Radiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online before print June 28, 2002, 10.1148/radiol.2242011022
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
2242011022v1
224/2/577    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Du, W.
Right arrow Articles by Karczmar, G. S.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Du, W.
Right arrow Articles by Karczmar, G. S.
(Radiology 2002;224:577-585.)
© RSNA, 2002

Technical Developments

Breast MR Imaging with High Spectral and Spatial Resolutions: Preliminary Experience1

Weiliang Du, MS, Yiping P. Du, PhD, Ulrich Bick, MD, Xiaobing Fan, PhD, Peter M. MacEneaney, FFR, RCSI, Marta A. Zamora, BA, Milica Medved, PhD and Gregory S. Karczmar, PhD

1 From the Department of Radiology, MC2026, University of Chicago, 5841 S Maryland Ave, Chicago, IL 60637. Received June 11, 2001; revision requested August 3; revision received October 31; accepted December 20. Supported by National Cancer Institute grants RO1CA75476 and RO1CA78803, Army Breast Cancer Research Program grant DAMD 17-99-1-9121, and GE Medical Systems. Address correspondence to G.S.K. (e-mail: gskarczm@midway.uchicago.edu).


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 Materials and Methods
Results
 Discussion
 REFERENCES
 
The authors evaluated magnetic resonance (MR) imaging with high spectral and spatial resolutions (HSSR) of water and fat in breasts of healthy volunteers (n = 6) and women with suspicious lesions (n = 6). Fat suppression, edge delineation, and image texture were improved on MR images derived from HSSR data compared with those on conventional MR images. HSSR MR imaging data acquired before and after contrast medium injection showed spectrally inhomogeneous changes in the water resonances in small voxels that were not detectable with conventional MR imaging.

© RSNA, 2002

Index terms: Breast, MR, 00.121415, 00.121416, 00.12145 • Breast neoplasms, calcification, 00.81 • Breast neoplasms, diagnosis, 00.32, 00.81 • Breast neoplasms, MR, 00.121415, 00.121416, 00.12145


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
Materials and Methods
Results
 Discussion
 REFERENCES
 
In this article, we describe the use of echo-planar spectroscopic (EPS) imaging for magnetic resonance (MR) imaging of the breast with high spectral and spatial resolutions (HSSR). The goal of this approach is to acquire data with the spatial resolution of typical anatomic images and with a high-resolution spectrum of the water and fat signals for each image pixel. Images calculated from these data sets often provide improved image contrast, anatomic accuracy, and functional information (1–8). HSSR may be particularly useful for MR imaging of the breast because they may improve separation of water and fat signals, contrast, edge delineation, and sensitivity to contrast media and, thus, increase the sensitivity for early carcinomas.

The present application of HSSR MR imaging is a natural extension of previous work that demonstrated the importance of incorporating spectroscopic information into MR images. Dixon first demonstrated that two points of spectral resolution often allow good separation of water and fat signals (9). Glover and Schneider (10), Glover (11), and others (12,13) produced further improvements in fat-water separation and correction for B0 inhomogeneity with several points of spectral resolution. More recently, advances in MR hardware and software have allowed rapid acquisition of data with use of EPS MR imaging (14,15) at very high spectral and spatial resolutions.

With EPS MR imaging, details of the water and fat line shapes in each small image pixel can be resolved (4,16) with reasonable acquisition times. HSSR MR imaging is important when the water and fat resonances in small pixels have complicated line shapes. Inhomogeneous broadening due to local magnetic susceptibility gradients is expected on theoretic grounds (17–20) and has been demonstrated experimentally at high spatial resolution (1,3,4,21–23). This effect is especially strong in and near tumors, where deoxygenated blood and a generally heterogeneous environment produce broad complicated water lines (1,22). The detailed structure of the water line can be an important source of information regarding local (subpixel) anatomy and physiology and can be analyzed to improve contrast and accuracy on HSSR MR images (1,24).

The EPS MR imaging approach originally developed by Doyle and Mansfield (14) and Mansfield (15) and subsequently refined by others (25) allows rapid acquisition of HSSR MR imaging data sets. For example, 256 x 256 matrices in the spatial domain combined with 128–256 points in the spectral dimension and 2-Hz spectral resolution can be acquired in less than 2 minutes with the simplest k-space sampling methods. Much more rapid data acquisition is possible when k space is sampled along two or three axes during readout. Results to date suggest that the rapidly switching gradients required by EPS MR imaging on clinical MR imagers produce relatively little distortion in proton line shapes, even at resolutions of 1–2 Hz. These results suggest that the effects of eddy currents are relatively small when state-of-the-art self-shielded gradient coils are used (4,6). Thus, it is realistic to consider application of HSSR MR imaging to evaluate suspicious breast lesions in a clinical setting. The purposes of this study were to perform a preliminary evaluation of HSSR MR imaging of the breast and suspicious breast lesions and to compare HSSR MR images with conventional T2-weighted and T1-weighted fat-saturated MR images.


   Materials and Methods
TOP
 ABSTRACT
 INTRODUCTION
 Materials and Methods
Results
 Discussion
 REFERENCES
 
Data Acquisition
Six healthy female volunteers (age range, 28–43 years; mean age, 37 years ± 5 [mean ± SD]) and six female patients (age range, 40–61 years; mean age, 47 years ± 7) were enrolled in this study. The healthy volunteers (ie, women with no history of abnormalities at mammography or palpation) were recruited from among the staff of the radiology department. The patients were recruited through the breast clinic on the basis of having suspicious lesions with a largest diameter of 4 mm to 3 cm on mammograms. All six patients were considered to have high risk of breast cancer. Three patients (selected by U.B.) underwent breast biopsy (performed by the clinical pathology service); one of them was found to have a benign lesion, and the other two were determined to have infiltrating ductal carcinomas. The other three patients did not undergo biopsy. In three of the six patients, abnormalities were not depicted clearly on the conventional MR images, probably because only a limited number of sections were imaged. MR imaging in all subjects was performed according to a protocol approved by the institutional review board after informed consent had been obtained.

The patients underwent MR imaging before and after intravenous injection of a gadolinium chelate (Omniscan; Nycomed Amersham, Princeton, NJ) at a dose of 0.2 mmol per kilogram of body weight. Approximately 20 mL of the contrast media solution was injected by using an automatic injector (Medrad, Pittsburgh, Pa) at a rate of 2 mL/sec. The healthy volunteers did not receive contrast media.

The MR images were acquired with a 1.5-T clinical system (Signa; GE Medical Systems, Milwaukee, Wis) equipped with gradients with a maximum slew rate of 120 mT/m/sec and maximum amplitude of 23 mT/m. Signal was detected by using a dedicated phased-array breast coil, and shimming was performed by using the signal from the entire sensitive volume of the coil. HSSR MR images were acquired with EPS MR imaging sequences. After section selection (section thickness, 4 mm) and phase encoding (256 steps), 128 gradient echoes were acquired by using trapezoidal gradient pulses with alternating polarity. A crusher gradient was applied at the end of the echo train to eliminate artifacts due to residual transverse magnetization. Each gradient echo was sampled at 256 points. The data were digitized at a bandwidth of plus or minus 62.5 kHz, and the time between the centers of gradient echoes was 3.0 msec. The proton free induction decay was sampled for 384 msec, and the repetition time was 500 msec. The resulting data had an in-plane spatial resolution of less than 1 mm (field of view, <=24 cm), spectral resolution of 2.6 Hz, and spectral bandwidth of 333 Hz. This bandwidth was sufficient to resolve the water and fat resonances at 1.5 T. Some spectral wraparound was corrected. Both spectral and spatial resolutions were sufficient to avoid truncation artifacts. Sagittal sections were imaged with the readout gradient applied in the anteroposterior direction to minimize artifacts due to respiratory and cardiac motion.

A T2-weighted multisection fast spin-echo sequence was used to determine locations of suspicious lesions (repetition time msec/echo time msec [effective] of 4,000/176; echo train length, 12; section thickness, 4 mm; matrix, 256 x 256; field of view, ~24 mm; two signals acquired). T1-weighted MR images with fat saturation were acquired with either the fast spin-echo method (600/14 [effective]; echo train length, 12; section thickness, 4 mm; matrix, 256 x 256; field of view, ~24 mm; eight signals acquired) or the conventional spin-echo method (617/14; section thickness, 4 mm; matrix, 256 x 256; field of view, ~24 cm; one signal acquired). HSSR MR images of the selected section were acquired as described earlier, that is, two images before contrast media injection and one image between 1 and 3 minutes after injection. The acquisition time for one HSSR MR image was 128 seconds. T1-weighted fat-saturated MR imaging was repeated at 3 minutes after contrast media injection.

To evaluate the performance of the EPS MR imaging pulse sequence, spherical (~10-cm-diameter) water phantoms containing 10 mmol/L copper sulfate were imaged. The water signal in small voxels in these phantoms (as measured with single-voxel spectroscopy) had a line width of approximately 1–2 Hz. The line width in small voxels of high-resolution spectroscopic images of the phantoms acquired with EPS MR imaging was an accurate indicator of the eddy currents produced by the EPS MR imaging pulse sequence. Eddy currents would significantly broaden the lines and cause phasing artifacts due to frequency shifts during the decay of the free induction decay. EPS MR images were acquired with the same parameters as were used for breast imaging, as described earlier.

Data Analysis and Synthesis of Images
Data analysis was performed with in-house software written in IDL (Interactive Data Language; Research Systems, Boulder, Colo). Raw data generated with EPS MR imaging were processed by using methods described elsewhere (4). In summary, a three-dimensional Fourier transform with respect to two k-space axes and the evolution of the free induction decay provide a data set in which spectral and spatial information are well separated. A high-resolution proton spectrum is associated with each pixel in the image. After the water and fat signals are separated, a variety of images based on the water and fat line shapes can be produced (1,4,8,22,26).

For the purposes of the present study, images were synthesized with signal intensity proportional to the signal peak heights of water and fat (for mixed T2* and T1 contrast), integrals (for T1 contrast), line widths (for T2* contrast), and resonance frequencies. In addition, images were produced to highlight pixels with highly asymmetric water peaks or with multiple resolved components of the water resonance. To represent the effects of contrast media, images were calculated to show changes in peak height, frequency shifts, and spectrally heterogeneous effects.

Separation of Water and Fat Signals
Nonuniformity of the magnetic field across the breast caused variations in the resonance frequencies of water and fat. To identify water and fat signals in each pixel, a frequency map was constructed by using the following algorithm.

1. The largest spectral peak was automatically identified in all image pixels with signal intensities three times greater than the noise level. A rough frequency map was constructed to represent the resonance frequencies of the identified peaks. This map showed regions that primarily contained water (hereafter, a water region) and regions that primarily contained fat (hereafter, a fat region). The water and fat regions composed most but not all of the breast. Some isolated pixels were not included in either region. Abrupt transitions in resonance frequency of the largest spectral peak (~220 Hz) occurred at the interfaces between water and fat regions. Other sharp transitions equal to the spectral bandwidth (333 Hz) occurred as a result of fold back. Within each region, the frequencies of water or fat were assumed to vary slowly (ie, there were no sharp magnetic field gradients).

2. A seed pixel was manually selected (W.D. and G.S.K. together) in each of one to three fat regions and in each of one to three water regions. Water or fat signals were identified in pixels adjacent to the seed pixel by searching within a narrow bandwidth around the water or fat resonance frequencies in the seed pixel. This process was implemented iteratively by using a region-growing program written (W.D.) in IDL. Within each fat region, the water resonance frequency was calculated in each pixel by using the known fat resonance frequency and the chemical shift offset (+220 Hz). Similarly, the fat resonance frequency was determined in water regions. Fold-back effects were identified on the basis of the sharp change in water or fat resonance frequencies and were corrected by adding or subtracting the spectral bandwidth (333 Hz, or 128 frequency bins) to the frequency of the folded resonance. As a result of this correction, the resonance frequencies for water and fat varied smoothly across all fat and water regions.

3. In pixels outside the fat and water regions—usually pixels with low signal-to-noise ratios—the water and fat resonance frequencies were estimated from the frequency values in the water or fat regions by using a morphologic dilation operation. The gray-value dilation operation was repeatedly applied on the frequency map until the gaps between the water or fat regions were filled, and the water and fat resonance frequencies in all pixels throughout the breast were determined.

The water spectrum was taken to be the signal within plus or minus 10 frequency bins around the central water frequency, as obtained from the frequency map. To calculate water signal peak height and integral, the water line was fit to the magnitude of a Lorentzian line plus a linear baseline by using a nonlinear least squares curve-fitting algorithm. The water peak height was taken as the highest point of the water line after the baseline was removed with the spectral fitting routine. In principle, this procedure could not precisely fit water lines in many pixels, particularly those that were strongly asymmetric or contained multiple resolved components (23). In practice, however, the fitting procedure provided good approximations for calculation of peak-height and integral images. The procedure allowed accurate detection of a small amount of water, even in the presence of a sloping baseline due, for example, to large fat signals with broad wings or signals that leaked from other voxels.

Texture Analysis
Image texture was measured in a region of interest (ROI) selected to include as much as possible of the breast but to exclude the chest wall. ROIs were selected manually by MR physicists (X.F. and W.D. together) with advice from a radiologist (P.M.M. or U.B.). Image signal intensity in the ROI was normalized by the average value of image signal intensity in the ROI so that images from different examinations could be compared. Image signal intensity was expressed as a function (f) of the column (x) and row (y) index for each pixel. Mathematically, f(x, y) represents a surface spanning an ROI on an image coordinate plane x-y. The area of the surface, S, was found from

where the summation is over the ROI, fx = {delta}f/{delta}x, and fy = {delta}f/{delta}y. We defined the normalized surface area (Sr) as

where S(ROI) is the surface area calculated from Equation (1) and Sxy is the area of S(ROI) projected onto the x-y plane. Thus, Sr is a measure of the variation or roughness in image signal intensity (eg, if the image signal intensity is constant, then Sr is 0.0).

Locating Pixels with Multiple Peaks and Asymmetric Peaks
Multiple peaks and asymmetric peaks were identified automatically with software written in IDL. To find the number of resolved peaks in the water resonance in each pixel, all possible peaks in the water line were automatically located with the software (ie, those points in the spectrum where the signal was larger than that at the adjacent points). Then, with the program, spectral amplitude was measured at the local minima between these candidate peaks. Candidate peaks were selected as true peaks if the difference between the local maximum and minimum was two times larger than the root-mean-square of the noise level. The number of resolved peaks was counted for each pixel, and images were constructed to show the location of pixels that contained multiple water peaks.

To determine the asymmetry of magnitude spectra, the sides of the spectra on either side of the global maxima were compared. Spectra were identified as asymmetric if the difference between the integrals of the high- and low-frequency sides of the resonance was larger than 10% of the total integral.

Analysis of Effects of Contrast Media on HSSR Data
Before the effects of contrast media were evaluated, the HSSR data sets were corrected for motion in two dimensions. In-plane breast motions, including translation, rotation, and scaling, were modeled with a six-parameter coordinate transform. The postcontrast peak-height image was taken as the reference to which the precontrast peak-height image was registered. Six to 12 corresponding points that represented shared anatomic features were manually identified on the two peak-height images. The parameters of the coordinate transform were determined by minimizing the squared difference between the coordinates of the corresponding points. Finally the coordinate transform and bilinear interpolation were applied to obtain the motion-corrected precontrast HSSR data set.

Software was written in IDL to measure the change in the water signal integral, peak height, frequency, and T2* caused by the contrast media. The integral of the water magnitude spectrum before contrast media injection was subtracted from that after contrast media injection. The frequency shift due to the contrast media in each pixel was calculated after zero filling of the free induction decays to 2,048 points. T2* was estimated by dividing the peak height by the peak integral of the water spectrum for each pixel. The changes in T2* due to the contrast media were then computed.

To study the spectrally inhomogeneous effects of the contrast media (1,3,4,8,22,26), images were generated on the basis of the difference spectrum obtained by subtracting the precontrast spectrum in each pixel from the postcontrast spectrum. Both the pre- and postcontrast free induction decays were zero filled to 512 points before being Fourier transformed. The postcontrast spectrum from each pixel was then shifted until the absolute difference between the pre- and postcontrast spectra was minimized. This shift corrected for contrast media–induced changes in bulk susceptibility that shifted the entire water signal. The resulting difference spectra were used to synthesize images on which signal intensity was proportional to the greatest positive and negative amplitudes in the difference spectra and also to the greatest absolute change in spectral amplitude in each pixel. In addition, we synthesized images proportional to the product of absolute values of the greatest positive and negative changes in the difference spectrum.

Evaluation by Experienced Radiologists
The conventional and HSSR MR images of the breast in all subjects (n = 12) were reviewed independently by two radiologists (U.B. and P.M.M., who have 10 and 5 years of experience, respectively). The radiologists were provided with the medical histories including the mammographic findings, if any, but excluding the biopsy results. The radiologists evaluated the diagnostic usefulness of the HSSR MR images over that of the conventional images in terms of the separation of fat and water signals, the depiction of normal tissues, lesions, and surrounding vascular structures, and the sensitivity of the imaging technique to the contrast media.

Statistical Analysis
Nonparametric (Wilcoxon signed-rank) tests were performed with commercially available software (Minitab, State College, Pa) with data from the normal breasts. The tests were performed to compare the effectiveness of fat suppression and image texture or sharpness on water HSSR MR images and conventional T1-weighted fat-saturated MR images. Differences with P <= .05 were considered statistically significant.


   Results
TOP
 ABSTRACT
 INTRODUCTION
 Materials and Methods
Results
Discussion
 REFERENCES
 
Accurate high-resolution spectroscopic imaging with EPS MR imaging requires that the gradient waveforms be produced with minimal eddy currents and that shimming is adequate. EPS MR imaging of water phantoms produced narrow (<=2-Hz) symmetric water lines with no phasing artifacts. These images demonstrated that eddy current and shimming effects were small and did not distort spectra, even at resolution of less than 2 Hz.

The high-resolution EPS MR images demonstrated strong spectral inhomogeneity of water signal line shapes in the breast. Water signals from individual small voxels often contained multiple resolved components or were clearly asymmetric. For example, the water resonance in 12.5% ± 9.0 (mean ± SD) of all pixels in six normal breasts was composed of two or more multiple resolved components. A larger percentage of pixels in these breasts contained asymmetric pixels: 20.4% ± 10.8. These parameters were similar in the group of women with breast lesions: 20.7% ± 8.2 of pixels contained water resonances with multiple resolved components and 30.7% ± 10.9 contained asymmetric resonances.

These percentages are based on the total number of pixels in the breast and, thus, are low owing to the fact that some pixels do not contain detectable water. In pixels that contained a significant amount of MR-detectable water, approximately 30.4% ± 6.3 of the water resonances contained multiple resolved components. After contrast media injection (n = 6), 16.9% ± 8.8 of pixels in the breasts contained water signals with multiple resolved components. The spectral inhomogeneity of the water resonance is illustrated in Figure 1. An image synthesized from water signal peak height and representative spectra acquired before and after contrast media injection from selected pixels is shown. Inhomogeneous broadening of the water peak is clear in many of the spectra, and some water resonances have two or more resolvable components (spectra 3, 4, 5, and 7). The response of the water resonance to the contrast media is different from that of the fat resonance. In spectrum 1, for example, the water peak shifts to higher frequency and its magnitude increases, while the fat peak does not shift and its magnitude decreases slightly. In spectra 4, 5, and 7, the contrast agent produces a well-defined shoulder or fully resolved component on the high-frequency side (low-field-strength side) of water. The increases in overall peak height of the water signal are expected because the acquisitions are T1 weighted.



View larger version (50K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 1. Representative spectra from selected pixels before (dotted lines) and after (solid lines) contrast medium injection. Reference image shows a sagittal section through a breast with ductal carcinomas (sampled with spectra 1, 4, 5, and 7). All spectra are on the same scale and are referenced to the same carrier frequency. Frequency shifts due to B0 variation across the breast are apparent. The water peak (w) and fat peak (f) are labeled for each spectrum. Spectra in some pixels have approximately the same shape before and after contrast media injection (spectra 1, 2, 3, and 6), which indicates that effects of interimage motion have been minimized.

 
Inhomogeneous broadening of the water resonance can be used to produce images that highlight anatomic features. For one of the breasts imaged, the spatial distribution of pixels with non-Lorentzian water spectra is illustrated in Figure 2, which highlights pixels in which the water resonance has multiple components (Fig 2, B) or is asymmetric (Fig 2, C). In this breast, the water resonances in approximately 21% of the tumor pixels contain two or more resolved components, and the spectra in 39% of pixels are asymmetric. Pixels with complicated water resonances tended to be near the tumor edges or near large vessels.



View larger version (35K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 2. A, Pixels with non-Lorentzian spectral shapes are overlaid on a sagittal reference water peak-height HSSR MR image of a breast with a lesion (arrow). B, Pixels in which the water resonance contains multiple resolved peaks are highlighted. Only peaks with signal-to-noise ratio of at least 2 were identified as secondary peaks. C, Pixels with strongly asymmetric water resonances are highlighted.

 
Although asymmetry and multiple resolved components are clear indicators of inhomogeneous broadening, even relatively symmetric water resonances are often inhomogeneously broadened. Spectral inhomogeneity is clear from the effects of contrast media on distinct components of relatively symmetric water resonances (eg, spectra 3, 4, 5, and 7 in Fig 1). The effects of contrast media in many pixels were spectrally inhomogeneous, that is, the changes caused by the contrast media were not uniform across the water spectrum. The largest change in the spectrum was often not at the center of the water resonance.

To evaluate fat suppression on water HSSR MR images, the signal in parenchymal pixels on these images (ie, pixels that primarily contained water) was divided by the signal in pixels that primarily contained fat. This ratio is referred to as the fat-suppression ratio in the following data. The fat-suppression ratios were significantly higher (Wilcoxon signed-rank test, two-tail P = .036) on water HSSR MR images (3.7 ± 0.1 [ mean ± standard error of the mean]) than on T1-weighted fat-saturated images (1.4 ± 0.2). Regions of insufficient fat suppression on the conventional images were evident, but in addition the general background in predominantly fat regions on conventional fat-saturated MR images was brighter than that on water HSSR MR images. This difference is illustrated in Figure 3, which shows conventional and HSSR MR images of single sagittal sections in six healthy breasts.



View larger version (91K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 3. Conventional and HSSR MR images of six healthy breasts. For each breast (one row), four sagittal images are shown. From left to right: T2-weighted fast spin-echo MR images, T1-weighted fat-saturated fast spin-echo MR images, water peak-height HSSR MR images, and fat peak-height HSSR MR images. Image signal intensities are adjusted to a similar level, and images are displayed with identical window settings. Conventional and HSSR MR images were each acquired in approximately 1.5-2.0 minutes.

 
Improved fat suppression on water HSSR MR images enhances contrast in the parenchyma. The Table shows quantitative comparisons of the T1-weighted fat-saturated MR images and water HSSR MR images presented in Figure 3. Because the signal-to-noise ratios were similar on HSSR and T1-weighted MR images (Table), the normalized surface area (Sr in Eq [2]) can be used to compare texture and sharpness (Sr) of the HSSR MR images with those of the T1-weighted MR images. Sr values, tabulated in the Table, are larger (Wilcoxon signed-rank test, two-tailed P = .036) on water HSSR MR images than those on T1-weighted MR images. The results demonstrate that HSSR MR images generally provide better fat-water separation and better contrast and texture compared with those on conventional T1-weighted MR images with fat saturation.


View this table:
[in this window]
[in a new window]

  		Comparison of Fat-Suppression Ratio, Normalized Surface Area, and Signal-to-Noise Ratio in Six Normal Breasts

 
Strong enhancement was found on MR images obtained in the two patients with histopathologically proved invasive ductal carcinomas. Two experienced radiologists (blinded independent evaluation by U.B. and P.M.M.) agreed that image contrast was improved on HSSR MR images compared with that on conventional images and that tumor boundaries were more clearly defined. Fat signal was more effectively suppressed on the water HSSR MR images so that the water-containing structures were clearly seen against a dark background. These findings are illustrated in Figure 4, where conventional and HSSR MR images are compared. The contrast ratio for the first lesion (arrow in part A) compared with surrounding tissue (average signal in lesion divided by average signal in surrounding tissue) is 4.4 on the water peak-height HSSR MR image before contrast media injection (part C), 4.8 after contrast media injection (part D), and 1.9 on the T1-weighted fat-saturated MR images after injection (part B). The vascular network (arrow in part D) in the first breast is better depicted on HSSR MR images. Further, U.B. and P.M.M. found that the edges of the second lesion (arrow in part E) are more clearly delineated on peak-height HSSR MR images than on conventional MR images, especially near the chest wall (arrow in part H). In this case, HSSR MR images provide a clearer view of whether or not the lesion is infiltrating the chest wall.



View larger version (60K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 4. A-H, Sagittal images of breasts with carcinomas (arrows). Top row: Images in a 47-year-old patient with infiltrating ductal carcinoma, Scarff-Bloom-Richardson grade II, in the left breast. Bottom row: Images in a 47-year-old patient with infiltrating ductal carcinoma, Scarff-Bloom-Richardson grade II, with a moderate component of ductal carcinoma in situ, low grade, in the right breast. A and E are T2-weighted fast spin-echo MR images before contrast medium injection. B and F are T1-weighted fat-saturated MR images acquired 3 minutes after contrast medium injection. C and G are HSSR MR images proportional to water resonance peak height before contrast medium injection. D and H are water peak-height images obtained 1-3 minutes after the contrast medium injection. Evaluation by experienced radiologists suggested that vasculature and lesion edges are more clearly delineated on the HSSR MR images (C, G).

 
The effects of contrast agents on water signal line shape and resonance frequency can be used to produce novel images. For example, Figure 5 illustrates the effects of contrast media on HSSR data sets. Figure 5, A shows changes on an image derived from the first gradient echo of the EPS echo train (postcontrast image minus precontrast image). This image simulates the effect of the contrast media on conventional T1-weighted gradient-echo MR images. Strong signal enhancement is seen in blood vessels at the tumor rim and in some pixels inside the tumor. Figure 5, B shows the change in the water peak-height image. The inherent T2* weighting in this image shows regions of decreased or increased signal intensity. Figure 5, C shows a map of the water peak-frequency change after contrast media injection. These changes were greatest in large vessels. Frequency changes are evident in elongated structures in the tumor interior that perhaps indicate the primary veins and arteries in the tumor center. Finally, image signal intensity in Figure 5, D is proportional to the product of the absolute values of the largest positive and negative changes in the water spectrum. This product image highlights pixels with spectrally heterogeneous contrast media effects and shows changes beyond the edge of the upper quadrant of the lesion that are not clear on the conventional image (Fig 5, A). The images derived from HSSR data show strong features that are not evident on conventional images (Fig 4, A, B).



View larger version (104K):
[in this window]
[in a new window]
[Download PPT slide]
  		Figure 5. Effects of contrast medium on HSSR MR images. These sagittal images were obtained of the lesion shown in Figure 4, A-D. A, Changes in the image synthesized from the first echo in the HSSR echo train are depicted. This image simulates a conventional T1-weighted subtraction MR image. B, Changes in peak height of the water magnitude spectra are depicted. C, Changes in water resonance frequency (in hertz) are depicted. Increase in frequency gives rise to higher signal intensity on the image. D, Pixels are highlighted in which the difference spectra (postcontrast spectrum minus precontrast spectrum) show spectrally heterogeneous changes.

 

   Discussion
TOP
 ABSTRACT
 INTRODUCTION
 Materials and Methods
Results
 Discussion
REFERENCES
 
These preliminary results demonstrate the application of HSSR imaging to human breast MR imaging and evaluation of suspicious lesions. We evaluated both conventional properties (ie, contrast, texture, and fat suppression) and image characteristics obtained from high-resolution spectral information that is available only with HSSR MR imaging. In both cases, our results suggest that HSSR MR imaging has some advantages over conventional MR imaging. Results of both quantitative analysis and evaluations by experienced radiologists suggest that HSSR MR images may show anatomy that is more clearly defined than that on conventional MR images. The HSSR MR images (eg, Fig 4) show well-defined lesion boundaries, which are important clinically because irregular lesion contours are a sign of malignancy (27,28). In addition, detailed spectral data allowed accurate fitting of the water peak and separation of the water signal from the spectral baseline. As a result, small water signals could be accurately detected in the presence of very large fat signals, even when the wings of fat spectra extended into the water region, owing to very rapid decay in the time domain. The improved separation of the water and fat signals increases sensitivity to small amounts of water, and this increase may facilitate detection of cancer.

The HSSR MR images were acquired 1–3 minutes after contrast media injection, whereas the conventional T1-weighted fat-saturated MR images were acquired 3–5 minutes after injection, which caused some bias in favor of HSSR MR images. However, there was no data acquisition at the peak of the contrast media bolus; both HSSR and conventional MR images were acquired during a period after the peak enhancement is usually reached, and signal intensity changes during this period are generally slow. Therefore, comparison of the HSSR and conventional MR images acquired after contrast media injection is reasonable.

The HSSR data demonstrate that the water line in breast tissue is inhomogeneously broadened and often contains multiple resolved components. It is very unlikely that this spectral inhomogeneity is a consequence of poor shimming. In these very small image pixels, magnetic field gradients due to shimming are approximately linear and, therefore, broaden the water line but do not cause detectable asymmetry or multiple resolved components. Spectrally inhomogeneous effects of the contrast media were detected in many pixels; this finding is consistent with those in previous studies of animal models of cancer (1,3,8,22) and with theoretic predictions (17,19,20).

We sometimes found strong but opposing changes in the amplitudes of different spectral components in the same image pixel. Thus, high spectral resolution makes it possible to detect strong effects of contrast media that are missed with conventional MR imaging. Large changes in a single small component of an inhomogeneously broadened resonance may reflect subpixel regions of high vascular density associated with angiogenesis. This hypothesis is consistent with the fact that spectrally inhomogeneous contrast media effects appear to cluster near the edges of tumors, where strong angiogenic activity is expected (Fig 5, D; references 1 and 3).

With the current relatively slow method of data acquisition, high-resolution EPS MR images may be useful for improved characterization of suspicious lesions identified with conventional MR imaging or other modalities. However, much faster versions of the method can be implemented that involve the use of two- or three-dimensional trajectories through k space during the proton free induction decay (15). Other methods for more efficient and rapid data acquisition are also available. Methods (eg, simultaneous acquisition of spatial harmonics [29]) for parallel data acquisition can dramatically increase speed, and spectral resolution can be reduced at the edges of k space. Thus, it is likely that high-quality HSSR MR images can be obtained more rapidly or from a large volume of tissue with adequate signal-to-noise ratio.

A definitive evaluation of HSSR MR imaging would include direct comparison with optimal multiple-point Dixon methods (9–11), optimal fat-suppression methods (eg, rotating delivery of excitation off resonance [30]), and other MR imaging methods (eg, magnetization transfer contrast material–enhanced MR imaging [31]). Such an evaluation is feasible since EPS MR imaging or related methods are available in a number of laboratories where they are used to improve the accuracy and sensitivity of anatomic and functional MR imaging (2,4,7,18,21,32–34). In addition, the hardware and software required to produce HSSR EPS data sets are either commonly available or can be implemented easily. Thus, more comprehensive clinical testing of this approach for MR imaging of the breast is practical.

In conclusion, these preliminary results, as well as findings in previous work (1,3,6–8,22,24,26,33), suggest potential advantages of HSSR MR imaging of the breast. Advantages of HSSR MR imaging include improved anatomic accuracy and edge delineation, improved fat suppression, increased sensitivity to small amounts of water-containing tissue, strong T2* contrast, increased sensitivity to contrast media, and increased sensitivity to subpixel environments (1,3). Future improvements in HSSR MR imaging will reduce data acquisition time. Even without these improvements, we believe the information provided with HSSR MR imaging may justify the longer acquisition times.


   ACKNOWLEDGMENTS
 
G.S.K. is grateful for the very helpful advice of Dr Alan Koretsky, Dr Frederick Kelcz, Dr Edward Hendrick, and Dr Martin Lipton.


   FOOTNOTES
 
Abbreviations: EPS = echo-planar spectroscopic, HSSR = high spectral and spatial resolutions, ROI = region of interest

Author contributions: Guarantors of integrity of entire study, G.S.K., U.B.; study concepts, G.S.K., Y.P.D., U.B., P.M.M.; study design, G.S.K., W.D.; literature research, G.S.K., W.D.; clinical studies, W.D., Y.P.D., P.M.M., U.B., M.A.Z., G.S.K.; experimental studies, W.D., Y.P.D., P.M.M., M.A.Z., M.M., G.S.K.; data acquisition, W.D., Y.P.D., P.M.M., M.A.Z., M.M., G.S.K.; data analysis/interpretation, W.D., X.F., P.M.M., U.B., M.M., G.S.K.; statistical analysis, W.D., X.F., M.M.; manuscript preparation, W.D., X.F., M.M., G.S.K., P.M.M., U.B.; manuscript definition of intellectual content, W.D., Y.P.D., U.B., P.M.M., G.S.K.; manuscript editing, W.D., X.F., M.M., G.S.K.; manuscript revision/review, all authors; manuscript final version approval, G.S.K.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 Materials and Methods
Results
 Discussion
 REFERENCES
 

  1. Karczmar GS, Fan X, Al-Hallaq HA, et al. Uptake of a superparamagnetic contrast agent imaged by MR with high spectral and spatial resolution. Magn Reson Med 2000; 43:633-639.[CrossRef][Medline]
  2. Yang QX, Demeure RJ, Dardzinski BJ, Arnold BW, Smith MB. Multiple echo frequency-domain image contrast: improved signal-to-noise ratio and T2 (T2*) weighting. Magn Reson Med 1999; 41:423-428.[CrossRef][Medline]
  3. Karczmar G, Fan X, Al-Hallaq H, et al. Functional and anatomic imaging of tumor vasculature: high resolution MR spectroscopic imaging combined with a superparamagnetic contrast agent. Acad Radiol 2002; 9(suppl 1):S115-S118.
  4. Kovar DA, Al-Hallaq HA, Zamora MA, River JN, Karczmar GS. Fast spectroscopic imaging of water and fat resonances to improve the quality of MR images. Acad Radiol 1998; 5:269-275.[CrossRef][Medline]
  5. Kuperman V, River JN, Karczmar GS. High resolution spectroscopic images of tumors (abstr) In: Proceedings of the Society of Magnetic Resonance in Medicine and the European Society of Magnetic Resonance in Medicine and Biology. Berkeley, Calif: Society of Magnetic Resonance in Medicine, 1995.
  6. Sarkar S, Heberlein K, Metzger GJ, Zhang X, Hu X. Applications of high-resolution echoplanar spectroscopic imaging for structural imaging. J Magn Reson Imaging 1999; 10:1-7.[CrossRef][Medline]
  7. Posse S, Wiese S, Gembris D, et al. Enhancement of BOLD-contrast sensitivity by single-shot multi-echo functional MR imaging. Magn Reson Med 1999; 42:87-97.[CrossRef][Medline]
  8. Oikawa H, Al-Hallaq HA, Lewis MZ, River JN, Kovar DA, Karczmar GS. Spectroscopic imaging of the water resonance with short repetition time to study tumor response to hyperoxia. Magn Reson Med 1997; 38:27-32.[Medline]
  9. Dixon WT. Simple proton spectroscopic imaging. Radiology 1984; 153:189-194.[Abstract/Free Full Text]
  10. Glover GH, Schneider E. Three-point Dixon technique for true water/fat decomposition with B0 inhomogeneity correction. Magn Reson Med 1991; 18:371-383.[Medline]
  11. Glover GH. Multipoint Dixon technique for water and fat proton and susceptibility imaging. J Magn Reson Imaging 1991; 1:521-530.[Medline]
  12. Szumowski J, Coshow WR, Li F, Quinn SF. Phase unwrapping in the three-point Dixon method for fat suppression MR imaging. Radiology 1994; 192:556-561.
  13. Wang Y, Li D, Haacke EM, Brown JJ. A three-point Dixon method for water and fat separation using 2D and 3D gradient-echo techniques. J Magn Reson Imaging 1998; 8:703-710.[Medline]
  14. Doyle M, Mansfield P. Chemical shift imaging: a hybrid approach. Magn Reson Med 1987; 5:255-261.[Medline]
  15. Mansfield P. Spatial mapping of the chemical shift in NMR. Magn Reson Med 1984; 1:370-386.[Medline]
  16. Kovar DA, Lewis MZ, River JN, Lipton MJ, Karczmar GS. In vivo imaging of extraction fraction of low molecular weight MR contrast agents and perfusion rate in rodent tumors. Magn Reson Med 1997; 38:259-268.[Medline]
  17. Chu SC, Xu Y, Balschi JA, Springer CSJ. Bulk magnetic susceptibility shifts in NMR studies of compartmentalized samples: use of paramagnetic reagents. Magn Reson Med 1990; 13:239-262.[Medline]
  18. Xu Y, Balschi JA, Springer CSJ. Magnetic susceptibility shift selected imaging: MESSI. Magn Reson Med 1990; 16:80-90.[Medline]
  19. Bauer WR, Nadler W, Bock M, et al. Theory of the BOLD effect in the capillary region: an analytical approach for the determination of T2 in the capillary network of myocardium. Magn Reson Med 1999; 41:51-62.[CrossRef][Medline]
  20. Yablonskiy DA. Quantitation of intrinsic magnetic susceptibility-related effects in a tissue matrix: phantom study. Magn Reson Med 1998; 39:417-428.[Medline]
  21. Zhong K, Li X, Shachar-Hill Y, Picart F, Wishnia A, Springer CSJ. Magnetic susceptibility shift selected imaging (MESSI) and localized (1)H(2)O spectroscopy in living plant tissues. NMR Biomed 2000; 13:392-397.[CrossRef][Medline]
  22. Al-Hallaq HA, Zamora M, Fish BL, Farrell A, Moulder JE, Karczmar GS. MRI measurements correctly predict the relative effect of tumor oxygenating agents on hypoxic fraction in rodent BA1112 tumors. Int J Radiat Oncol Biol Phys 2000; 47:481-488.[CrossRef][Medline]
  23. Al-Hallaq HA, Fan X, Zamora M, River JN, Moulder JE, Karczmar GS. Spectrally inhomogeneous BOLD contrast changes detected in rodent tumors with high spectral and spatial resolution MRI. NMR Biomed 2002; 15:28-36.[CrossRef][Medline]
  24. Kovar DA, Lewis M, Karczmar GS. A new method for imaging perfusion and contrast extraction fraction: input functions derived from reference tissues. J Magn Reson Imaging 1998; 8:1126-1134.[Medline]
  25. Nelson SJ, Vigneron DB, Star-Lack J, Kurhanewicz J. High spatial resolution and speed in MRSI. NMR Biomed 1997; 10:411-422.[CrossRef][Medline]
  26. Al-Hallaq HA, Karczmar G. High resolution 1H spectroscopic imaging of the water and fat resonances in human breast (abstr) In: Proceedings of the Fifth Meeting of the International Society for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 1997; 1055.
  27. Schnall MD, Ikeda DM. Lesion Diagnosis Working Group report. J Magn Reson Imaging 1999; 10:982-990.[CrossRef][Medline]
  28. Orel SG, Schnall MD, LiVolsi VA, Troupin RH. Suspicious breast lesions: MR imaging with radiologic-pathologic correlation. Radiology 1994; 190:485-493.[Abstract/Free Full Text]
  29. Bankson JA, Griswold MA, Wright SM, Sodickson DK. SMASH imaging with an eight element multiplexed RF coil array. MAGMA 2000; 10:93-104.
  30. Soderstrom CE, Harms SE, Copit DS, et al. Three-dimensional RODEO breast MR imaging of lesions containing ductal carcinoma in situ. Radiology 1996; 201:427-432.[Abstract/Free Full Text]
  31. Balaban R, Ceckler TL. Magnetization transfer contrast in magnetic resonance imaging. Magn Reson Q 1992; 8:116-137.[Medline]
  32. Kovar DA, Karczmar GS. Fast spectroscopic imaging of water and fat proton resonances improves image contrast and signal-to-noise ratio (abstr) In: Proceedings of the Fifth Meeting of the International Society for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 1997; 1834.
  33. Hilaire L, Wehrli FW, Song HK. High-speed spectroscopic imaging for cancellous bone marrow R(2)* mapping and lipid quantification. Magn Reson Imaging 2000; 18:777-786.[CrossRef][Medline]
  34. Speck O, Hennig J. Functional imaging by I0- and T2*-parameter mapping using multi-image EPI. Magn Reson Med 1998; 40:243-248.[Medline]



This article has been cited by other articles:


Home page
 Am. J. Roentgenol.Home page
M. Medved, G. M. Newstead, H. Abe, M. A. Zamora, O. I. Olopade, and G. S. Karczmar
High Spectral and Spatial Resolution MRI of Breast Lesions: Preliminary Clinical Experience
Am. J. Roentgenol., January 1, 2006; 186(1): 30 - 37.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
2242011022v1
224/2/577    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Du, W.
Right arrow Articles by Karczmar, G. S.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Du, W.
Right arrow Articles by Karczmar, G. S.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
RADIOLOGY RADIOGRAPHICS RSNA JOURNALS ONLINE