Magnetic resonance imaging with T₁ dispersion contrast

Abstract

Prepolarized MRI uses pulsed magnetic fields to produce MR images by polarizing the sample at one field strength (∼0.5 T) before imaging at a much lower field (∼50 mT). Contrast reflecting the T₁ of the sample at an intermediate field strength is achieved by polarizing the sample and then allowing the magnetization to decay at a chosen “evolution” field before imaging. For tissues whose T₁ varies with field strength (T₁ dispersion), the difference between two images collected with different evolution fields yields an image with contrast reflecting the slope of the T₁ dispersion curve between those fields. Tissues with high protein content, such as muscle, exhibit rapid changes in their T₁ dispersion curves at 49 and 65 mT due to cross-relaxation with nitrogen nuclei in protein backbones. Tissues without protein, such as fat, have fairly constant T₁ over this range; subtracting images with two different evolution fields eliminates signal from flat T₁ dispersion species. T₁ dispersion protein-content images of the human wrist and foot are presented, showing clear differentiation between muscle and fat. This technique may prove useful for delineating regions of muscle tissue in the extremities of patients with diseases affecting muscle viability, such as diabetic neuropathy, and for visualizing the protein content of tissues in vivo. Magn Reson Med 2006. © 2006 Wiley-Liss, Inc.

The ability to manipulate image contrast is one of MRI's strongest advantages over other imaging modalities. The search for new contrast mechanisms and better visualization of various tissues and pathologies drives a substantial segment of MRI research. Cooperative research efforts between the world of medicine and that of physics and engineering have sought and found an ever-increasing number of endogenous tissue parameters that can produce useful contrast for specific medical applications, including T₁ relaxation, T₂ relaxation, diffusion, perfusion, blood oxygenation, and others. T₁ dispersion contrast is a new contrast mechanism that uses the variation in T₁ with magnetic field strength to probe the behavior of macromolecules in tissue, similar to T_1ρ imaging (1) or magnetization transfer contrast (MTC) (2, 3).

The values of T₁ and T₂ for a given tissue depend on many factors; some are intrinsic, such as the molecular makeup of the tissue, and some are extrinsic, relating to the local cellular environment of the tissue. In particular, the T₁ of many tissues depends strongly on the local magnetic field strength. T₁ values usually get longer as the field strength increases, but the rate of change in T₁, known as T₁ dispersion (4), varies from tissue to tissue. T₁ dispersion is therefore a potential contrast mechanism that may yield different information than standard T₁ contrast; rather than comparing the relative values of T₁ at a particular field strength, T₁ dispersion contrast would reveal how those values of T₁ change as the field increases.

To access T₁ dispersion as a contrast mechanism, we must be able to change the field strength while the image is being acquired. This requires a pulsed-field MRI scanner using a technique such as prepolarized MRI (PMRI) (5-7) or field-cycling MRI (8, 9). If we use a pulsed-field scanner to collect two images with T₁-weighted contrast at different field strengths (normalized to account for the different equilibrium magnetizations) and then subtract those two images, the resulting image will have contrast that comes from the difference in T₁ between those two field strengths.

Tissues whose T₁ did not change between the two fields (a flat T₁ dispersion curve) will have the same intensity in both images, and their signal will subtract away. Tissues whose T₁ has changed substantially between the two fields will have different intensities in the two images, and their intensity in the subtracted image will be indicative of the slope of their T₁ dispersion curve.

The T₁ dispersion curves of protein-rich samples have a well-known feature known as the “nitrogen dip,” a rapid change in T₁ that causes a steep T₁ slope over a small field range (10-15). At field strengths around 50 and 65 mT, protons in tissue can undergo cross-relaxation with nitrogen nuclei in the backbones of proteins, shortening their T₁ and leading to “dips” in the T₁ dispersion curve. These dips are only present in the T₁ dispersion curves of tissues that have a substantial amount of protein; tissues without protein, such as fat, will have a flat T₁ dispersion curve over this field range (16). Therefore, an image with T₁ dispersion contrast using the field range around the nitrogen dip will display tissues with an intensity corresponding to their protein content (9, 17, 18). T₁ relaxation data from in vitro muscle and fat samples is shown in Fig. 1; the physical basis for these dips will be discussed further under Theory.

Figure 1

T₁ measurements performed using our prepolarized MRI scanner on samples of muscle tissue and fat tissue (both from chicken), showing the characteristic nitrogen cross-relaxation dips in the T₁ dispersion curve of the protein-rich muscle tissue.

The T₁ dispersion imaging technique is accessing a similar cross-relaxation mechanism as that used in MTC (2, 3). In MTC, off-resonance RF irradiation is used to saturate the pool of protons associated with macromolecules, which then undergo saturation transfer with the pool of liquid, detectable protons. T₁ dispersion instead takes advantage of the intrinsic cross-relaxation phenomenon that occurs at the nitrogen quadrupolar resonance field strength, requiring no additional RF excitation. In MTC, some direct saturation of the liquid pool of spins inevitably occurs due to the applied RF pulse; with T₁ dispersion, there is no such “direct” relaxation effect in addition to the indirect cross-relaxation. Therefore, T₁ dispersion provides a similar form of contrast from the same cross-relaxation effect, but without the additional direct saturation effect that must be accounted for in MTC. As such, T₁ dispersion may be useful for many of the same applications for which MTC has been used, including imaging of diabetic neuropathy (19), musculoskeletal tumors (20), and multiple sclerosis lesions (21, 22), among others. This T₁ dispersion imaging technique is also similar to T_1ρ imaging in that it benefits from the SNR of a higher polarizing field while producing images with the contrast of a lower-field relaxation time (1). However, T₁ dispersion imaging is able to access a much higher range of field strengths than T_1ρ imaging, as it is not limited by the application of a spin-locking RF field. RF power absorption limitations would prevent T_1ρ imaging at the fields where nitrogen cross-relaxation occurs.

THEORY

Nitrogen Cross-Relaxation

Nitrogen atoms in the backbones of proteins can act as relaxation sinks for surrounding water protons via cross-relaxation. As a spin-1 nucleus, nitrogen has both a dipole moment and a quadrupole moment; the quadrupole moment interacts with fluctuating electric field gradients in a comparable way to the interaction of the dipole moment with magnetic fields. The fluctuating electric fields caused by motion of electrons in molecules induce relaxation much more efficiently than the fluctuating magnetic fields caused by nuclear magnetic moments. As a result, quadrupolar nuclear relaxation of nitrogen nuclei in a protein is on the order of 10⁻⁶ s (23), which is essentially instantaneous in comparison to MRI timescales.

The necessary conditions for cross-relaxation between a proton and a nitrogen nucleus are a strong spin coupling and a match between the proton and nitrogen nuclear energy levels. Strong dipolar coupling is present within NH pairs in the backbone of a protein, where the separation distance is only about 1 Å (24). The energy level matching can occur only at low magnetic field strengths, where the energy levels of the quadrupolar ¹⁴N nucleus are dominated by local static electric field gradients rather than the presence of a magnetic field. When the Larmor frequency is much less than the quadrupolar splitting, the three energy levels of the nitrogen nucleus (corresponding to m = 0 and linear combinations of m = ±1) remain fixed regardless of the field strength; transitions among these three energy levels for nitrogen nuclei in proteins correspond to frequencies of roughly 0.75, 2.15, and 2.82 MHz (23). When the Larmor frequency of the protons equals one of these quadrupolar transition frequencies, cross-relaxation can occur rapidly within the NH pair.

The NH proton couples to other nearby protons in rapidly exchanging water molecules and can undergo a comparable relaxation interaction with a water proton, either via dipolar coupling or via chemical exchange (3). In this interaction, the NH proton relaxes the proton of the exchangeable water and absorbs its energy and then transfers that energy to the nitrogen nucleus, effectively acting as an energy conduit to transfer energy from water protons to the lattice. A single nitrogen nucleus can therefore act as a relaxation sink for multiple rapidly exchanging water protons in the surrounding tissue. This cross-relaxation speeds up the T₁ process of the water protons at the frequencies where nitrogen cross-relaxation can occur, resulting in a dip in the T₁ dispersion curve at those frequencies (or a “peak” in the R₁ = 1/T₁ dispersion curve).

Tissues that contain protein will exhibit nitrogen dips in their T₁ dispersion curves. Nitrogen dips have been measured in muscle tissue (5, 15, 23, 24), collagen fibers (25), brain tissue (5, 26), multiple sclerosis plaques (15), and eye lenses (23), among others. The depth of the dip in the T₁ dispersion curve corresponds to the protein content of the sample (14) and has been used to quantitatively measure the protein concentration of muscle tissue in vitro with a nonimaging bulk measurement (27). Tissues without protein, such as fat, have a nearly flat dispersion curve in the field range of the nitrogen dips.

Magnetization Manipulation for T₁ Dispersion Contrast

To produce T₁ dispersion contrast images, two images with T₁ contrast from different field strengths must be subtracted, producing an image whose contrast comes from the difference in T₁ between those field strengths. For protein-content contrast, the two fields we choose for maximizing the difference in T₁ (or maximizing the slope of the T₁ dispersion curve) are at the “top” and “bottom” of the nitrogen dip at 65 mT (indicated in Fig. 1). Tissues with a substantial protein content will have a large change in their T₁ between those two fields, leading to a large signal difference, while tissues without protein will have little or no change in their T₁ over that field range and will subtract away.

To produce images with pure T₁ contrast at the two field strengths of interest, we must collect two images at each field strength, meaning that a total of four images are collected to produce the final image with T₁ dispersion contrast. The reason for this can be seen from an examination of the magnetization behavior, shown qualitatively in Fig. 2, as it evolves in the changing fields during magnet pulse sequence. First, the polarizing field is switched on for a time on the order of 3T₁, during which the magnetization approaches a saturation value M_p. The polarizing field is then switched off, and the evolution field is switched on to one of the two evolution field strengths for a time t_e, during which the magnetization decays toward an equilibrium value M_e. After the polarization and evolution intervals, the magnetization M can be written as

(1)

where M_p is the sample magnetization after the polarization period, M_e is the equilibrium magnetization at the evolution field strength, t_e is the evolution time, and T is the relaxation time of the sample at the evolution field, which is the desired source of contrast.

Figure 2

Qualitative behavior of the sample magnetization during polarization and evolution for two different evolution field strengths. The sample approaches the polarization magnetization M_p during the polarization interval and then decays toward the high evolution field (solid line) or the low evolution field (dashed line). If an RF inversion pulse is applied to flip the magnetization into the negative plane after the polarization interval (dotted line) the signal after inversion can be subtracted from the signal without inversion to remove the dependence of the final magnetization on M_e.

An image collected after this polarization-evolution sequence would have contrast that came from both T and M_e, as well as some dependence on the initial polarization magnetization M_p. If we attempted to create a T₁ dispersion contrast image by taking two such images with different evolution field strengths and subtracting them directly, the resulting contrast would depend on the difference between the two values of M_e as well as the difference between the two values of T. True T₁ dispersion contrast should depend only on changes in T₁ as the field strength changes, not on changes in the field strength itself.

To remove this dependence on M_e, we collect another image with a 180 ° inversion after the initial polarization. The inversion pulse flips the magnetization into the negative plane, where it then evolves as before. The resulting magnetization after an evolution time t_e is then

(2)

As the first term in this expression is now negative, while the second term remains positive, we see that a complex subtraction of Eq. [2] from Eq. [1] will eliminate the second term entirely, as it is identical in both cases. The resulting magnetization will be

(3)

which depends only on the initial polarizing magnetization M_p and the T₁ at the evolution field, T. Thus, subtracting images with and without an inversion will yield an image whose contrast does not depend on the strength of the evolution field directly, but only on the T₁ at that evolution field.

Therefore, if two such images are acquired using different evolution field strengths, with correspondingly different values of T subtracting these two images should produce an image whose contrast reflects the difference in T between the two field strengths. Since both images will undergo the same initial polarization, they will have the same M_p. Supposing a “high” evolution field (H) and a “low” evolution field (L) are used, the magnetization of such a subtraction can be written as

(4)

where T and T are the T₁ values at the two evolution fields. An image taken after this type of sample magnetization has been achieved will have the desired T₁ dispersion contrast.

Partial Volume Effects

To examine partial volume effects with this technique, we look at Eq. [4] for the case of a voxel containing multiple tissues with different T₁ values. Suppose a voxel contains both muscle and fat, with relaxation values T and T. Equation [4] would then have the form

(5)

where M_m and M_f are the magnetizations of muscle and fat after polarization. However, since fat has a flat dispersion curve, T = T, and the two terms relating to the magnetization of the fat portion will subtract out. The magnetization left after subtraction will be

(6)

the same as if the voxel had contained only muscle tissue with a single T₁ value. However, since the volume of muscle tissue in the voxel is smaller than it would be if the voxel were completely filled with muscle, the magnitude of the polarization magnetization M_m will be reduced, resulting in a lower intensity pixel. Partial volume voxels will therefore appear as intermediate intensity pixels in the final image, accurately representing the amount of protein in the voxel.

MATERIALS AND METHODS

All of the data and images shown here were taken using a homebuilt prepolarized MRI scanner (see Fig. 3). The system consists of two pulsed electromagnets: a 0.4 T polarizing magnet (28), which is pulsed on to create the sample polarization, and a homogeneous readout magnet (29-31), which is pulsed on to 50 mT for signal acquisition (5, 6). This prepolarizing scheme can produce images with the theoretical SNR of the polarizing field, while retaining the advantages of a low-field readout. The signal strength comes from the prepolarization pulse, and provided that the readout field is high enough to achieve body noise dominance (the sample being the dominant noise source rather than the coil), the image SNR depends only on the strength of the polarizing field and is independent of the readout field strength (32). Our PMRI system uses a Tecmag console; RF coils and magnet control electronics are homebuilt (33).

Figure 3

Cross-sectional schematic of our prepolarized MRI scanner.

To produce T₁ dispersion images, the magnet pulse sequence shown in Fig. 4 was used. To maximize the signal of such an image, we must choose imaging parameters to maximize Eq. [4]. Since the magnetization depends directly on the initial polarization magnetization, the polarization time is chosen such that the sample magnetization is at least 90% of its equilibrium value. For protein-content contrast, we choose the two evolution field strengths to be at the top and bottom of the higher nitrogen dip (65 and 74 mT), as indicated in Fig. 1. The images taken at each evolution field strength are interleaved to reduce motion artifacts in the final subtraction images.

Figure 4

Magnet pulse sequence used to acquire T₁ dispersion images. An initial polarization pulse provides the sample polarization. After the polarizing magnet is turned off, an RF inversion pulse is applied on every other interleave before the readout magnet is ramped up to one of the two evolution fields. After the evolution interval, the readout magnet is ramped back down to 2.2 MHz and a 2DFT spin-echo imaging sequence is applied. The gradients applied during the spin-echo imaging sequence are not shown.

We then chose the evolution time to maximize the difference of the two exponential terms in Eq. [4]. Using T₁ measurements for muscle from the data shown in Fig. 1, we calculated that the maximum signal difference occurs at an evolution time of 167 ms. The peak of the signal difference curve is broad enough that a t_e of even 115 ms will yield 90% of the available signal strength, so the evolution time need not be set precisely. However, there is no advantage to using longer values of t_e; using an evolution time longer than the peak value will only result in less signal as the magnetizations decaying in the high and low evolution field approach each other.

In Vitro Experiments

We performed initial in vitro T₁ measurements and tests of the T₁ dispersion imaging technique using tissue samples of chicken muscle and fat, as well as water doped with copper sulfate to have a T₁ of ∼250 ms at 1 MHz. The sample tubes for imaging experiments were 1 cm in diameter and 3 cm in length.

To demonstrate the effects of partial-volume voxels on the T₁ dispersion technique, we constructed a phantom composed of five sample tubes filled with varying proportions of muscle tissue and fat (100, 75, 50, 25, and 0% muscle, with the remaining volume filled with vegetable oil). Muscle tissue was measured by weight such that the tubes contained 2.0, 1.5, 1.0, 0.5, and 0 g of muscle; oil was then added in inverse proportions using 0, 0.5, 1.0, 1.5, and 2.0 mL of oil. This phantom was imaged using a 2D projection, so that the full 2-cm length of the tube comprised a single “voxel,” and with a coarse 16 × 16 matrix so that the 1-cm diameter of each tube was spread over only four pixels in the image. This is a macroscopic way to explore partial volume effects, as each tube containing both muscle and fat is treated as only a few voxels.

We also performed in vitro experiments on protein gels of varying concentrations. A solution of 30–40% bovine serum albumin (BSA) (Sigma A-2934) was prepared in deionized H₂O and then diluted to 5 mL in the appropriate amounts for final concentrations of 10–20% BSA in H₂O for fixation in a 10-mL volume. The solutions were placed in 15-mL screwcap polypropylene tubes (2 cm in diameter and 6 cm in length) and chilled in an ice bucket. Then 5 mL of ice-cold 50% glutaraldehyde (photographic grade, Sigma G-6403) was added to the BSA solution, and the tube was immediately inverted six times and placed upright in an ice bucket for 10 min. The tubes could then be removed from the ice for imaging at room temperature (34).

To perform T₁ measurements on in vitro samples, a 1-s polarization interval was followed by a variable evolution time (25–1000 ms) at the field strength for which the T₁ was being measured. The magnitudes of spin echoes collected after the evolution time were fit to a curve describing the magnetization decay after polarization to extract the T₁ at that field strength.

In Vivo Experiments

Imaging studies were performed on two normal volunteers, using the pulse sequence shown in Fig. 4, with a 1-s polarizing time at 0.4 T and a 160-ms evolution time. We allow 48 ms for the polarizing magnet to ramp down before applying the first RF pulse, to allow any ringing in the readout magnet to dissipate. Readout was performed at 2.2 MHz with an 8-cm FOV and 1-cm slice thickness.

Processing of in vivo images was done in two ways: a “proportional” method and a “high contrast” method. For the proportional method, the high evolution and low evolution images are directly subtracted; a 2 × 2 median filter is then applied to remove noise introduced by the subtraction. This yields a T₁ dispersion image whose intensity depends on the amount of protein in a voxel as given by the magnetization expression in Eq. [4].

For the high contrast method, a masking procedure was applied using cluster analysis to separate the pixels in the high and low evolution images into two groups: pixels whose intensity was the same in both images and pixels whose intensity changed significantly between the two images (35). Pixels whose intensity remained the same in both images were then set to zero in the original high evolution image, leaving a “masked” T₁ dispersion image that shows regions of high T₁ dispersion tissue, but whose intensity retains the SNR of the original high evolution image and is not proportional to protein content. The masking threshold was calculated from the ratio of the high evolution and low evolution magnetizations as given in Eq. [3], using T₁ values for muscle at 74 and 65 mT taken from the data in Fig. 1. We calculated a maximum signal ratio of 1.23; pixels whose intensity is the same in both images would have a signal ratio of 1.00. The masking threshold was therefore selected to be halfway between these two values, or 1.115. Pixels whose intensity in the high evolution image is less than 1.115 times their intensity in the low evolution image are set to zero, and a 2 × 2 median filter is applied. This high contrast masking technique shows good delineation of regions of protein-rich muscle tissue, but does not provide additional information about the concentration of protein in the muscle beyond the fact that it is above a certain threshold.

RESULTS

In Vitro Experiments

We tested our T₁ dispersion imaging sequence on a phantom of test tubes containing muscle tissue, fat tissue, and water doped with copper sulfate. The resulting projection image is shown in Fig. 5. All three tubes are clearly visible in the images taken at each evolution field, shown in Fig. 5a and b. Subtracting the low evolution image (Fig. 5b) from the high evolution image (Fig. 5a) yields the proportional T₁ dispersion image in Fig. 5c, in which only the muscle tissue is visible. This tissue phantom test confirms that our technique produces protein-content contrast.

Figure 5

T₁ dispersion projection image of 1-cm-diameter tubes containing muscle tissue, fat tissue, and doped water. In the subtracted T₁ dispersion image, the muscle tissue remains bright, while the signal from fat and unbound water has been subtracted away. Image parameters include the following: polarization interval = 1 s, t_e = 160 ms, FOV = 4 × 4 cm, matrix = 64 × 64, BW = ±4 kHz, scan time 5:45.

To demonstrate the effects of partial-volume voxels on the T₁ dispersion technique, we imaged a phantom composed of five sample tubes filled with varying proportions of muscle tissue and fat, using a coarse image resolution to ensure that each voxel contained both muscle and fat. The results are shown in Fig. 6. The proportional T₁ dispersion image shows that the intensity of each tube increases as the proportion of muscle to fat increases. A relative SNR measurement, done by comparing a 4 × 4 pixel region around each sample and normalizing to the SNR of the 100% muscle sample, shows this behavior quantitatively; the relative SNR values are given in Table 1. Correlation between the percentage of muscle tissue and the relative SNR is good, given some imprecision in measuring the exact quantity of muscle tissue in each tube. The normalized SNR for the mixed-voxel samples is higher than the actual muscle percentage because the fat present in each sample does not subtract away completely to zero, as shown by the 100% fat sample, which still retained enough signal to yield a normalized SNR of 0.06.

Figure 6

T₁ dispersion image of 1-cm-diameter tubes containing different proportions of muscle tissue and fat (vegetable oil) to mimic the effect of partial volumes on the T₁ dispersion technique. A projection image was taken with very coarse resolution (0.25-cm resolution, 4-cm FOV) to effectively treat each sample as only a few voxels with mixed fat–muscle content. The relative SNR of each sample in the image (see Table 1) shows that the intensity of a voxel is proportional only to the amount of muscle in the voxel. Image parameters are as follows: matrix = 16 × 16, NEX = 4, scan time 7:30, other parameters are identical to those in Fig. 5.

Table 1. Correlation of SNR with Muscle Proportion

Muscle percentage	Normalized SNR
100	1
75	0.88
50	0.52
25	0.28
0	0.06

• The normalized SNR of a set of samples with varying proportions of muscle tissue and fat. Relative SNR measurements of the entire region of the sample tubes were taken using the subtraction image in Fig. 6 and normalized to the SNR of the 100% muscle sample. The relative SNR shows good correlation with the amount of muscle tissue in each sample; the normalized SNR is higher than the actual muscle percentage because the fat present in each sample does not subtract completely to zero.

As an additional test of how the T₁ dispersion technique depends on protein concentration, we measured the relaxation rate of protein gel samples; results from two samples (8 and 12% protein) are shown in Fig. 7 in terms of the relaxation rate R₁; the shortening of T₁ due to nitrogen cross-relaxation appears as peaks on the R₁ relaxation rate curve. The effect of the nitrogen cross-relaxation can be isolated by subtracting off the overall R₁ dispersion rate. The data outside the region of the nitrogen effect were fit to a function of the form R₁ = A ω^−b (36, 37); the theoretical fits are shown in Fig. 7a as dotted curves. Subtracting these curves from the data yields the isolated effect of the nitrogen cross-relaxation, shown in Fig. 7b.

Figure 7

(a) Relaxation rate measurements performed using our prepolarized MRI scanner on two protein gel samples of different protein concentrations, showing the increase in size of the nitrogen peaks with higher protein concentration. (b) The overall R₁ dispersion curves of the proteins (dotted lines) have been subtracted from the data to isolate the effect of the nitrogen cross-relaxation on the relaxation rate.

An image of the protein gel samples is shown in Fig. 8. In the high evolution image (Fig. 8a), the protein gels have much lower signal than the water and fat samples. In the subtraction image (Fig. 8c), the water and fat samples have subtracted out, but the 12% gel is of only comparable intensity to the 8% gel, despite having a larger change in T₁ due to the nitrogen effect as demonstrated in Fig. 7b. This is because the 12% gel, while having a larger nitrogen peak in its R₁ curve, also has a higher overall relaxation rate, seen in Fig. 7a. The faster relaxation rate causes it to lose more signal during the evolution time, resulting in lower overall signal in the final image. However, if the subtraction image is divided by the sum of the original evolution images to remove the overall magnetization and show a normalized “percentage change” image, we see in Fig. 7d that the 12% gel is indeed significantly brighter than the 8% gel. This division amplifies the image noise, so in Fig. 7d the regions of noise (determined by the high evolution image) have been masked out. The 8% sample has a relative SNR of 0.68 when normalized to the SNR of the 12% sample, confirming that the image intensity scales with protein concentration. In Vivo Experiments

Figure 8

T₁ dispersion image of 2-cm-diameter tubes containing different percentages of protein gels to mimic the effect of varying protein concentrations on the T₁ dispersion technique. The relative SNR of the 8% protein gel sample is 0.68 when normalized to the SNR of the 12% sample, showing that the intensity of a voxel scales with the protein concentration. Image parameters are as follows: polarization interval = 1 s, t_e = 100 ms, FOV = 6 × 6 cm, matrix = 64 × 64, slice thickness = 2 cm, BW = ±4 kHz, scan time 5:45.

We applied the T₁ dispersion technique to multiple normal volunteers by imaging a cross-section of the wrist. Representative results are shown in Fig. 9. The high and low evolution images are shown above, with the T₁ dispersion images processed with subtraction and with masking shown below. Regions of fat that are bright in the evolution images, including bone marrow and pockets of subepidermal fat, subtract out in the T₁ dispersion images; regions of muscle tissue remain bright. The high contrast masked image shows clear delineation of the muscle regions, while the proportional subtraction image has lower SNR but shows variations in intensity within the muscle that may provide additional information about the muscle protein content.

Figure 9

Coronal T₁ dispersion image of the forearm of a normal volunteer. (a) and (b) show the high and low evolution images, respectively. (c) The direct subtraction of the two evolution images results in a protein-content contrast image whose intensity depends on the amount of protein in each voxel. (d) A “masked ” version of the high evolution image in (a), in which pixels whose intensity did not vary between (a) and (b) have been masked out, shows regions of protein-rich tissue, but the image intensity does not depend on the quantity of protein in the voxel. Image parameters are as follows: polarization interval = 1 s, t_e = 160 ms, FOV = 8 × 8 cm, matrix = 128 × 64, slice thickness = 1 cm, BW = ±8 kHz, scan time 5:45.

We have also imaged the foot of a normal volunteer. A slice through the ball of the foot near the metatarsal heads is shown in Fig. 10. There is some signal loss at the edges of the image due to the limited size of the RF coil; the images were taken using a “clamshell”-shape coil, 3 inch in width and 2 3/8 inch in height, while the subject's foot was 3 ¼ inch wide and extended just beyond the coil's sensitivity. However, the muscular regions in the center of the foot are shown clearly in both the subtracted and the masked T₁ dispersion images.

Figure 10

Coronal T₁ dispersion image of the foot of a normal volunteer, near the metatarsal heads. (a), (b), (c), and (d) are as in Fig. 9. Image parameters are as follows: matrix = 64 × 64, BW = ±4 kHz, other parameters are identical to those in Fig. 9.

DISCUSSION

T₁ dispersion protein-content imaging has certain similarities to MTC. Both techniques use the cross-relaxation process between water protons and protons in macromolecules to produce contrast. However, unlike MTC, the T₁ dispersion technique requires no additional RF irradiation; it depends only on the cross-relaxation effect that occurs naturally at particular B₀ field strengths. The RF irradiation used for MTC inevitably produces some direct saturation of the water protons in the process of saturating the protons in the macromolecules (3). Thus, the contrast produced by magnetization transfer is actually a blend of the change in magnetization due to cross-relaxation and the change in magnetization due to direct RF irradiation. For T₁ dispersion contrast, on the other hand, no additional RF is applied, so the change in magnetization comes entirely from the cross-relaxation of water protons and protons in protein. This means that T₁ dispersion contrast is a way to visualize the strength of the cross-relaxation interaction in a more pure form, without the additional effect of direct RF saturation.

MTC can work at any frequency, because the source of relaxation for the bound protons is an applied off-resonance pulse. Nitrogen cross-relaxation can occur only at the fixed frequencies of the nitrogen quadrupolar resonance. However, the effect of the nitrogen cross-relaxation on the T₁ of the liquid protons occurs immediately once the proton Larmor frequency equals the nitrogen transition frequency. The MTC effect takes some time to develop, since the applied RF must first saturate the bound pool of protons and then transfer to the liquid pool; MTC RF saturation pulses are often hundreds of milliseconds in duration. The lack of an RF saturation pulse removes some time from the imaging sequence, although T₁ dispersion imaging does require an evolution time for the cross-relaxation process to occur. The maximum T₁ dispersion signal is achieved within 100–200 ms depending on the T₁ of the particular tissue, as is explained under Materials and Methods. The lack of RF irradiation in T₁ dispersion imaging also removes the specific absorption rate (SAR) constraints that can be limiting for MTC. Of course, it should be noted that SAR is never a problem for PMRI, since all RF takes place at the low readout field (38); there are no SAR concerns when using PMRI to do either MTC or T₁ dispersion.

Imaging time is still a drawback for T₁ dispersion imaging, however. Since four images must be acquired to produce a single T₁ dispersion image, the technique is somewhat time-intensive. Although the images presented here were acquired using a standard 2DFT imaging sequence, the required imaging time could be reduced by incorporating various fast imaging techniques that have been developed for other applications, such as RARE (39). In fact, virtually any pulse sequence can be incorporated into this technique simply by adding our preparatory polarization and evolution magnet pulses before the RF excitation of the pulse sequence. This flexibility means that the images presented here are by no means indicative of any hard limit on speed or SNR; they are simply an initial approach to a problem that may be addressed in the future with more efficient data acquisition methods for further optimization. With any sequence, the four acquired images should be interleaved to reduce motion artifacts in the subtracted images.

Many of the applications for which MTC is useful will also be interesting to pursue with T₁ dispersion imaging. MTC has been used to measure changes in the diabetic neuropathic foot (19); it has recently been shown that changes such as small muscle atrophy and fatty infiltration can be detected using ³¹P MRI to distinguish muscle from surrounding tissue before the neuropathy can be clinically detected (40). As PMRI is eminently suited for extremity imaging, the T₁ dispersion technique for distinguishing muscle from fatty tissue may prove to be a useful tool for early diagnoses of diabetic foot problems and monitoring of treatment response. Given the large and growing number of patients with diabetes, the low cost of PMRI could enable this technique to have a significant impact on the field of diabetes, both for yearly monitoring of patient progress and for testing possible therapies for diabetic neuropathy.

MTC is also used for characterization of white matter diseases such as multiple sclerosis (21). T₁ dispersion imaging may be another way to obtain contrast for identifying lesions and demyelination of white matter. Myelin has a substantial protein content, and nitrogen dips have been observed in white matter (26), so a reduction of myelin should correspond with a reduction in protein content that may be detectable with T₁ dispersion imaging.

CONCLUSIONS

We have demonstrated in vivo images with protein-content contrast using T₁ dispersion imaging. The T₁ dispersion imaging technique makes use of the flexibility of prepolarized MRI in choosing the field strength during an imaging sequence. The resulting T₁ dispersion contrast yields an image of the protein content of the sample, where tissues with little or no protein are suppressed and protein-rich tissues are bright. This technique may prove useful for monitoring diseases that affect muscle viability, such as diabetes or muscular dystrophy, as well as diseases that affect demyelination of white matter.

Acknowledgements

The authors thank additional research contributors to the Stanford PMRI project: Thomas Grafendorfer, Blaine Chronik, Hao Xu, and Patrick Morgan.