| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Developments |
1 From the Departments of Radiology, Brigham and Women’s Hospital (S.E.M., H.M., Y.M., F.A.J.) and Children’s Hospital (R.V.M.), Harvard Medical School, 75 Francis St, Boston, MA 02115; and the Diagnostics Center, Pannon University of Agriculture, Kaposvar, Hungary (P.B., G.B., I.R.). Received June 27, 2000; revision requested August 7; revision received September 6; accepted September 11. S.E.M. supported by grants from the Whitaker Foundation and the National Institutes of Health (NIH 1R01 NS39335-01A1). P.B. supported by Hungarian Scientific Research Fund (OTKA grants F 016343, T 034200). Address correspondence to S.E.M. (e-mail: stephan@bwh.harvard.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
Index terms: Brain, edema, 10.86 • Brain, gray matter • Brain, white matter • Brain neoplasms, MR, 10.121411, 10.12143 • Magnetic resonance (MR), contrast enhancement, 10.12143 • Magnetic resonance (MR), diffusion study, 10.12144 • Magnetic resonance (MR), tissue characterization
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
In several recent publications (4–9), diffusion-weighted imaging has been proposed as a novel mechanism for producing contrast in the demarcation of different cerebral tumors. Apparent diffusion coefficient (ADC) maps of brain tumors seem also to provide useful information about structural details of tumors (5,7,8). According to these reports, peritumoral edema and solid enhanced, solid necrotic nonenhanced, and cystic parts can be recognized on ADC maps. Diffusion tensor imaging adds information about the directional dependence of molecular diffusion that may also be helpful in the demarcation of tumor margins (10). Nevertheless, any of these new diffusion imaging methods used with contrast material–enhanced relaxation-weighted imaging fails to be specific enough in every case (5,6).
Routine diffusion imaging of the brain generally involves the use of b factors within the range of 0 to 1,000 sec/mm2. ADC maps are then generated, based on the assumption that the relationship between the MR signal and b factor is monoexponential. Recently, however, it was shown (11) that for rat brain, the signal decay with b factors in an extended range of up to 10,000 sec/mm2 is better described with a biexponential curve. Similar findings were made in human brain by using multiple b factors of up to 6,000 sec/mm2 (12). Both studies lack the anatomic details needed for clinical application, since diffusion was measured only within a localized volume (11) or along a column (12).
In the present study, our goal was to obtain diffusion-weighted images of the human brain with b factors ranging from 5 to 5,000 sec/mm2. Biexponential fits were applied to the measured signal of each voxel. With the four parameters that describe the biexponential fit, we attempted to characterize normal white and gray matter, edematous white matter, and brain tumors.
|
Materials and Methods |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
MR Diffusion Imaging
Each patient underwent the routine clinical imaging protocol. This protocol included T2-weighted (3,000–3,600/80–98, repetition time msec/echo time msec; field of view, 220–240 x 220–240 mm; section thickness, 3–5 mm) and T1-weighted (500–700/14–25; field of view, 220–240 x 220–240 mm; section thickness, 3–5 mm) spin-echo sequences before and after contrast enhancement (Magnevist [gadopentetate dimeglumine]; Berlex Laboratories, Wayne, NJ; 0.2 mL per kilogram of body weight administered in the cubital vein) to localize the tumor mass and other pathologic details. For that obvious reason, in most cases diffusion imaging was performed only after the administration of contrast material. In four patients, the section for diffusion imaging of tumor tissue was determined with nonenhanced images, and diffusion imaging was performed prior to the administration of contrast agent. One patient with tumor underwent diffusion imaging before and after the administration of contrast material.
Diffusion-weighted images with a wide range of b factors were obtained with line scan diffusion imaging. Aspects of the MR physics and the feasibility of this single-shot column-sampling technique have been presented (13). Patient studies with line scan diffusion imaging (14,15) have demonstrated the usefulness of this technique for conventional diffusion imaging. The line scan diffusion sequence was implemented by using a 1.5-T whole-body system (Signa Echospeed; GE Medical Systems, Milwaukee, Wis) with version 5.7 software. The maximum gradient strength was 22 mT/m. The standard birdcage coil was used, and neither cardiac gating nor head restraints were used.
Images were acquired with a rectangular field of view of 220 x 165 mm and a matrix size of 64 x 48 or 126 x 96 columns. The effective section thickness (13) was set at 6.0–7.3 mm. The receiver bandwidth was set at 6.25 kHz, which was found to be the best compromise in view of the decreased signal-to-noise ratio at higher bandwidths and the augmented image distortions caused by field inhomogeneities or chemical shifting at lower bandwidths. Sixteen images with linearly increasing diffusion weighting between 5 and 5,000 sec/mm2 were acquired.
For patient imaging, diffusion was measured along only a single direction, in a (1,1,1) gradient configuration to achieve maximal diffusion encoding with a minimal echo time of 94 msec. A b factor of 5,000 sec/mm2 was attained with trapezoidal gradient pulses of 36.2 mT/m amplitude, 40-msec pulse duration (
), and 46 msec between the onset of the first and second gradient pulses (
). In healthy subjects, data along six non-colinear directions were collected with the tensor configuration described by Basser and Pierpaoli (16) by using an echo time of 107 msec. The repetition time and effective repetition time (13) were 204 and 2,040–3,600 msec, respectively; the total imaging time was 3 minutes per section (one diffusion direction with 16 b factors). Shorter repetition and imaging times would have been possible had gradient heating not been a concern.
Data Analysis
Data analysis was performed offline at workstations (Sun Microsystems, Mountain View, Calif) by using MATLAB software (Math Works, Natick, Mass). A nonlinear least-squares Marquardt algorithm was used for each pixel to fit brain signal intensity decay S with diffusion-weighting b to a biexponential function of the following form: S = A1 exp(-ADC1b) + A2 exp(-ADC2b), where ADC1 and ADC2 are the ADCs, with the signal amplitudes for ADC1 (A1) and ADC2 (A2).
Data points were included in the fit only if their signal exceeded three times the noise baseline, defined by the mean signal intensity in the four corners of the image. Changes in the calibration of the MR apparatus do not permit a direct comparison among biexponential signal amplitudes measured during different acquisitions. We therefore calculated the relative fraction of the biexponential signal amplitude of the slow diffusing component as follows: A2/(A1 + A2). Preliminary experiments with a phantom that simulates the geometry and load of a human head (head SNR phantom model 46-287900G3; GE Medical Systems) showed that within a 170-mm diameter the signal variations are 4.4%. We considered these variations small enough to permit normalization of individual tissue values (A1T and A2T) with individual white matter values (A1W and A2W) measured at a different location.
Maps of the fit parameters were used for region-of-interest (ROI) analysis with dedicated image-analysis software (XPhase) developed at our institution. With this program, the contour drawing process can be simultaneously controlled on different image backgrounds (eg, T1-weighted, T2-weighted, and ADC). ROIs were drawn manually by two of the authors (P.B., G.B.) in consensus.
The ROI for tumor tissue was defined according to contrast enhancement on the conventional T1-weighted images. Separate ROIs were drawn for cystic parts of tumor lesions. The ROI for peritumoral edema was defined with the use of nonenhanced T2-weighted images and contrast-enhanced T1-weighted images. Nonpathologic periventricular white matter areas were selected as a healthy control. To reduce the influence of directional diffusion, the mean value of two independent and large white matter ROIs was used. Gray matter values were determined in the cortex. In all cases, ROI size and placement were selected so that bordering tissues of ambiguous origin were excluded as much as possible. Moreover, to avoid partial volume effects in the ROI analysis of the relatively thick sections used at line scan diffusion imaging, considerable effort was made to determine that the tissue of interest was also present on neighboring sections of the conventional images.
ROI sizes for tumors ranged between 5 and 29 pixels (mean, 15.4 pixels or 182 mm2). All other ROIs were, on average, larger; that is, there were 27.7 pixels for cysts, 24.6 pixels for peritumoral edema, 62.7 pixels for gray matter, and 25.4 pixels for gray matter. Significant differences among mean ROI values were verified with a two-sided Student t test, with P values less than .05 considered to indicate a significant difference.
|
Results |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
|
|
|
2 values that were more than an order of magnitude larger than those obtained from biexponential fits. Triexponential fitting functions yielded unstable parameters, with no significant decrease in 2 values compared with those obtained from biexponential fits. Mean ROI values and SDs of the fast-diffusing component (ADC1), the slow-diffusing component (ADC2), and the relative fraction of the slow-diffusing component are summarized in the Table. For each cystic lesion, only one ADC value was calculated from the slope of the monoexponential line fit to the logarithm of the MR signal intensity. The measured ADC for cystic lesions was, on average, comparable with the ADC of water at 37°C (3.0 µm2/msec [17]). The biexponential signal intensity ratio between different tissues and normal white matter are listed in the Table. Moreover, to provide a quantitative expression of the appearance of the pathologic areas on the diffusion-weighted images, corresponding ratios of the MR signal intensities at low, high, and very high diffusion weighting are also given in the Table. The data in one of the patients with a metastasis were not suitable for analysis because of severe motion artifacts. In one of the patients with glioblastoma, peritumoral edema was not depicted in the selected section.
|
|
|
Figure 5 shows images of a metastasis of a breast adenocarcinoma in the left occipital lobe. On the T2-weighted image, the tumor appears to have a signal intensity that is slightly higher than that of normal white matter. The lesion is surrounded by edema that extends to the thalamus and the dorsal limb of the external capsule. The A1 map, which represents the biexponential signal amplitude of the fast-diffusing component of tissue water, clearly delineates the edema without tumor. The tumor is seen separately on the A2 map that represents the biexponential signal amplitude of the slow-diffusing component.
|
|
|
|
|
Discussion |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
On the other hand, in areas near large bone structures, the sensitivity of single-shot echo-planar imaging to susceptibility variations and chemical shifting can, with inadequate shimming, result in ghosting artifacts, image distortions, or complete signal loss, whereas images obtained with line scan diffusion imaging do not show such artifacts (14,15). In addition, eddy currents that are caused by the application of the diffusion encoding gradients can generate distortion artifacts (19). One would expect that these distortion artifacts increase with higher b factors. With line scan diffusion imaging, however, no distortions were observed, even with the highest b factors.
These findings, in agreement with those of earlier reports (11,12), revealed that biexponential fits better describe diffusion-related signal loss in the brain. The deviation from a purely monoexponential model that becomes evident as the diffusion encoding is extended beyond the normal range should not be confounded with effects that arise from microcirculation (20,21). Signal loss due to microcirculation is observed only with b factors of less than 300 sec/mm2 and therefore cannot account for the deviations measured in various tissues. A reanalysis of the data without use of the lowest b factor (ie, only 15 b factors between 338 and 5,000 sec/mm2) revealed the same biexponential behavior. The normal human white matter parameters observed in the current study are in good agreement with those of an earlier study (12) in which a considerably higher number of b factors over a slightly larger range were used. Limitations in the interpretation of the slow ADC component as intracellular water and the fast ADC component as extracellular water, including a discrepancy between the ADC component volume fractions and reported values of intracellular and extracellular water volume fractions in brain, were discussed by Niendorf et al (11).
The novelty of this study, besides the acquisition of these fit parameters in image formats, is the observation of biexponential diffusion in pathologic brain tissues. Moreover, the four parameters that describe the biexponential fit seem to permit the distinction among the various tissues. In normal cortical gray matter compared with normal white matter, all parameters except the biexponential signal amplitude A2 were strongly elevated. The low spatial resolution of the collected data may have resulted in a contamination of the measured signal with cerebrospinal fluid signal. This signal contamination would primarily affect the measurement of the fast-diffusing component, since the monoexponential diffusion constant of cerebrospinal fluid is relatively high. In subacutely infarcted tissue, a lower ADC for the fast-diffusing component (Table) is expected, since it is known from conventional diffusion imaging that diffusion is reduced (22). In the present study, the ADC of the slow-diffusing component was also reduced. This observation, however, is not well corroborated, since only one patient with stroke participated in the study.
Characteristic changes in biexponential diffusion parameters, which remain to be explained, were observed in peritumoral edema and tumor tissue; compared with the value in white matter, the ADC1 value increased by almost 50%. The A1 value increased by 178% in edema and 118% in tumor tissue. While there were only minimal changes in the ADC value of the slow-diffusing component, the biexponential signal amplitude A2 decreased in peritumoral edematous white matter, whereas it increased in tumor tissue. We do not have an explanation for the contrary behavior of the biexponential signal amplitude of the slow-diffusing component in the two tissues, since the exact nature of the biexponential signal attenuation is not known. The amplitude of each diffusion constant is influenced by the spin density and relaxation time of each observed component. Since spin density and the relaxation times of the components are not necessarily the same, imaging protocols with different settings of the repetition and/or echo time may yield different amplitude values.
In another study (23), however, we found no statistically significant differences in T1 relaxation. It can be speculated that the slow-diffusing component is determined by the concentration of water-binding macromolecules, cellular size, and tissue architecture (tortuosity) (24–28), that is, factors that are indeed different among the investigated tissues. Thus ADC1, ADC2, and the biexponential signal amplitudes A1 and A2 permit the separation of the various tissues.
Changes in tumor tissue appear to be more variable, which most likely reflects the fact that tumors do not represent a single type of tissue. From the limited number of different tumor types studied with this technique to date, correlations between histologic type and biexponential diffusion parameters cannot be established. Cystic regions are easily distinguished by their monoexponential diffusion and high ADC.
Another limitation of the current study is the monodirectional diffusion encoding used for patient imaging. As pointed out in the Data Analysis section, to reduce the potential error in white matter tracts where restricted diffusion is most likely to be present, the mean value of two independent and large white matter ROIs was used. For each case (eg, Fig 6), we also verified that the observed tumor enhancement was not due to restricted diffusion in a white matter tract.
Diffusion-based tissue differentiation does not depend on the use of contrast agents, since the diffusion-weighted images are only minimally T1-weighted (effective repetition time, 2,040 msec) (29). The tumor-tissue contrast generated on multicomponent ADC maps is different from the enhancement produced by paramagnetic contrast agents. The contrast agents in use are not specific to tumor tissue, but rather, they enhance areas where the blood-brain barrier has become permeable due to neoplastic growth, surgical procedures, or radiation therapy (2,3). Moreover, not all tumor types enhance with the use of contrast agents. On multicomponent ADC maps, on the other hand, as demonstrated in Figures 6 and 7, enhancement occurs in the solid part of the tumor.
Diffusion-weighted imaging has been previously evaluated (5–7) in the characterization of brain tumors and associated pathologic structures. With this study, we demonstrated that only limited information is gained from diffusion imaging in the normal b factor range (Fig 3). With very high b factors, however, the signal in tumor tissue is elevated in comparison with that of surrounding edema and normal white matter; the elevated signal is due to the higher biexponential signal amplitude of the slow-diffusing component in tumor tissue. The data in the Table and the diffusion-weighted images in Figure 8 reiterate this observation: With low diffusion weighting, peritumoral edema and tumor tissue are enhanced; at high diffusion weighting (around 1,000 sec/mm2), the contrast among the different tissues is minimal; and at very high diffusion weighting, tumor tissue is selectively enhanced.
From the equation in the Materials and Methods section, it follows that for b equal to 0, the measured signal equals A1 plus A2. At very high diffusion weighting, the signal is dominated by the second diffusion component, that is, S 
A2 exp(-ADC2b). Consequently, for white matter, edema, and tumor, where ADC2 is similar (Fig 4b, Table), images obtained with very high diffusion imaging may be considered A2-weighted. Thus, for routine diagnostic imaging, it may be sufficient to acquire only one diffusion-weighted image with very high diffusion weighting. Signal averaging by using different diffusion encoding directions could then be performed to overcome a poor signal-to-noise ratio and limited spatial resolution. Trace diffusion weighting would be necessary to eliminate selective enhancement of white matter tracts (Fig 8), which could be misinterpreted as lesions.
In summary, we investigated a method for tissue characterization with multiexponential analysis of the diffusion-attenuated signal. We examined brain tumors, and our results indicated that the calculated images of slow- and fast-diffusing components helped distinguish edema and tumor tissue in a number of cases. Provided that more experience is gained with different tumor types, the technique could potentially be used without paramagnetic contrast agents for tumor localization. Clearly, further experiments are needed to reveal the biophysical basis of the described phenomenon and to optimize the method for routine clinical diagnosis.
|
FOOTNOTES |
|---|
Author contributions: Guarantor of integrity of entire study, S.E.M.; study concepts, S.E.M., R.V.M.; study design, S.E.M., P.B.; definition of intellectual content and editing, S.E.M., P.B., R.V.M.; literature research, P.B.; clinical studies, P.B.; data acquisition, P.B., S.E.M.; data analysis, P.B., G.B., S.E.M., H.M., Y.M.; statistical analysis, S.E.M., P.B., G.B., H.M., Y.M.; manuscript preparation, S.E.M., P.B.; manuscript review, R.V.M., F.A.J., I.R.; manuscript final version approval, S.E.M., R.V.M.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods ResultsDiscussionREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  H.S. Seo, K.-H. Chang, D.G. Na, B.J. Kwon, and D.H. Lee High b-Value Diffusion (b = 3000 s/mm2) MR Imaging in Cerebral Gliomas at 3T: Visual and Quantitative Comparisons with b = 1000 s/mm2 AJNR Am. J. Neuroradiol., March 1, 2008; 29(3): 458 - 463. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. M. Provenzale, S. Mukundan, and D. P. Barboriak Diffusion-weighted and Perfusion MR Imaging for Brain Tumor Characterization and Assessment of Treatment Response. Radiology, June 1, 2006; 239(3): 632 - 649. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. Le Bihan, S.-i. Urayama, T. Aso, T. Hanakawa, and H. Fukuyama Direct and fast detection of neuronal activation in the human brain with diffusion MRI PNAS, May 23, 2006; 103(21): 8263 - 8268. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. Goebell, S. Paustenbach, O. Vaeterlein, X.-Q. Ding, O. Heese, J. Fiehler, T. Kucinski, C. Hagel, M. Westphal, and H. Zeumer Low-Grade and Anaplastic Gliomas: Differences in Architecture Evaluated with Diffusion-Tensor MR Imaging Radiology, April 1, 2006; 239(1): 217 - 222. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Brugieres, P. Thomas, A. Maraval, H. Hosseini, C. Combes, A. Chafiq, L. Ruel, S. Breil, M. Peschanski, and A. Gaston Water Diffusion Compartmentation at High b Values in Ischemic Human Brain AJNR Am. J. Neuroradiol., May 1, 2004; 25(5): 692 - 698. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. F. L. Schneider, K. A. Il'yasov, E. Boltshauser, J. Hennig, and E. Martin Diffusion Tensor Imaging in Cases of Adrenoleukodystrophy: Preliminary Experience as a Marker for Early Demyelination? AJNR Am. J. Neuroradiol., May 1, 2003; 24(5): 819 - 824. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. Mardor, R. Pfeffer, R. Spiegelmann, Y. Roth, S. E. Maier, O. Nissim, R. Berger, A. Glicksman, J. Baram, A. Orenstein, et al. Early Detection of Response to Radiation Therapy in Patients With Brain Malignancies Using Conventional and High b-Value Diffusion-Weighted Magnetic Resonance Imaging J. Clin. Oncol., March 15, 2003; 21(6): 1094 - 1100. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. Bulakbasi, M. Kocaoglu, F. Ors, C. Tayfun, and T. Ucoz Combination of Single-Voxel Proton MR Spectroscopy and Apparent Diffusion Coefficient Calculation in the Evaluation of Common Brain Tumors AJNR Am. J. Neuroradiol., February 1, 2003; 24(2): 225 - 233. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
