brain activation studies. Equivalently the dephasing of the signal at twice field strength is detected during the evolution period, while the signal maintains the relaxation rates of the actual measurement field during detection. In BOLD fMRI, a T2*-weighted image will show neuronal activation, through the secondary and tertiary effects of blood oxygenation and blood flow respectively. The iMQCs themselves may be directly related to susceptibility variation and may therefore be more sensitive to changes caused by neuronal activation than BOLD contrast. For example, iDQC evolves at the sum of the SQC frequencies of the two spins involved. Therefore the iDQC signal intensity varies as a function of the distribution of susceptibility gradients. The BOLD signal, on the other hand, is a function of the average strength of those gradients. Furthermore, the choice of correlation distance provides an additional degree of freedom to optimize contrast, which BOLD-based methods do not possess. Potential gains in selectivity and sensitivity may be obtained in brain functional imaging as well with iMQC imaging. Non-specificity to the vessel size has been an intrinsic drawback for BOLD-based methods, since they do not have accurate spatial correspondence with the actual sites of enhanced neuronal activity. iMQCs offer the possibility of altering the distance scale over which the dipolar interactions lead to an observable signal, and hence the sensitivity to blood vessels of different sizes. In addition, iMQC imaging sequences based on the CRAZED sequences allow for the separation of the evolution of the iMQCs which is sensitive to the neuronal activation, and the acquisition period for which the most effective schemes can be used to enhance SNR. Since the iDQC signal lasts much longer than the SQC or , the iDQCs provide the possibility to used multi-echo and other effective acquisitions. In conventional BOLD activation studies, a compromise has to be made between the requirement of the best activation contrast-to-noise ratio and SNR enhancement for acquisition. It has been demonstrated that at high magnetic fields there is improved contrast for BOLD imaging, and more importantly, an elevated sensitivity to capillary level changes coupled to neuronal activity. This is achieved via stronger functional dependence of R2* on field strength in capillary networks. We believe that information acquired from iDQC imaging may be complimentary to BOLD imaging, and the advantages of activation studies at higher fields will be fully utilized with a combined iMQC and BOLD technique.

Preliminary studies have been performed with iDQC imaging for brain activation at 1.5T. An auditory stimulation was used in which the subject was listening to the same piece of country music during activation period. It demonstrated that it is feasible to obtain activation maps with iDQC imaging even at a relatively low field of 1.5T, when the imaging parameters are nearly optimized. When a SE acquisition is used with an iDQC excitation, activation is mostly due to effect of susceptibility changes during the iDQC evolution period.

Fig 9. Auditory activation by listening to a piece of country music. (Left) BOLD; (middle) iDQC with spin-echo acquisition; (right) iDQC with gradient-echo acquisition. Activated pixels are superimposed onto average images of corresponding types of scans in gray-scale.

[topic 3] [chapter top]

I.  iMQC diffusion-weighted imaging of animals at high fields and in human volunteers at 1.5T

Since the intermolecular MQC has an increased sensitivity to molecular diffusion compared to the single-quantum diffusion process, it may provide a more sensitive probe for the molecular diffusion process, and be very useful in diffusion weighted MRI. Previously, we have studied properties of iMQC diffusion via a spectroscopic measurement. A RF sequence was designed so that the confounding effects of radiation damping, relaxation, and varying dipolar correlation distances were all eliminated, and the signal decay solely with increasing diffusion weighting was measured. Similar design principles can be applied to the diffusion imaging sequences.