IMAGING & THERAPEUTIC TECHNOLOGYFree Access

Intensity-modulated Parametric Mapping for Simultaneous Display of Rapid Dynamic and High-Spatial-Resolution Breast MR Imaging Data

Abstract

Contrast material–enhanced magnetic resonance (MR) imaging of the breast has variable specificity for differentiation of breast cancer from other enhancing conditions. Two principal strategies to improve its specificity are rapid dynamic MR imaging and high-spatial-resolution MR imaging. A method was developed of combining contemporaneously acquired dynamic and high-spatial-resolution MR imaging data into a single integrated display. Whole-breast rapid dynamic data were condensed into a color map by using pharmacokinetic analysis. The pharmacokinetic results were combined with the high-spatial-resolution images with a new technique that preserves underlying morphologic details. This new method was evaluated by five radiologists for eight breast lesions, and the results were compared with those of the standard method of overlaying parametric map data. The radiologists' ratings showed a statistically significant preference for the intensity-modulated parametric map display method over the overlaid parametric display method for 10 of the 12 evaluation criteria. The new method enabled simultaneous visualization of pharmacokinetic and morphologic information, facilitated assessment of lesion extent, and improved the suppression of noise in the pharmacokinetic data. The ability to simultaneously assess both dynamic and high-spatial-resolution features may ultimately improve the specificity of breast MR imaging.

Introduction

Although contrast material–enhanced magnetic resonance (MR) imaging of the breast is exqui-sitely sensitive for detection of invasive breast cancer, its reported specificity is variable (^,1–^,6). Two principal strategies have evolved to improve specificity: rapid dynamic imaging of gadolinium enhancement and high-spatial-resolution imaging. Although the former technique provides valuable diagnostic information (^,7,^,8), whole-breast rapid dynamic MR imaging produces large, unwieldy data sets (^,4,^,7) that are difficult to correlate with separately acquired high-spatial-resolution images.

Previous studies have used pharmacokinetic modeling to condense dynamic MR imaging data into color maps of quantitative parameters of gad-olinium enhancement (^,9,^,10) that are overlaid onto corresponding gray-scale anatomic images (^,11). The overlaying of these maps may obscure morphologic features of enhancing lesions that are important predictors of the risk of malignancy (^,12,^,13). The purpose of this study was to develop and evaluate an efficient method of simultaneously displaying pharmacokinetic maps and high-spatial-resolution images in a manner that preserves visualization of both pharmacokinetic information and morphologic details in a single set of high-resolution color images. Specific topics discussed are the imaging protocol, the choice of pulse sequences, pixel-by-pixel pharmacokinetic modeling, parametric map presentation, radiologist evaluation of intensity-modulated parametric maps, simultaneous visualization of pharmacokinetics and morphology, assessment of lesion extent, noise suppression, and limitations.

Imaging Protocol

A 1.5-T scanner (Echospeed; GE Medical Systems, Milwaukee, Wis) and a phased-array breast coil (MRI Devices, Waukesha, Wis) were used. Whole-breast rapid dynamic MR images and high-spatial-resolution fat-nulled MR images were contemporaneously acquired by using a combination of dynamic three-dimensional (3D) spiral MR imaging (^,14) during the wash-in phase of an intravenous injection of gadolinium contrast material followed immediately by high-spatial-resolution 3D spectral-spatial excitation magnetization transfer (3DSSMT) imaging (^,15) followed immediately by more dynamic 3D spiral MR imaging during the washout phase of enhancement. Unlike some protocols, in which rapid dynamic imaging and high-spatial-resolution imaging are performed on different days, this interleaved protocol enables dynamic data and high-spatial-resolution data to be acquired from identical volumes of breast tissue during the same examination and after the same injection, thereby enabling precise spatial registration between the two data sets. Furthermore, the interleaved protocol allows timely acquisition of contrast-enhanced high-resolution data at a point when the dynamic data are changing relatively slowly. The hiatus in the dynamic data has little effect on the pharmacokinetic values because critical wash-in and washout data preceding and following the hiatus are sufficient for these calculations.

The initial rapid dynamic images were acquired at 32 sagittal section locations encompassing an entire single breast by using a T1-weighted 3D spoiled gradient-echo spiral pulse sequence (^,^,^,^,^,^,^,^,Fig 1a) (^,14). The parameters of this pulse sequence were 38/12.3 (repetition time msec/echo time msec), 40° flip angle, 20 spiral interleaves, effective matrix of 188 × 188 interpolated to 256 × 256, and 20-cm field of view. A water-selective spectral-spatial excitation prevented significant signal from fat. An on-resonance 1-2-1 binomial magnetization transfer pulse reduced signal from background fibroglandular tissue and muscle. Partial-Fourier phase encoding in the section direction (z) provided images at 32 locations within 10.64 seconds (10 positive k_z lines, four negative k_z lines, zero filling to a total of 32 k_z lines, resultant section thickness of 4.5–6 mm, section spacing of 3–4 mm). The 3D spiral acquisition was repeated 20 times for a dynamic series of 213 seconds (∼3.5 minutes) duration. Forty seconds after the start of the dynamic series, 0.1 mmol/kg gadopentetate dimeglumine (Magnevist; Berlex Laboratories, Wayne, NJ) was injected at 2.5 mL/sec via an antecubital vein. Flushing with 20 mL of saline solution immediately followed the contrast material injection. In patients with a typical circulation time, this process resulted in about 1.5 minutes of baseline data followed by 2 minutes of enhancement data. Because of the very rapid acquisition time, steady-state magnetization was not achieved until the second dynamic acquisition. Thus, data from the first acquisition were discarded, yielding 19 time points for analysis.

Sagittal high-spatial-resolution imaging was performed immediately after the dynamic wash-in series by using a 3D water-selective spectral-spatial spoiled gradient-echo acquisition with an on-resonance 1-2-1 binomial magnetization transfer pulse (3DSSMT) (^,^,^,^,^,^,^,^,Fig 1b). The parameters of this pulse sequence were 33/9, 50° flip angle, matrix of 512 × 192 interpolated to 512 × 512, 64 sections, 1.5–2-mm section thickness, 20-cm field of view, and acquisition time of 6 minutes 31 seconds (^,15). The 3DSSMT images were acquired at approximately twice the in-plane resolution of the dynamic spiral images (1.1 × 1.1-mm pixel vs 0.4 × 1.0-mm pixel) and with one-third the section thickness (1.5–2 mm vs 4.5–6 mm). Centric k-space encoding ensured that image contrast was achieved at the start of the acquisition, approximately 2 minutes after arrival of the contrast agent at the breast. Because of section-direction aliasing artifacts (wraparound), the outer two sections on each side of the 3DSSMT volume were discarded from further analysis, yielding 60 images through the breast.

Immediately after centric 3DSSMT imaging, dynamic spiral imaging was resumed for another 26 acquisitions (277 seconds) to assess for washout of the contrast agent. Again, data from the first acquisition were discarded, yielding 25 time points for analysis.

Choice of Pulse Sequences

For dynamic imaging, the 3D spiral method was preferred over echo-planar imaging (^,4) because magnetic field inhomogeneities result in blurring rather than spatial distortions. Since the pharmacokinetic maps calculated from the dynamic data will be precisely registered on a pixel-by-pixel basis with corresponding high-resolution images, errors due to distortion could potentially be catastrophic by causing high-resolution data to be correlated with the wrong dynamic data. However, blurring is acceptable because the dynamic data are essentially correct even if their spatial resolution is not as high.

The 3D spiral technique also uses a more efficient readout strategy that allows imaging of the whole breast in 10 seconds. Its repetition time of 38 msec is much longer than that of conventional fast 3D Fourier transform sequences, and therefore the signal-to-noise ratio is potentially higher. Furthermore, the dynamic 3D spiral sequence uses a water-selective spectral-spatial excitation pulse to minimize the signal from fat. This pulse is critical because even a small amount of transient motion could cause substantial signal variations at the margin of a lesion due to variable degrees of volume averaging of the tumor and adjacent fat with more typical dynamic imaging methods, in which fat has intrinsically high signal intensity. Such signal variations in the dynamic data would result in spurious pharmacokinetic calculations. The water-selective excitation has the additional benefit that within any given pixel, pharmacokinetic modeling is performed on only the aqueous signal, not the fatty signal; pharmacokinetic modeling on only the aqueous signal is appropriate because breast lesions are composed of aqueous, not fatty, elements.

Pixel-by-Pixel Pharmacokinetic Modeling

The combined wash-in and washout spiral data were linearly interpolated in the section direction to create dynamic data at locations corresponding to each 3DSSMT image. In-plane correlation between the dynamic and high-resolution data sets was achieved by registering each pixel in the 256 × 256 spiral images with the four corresponding pixels in the 512 × 512 3DSSMT images at the same spatial coordinates. Parametric calculations were made by curve fitting the dynamic time-dependent signal intensity data to a two-compartment pharmacokinetic model (^,9,^,10) on a pixel-by-pixel basis (^,^,^,^,^,^,^,^,Fig 1c, 1d). This model assumes that the rate of change of gadopentetate dimeglumine concentration in the extravascular extracellular tissue space (C_e) is proportional to the difference between the gadopentetate dimeglumine concentration in the blood plasma (C_p) and C_e and that gadopentetate dimeglumine does not enter the intracellular tissue space. The proportionality constant was termed k₂₁:

The pharmacokinetic parameter k₂₁, in this case, can be interpreted as a measure of how quickly C_p and C_e equalize; it is assumed to be proportional to the vascular permeability–surface area product of the vascular bed and is modulated by blood flow.

It was also assumed that C_p decreases exponentially from an initial gadopentetate dimeglumine concentration C₀ immediately after bolus injection and instantaneous mixing. The exponential decay constant was termed k_el:

Substituting this expression into Equation (1) and solving the differential equation results in the following expression for C_e:

For a repetition time much smaller than T1, it is assumed that the signal intensity over baseline (S/S₀) is equal to 1 plus the product of the extravascular extracellular gadopentetate dimeglumine concentration (C_e), the volume fraction of extravascular extracellular space in the tissue (F), the relaxivity of gadopentetate dimeglumine (R), and the native T1 of the tissue (T1₀) (^,9):

Therefore, the final equation used for curve fitting is as follows:

where A = C₀FRT1₀ (the maximum amplitude of enhancement above baseline) and t₀ is the time of initial enhancement. The parameters k₂₁, k_el, A, and t₀ were all variables in the curve-fitting procedure. The inclusion of k_el as a free parameter rather than an independently measured quantity could, in theory, lead to suboptimal estimation of other parameters, particularly if the underlying signal intensity data have unusual features (eg, the superimposition of slow and rapid wash in, recirculation phenomena, third-space distribution, or multiexponential elimination from the plasma) that cause the plasma concentration to deviate from the assumed monoexponential decay. Inaccuracies could also occur if the signal from the plasma compartment makes a substantial contribution to the overall signal from the lesion. Under these circumstances, the curve-fitting procedures described later in this section could potentially fit the model to these features, leading to incorrect k_el values, even negative k_el values, and correspondingly less accurate k₂₁ values. To minimize these potential sources of error, the fitted curves were constrained to return to baseline during the fitting procedure. This constraint limited the range of k_el to positive values, which are more physiologically plausible. More complex models have been proposed to address some of these concerns. A comparison of the theoretical assumptions and features of the simplified two-compartment model versus other pharmacokinetic models is the subject of a recent review (^,16).

Despite these potential limitations, this simplified pharmacokinetic model, and in particular the k₂₁ parameter, has been found to allow differentiation of invasive breast carcinomas from other lesions with excellent diagnostic performance (^,7). It had two other distinct advantages for dynamic breast MR imaging data. First, it is not necessary to provide a gadopentetate dimeglumine arterial input function, since a monoexponential decrease is assumed after bolus injection and the parameter t₀ accounts for the varying times until initial enhancement. Thus, the need for high-quality imaging of the large arteries outside the breast is avoided, enabling small-field-of-view sagittal plane dynamic imaging to be performed with the breast coil. This type of imaging, in turn, maximizes the image quality, efficiency, and spatial resolution of the acquisition. Second, a native T1 map is not required to calculate the k₂₁ values from the signal intensity curves because the parameter A accounts for T1₀.

Curve fitting was performed by using gradient-expansion algorithms (Interactive Data Language [IDL] version 3.6; Research Systems, Boulder, Colo) on a workstation (UltraSPARC 1; Sun Microsystems, Mountain View, Calif). To reduce the calculation time from 24 hours to 6–8 hours per case, pixels less than 3 standard deviations above the mean signal intensity of air in the 3DSSMT images were omitted from the calculations. This threshold ensured that the pharmacokinetic calculation time was devoted to pixels representing potentially enhancing tissue. Because of the highly water-selective nature of the 3DSSMT sequence, some pixels of pure adipose tissue also fell below this threshold. However, since breast cancers are composed of aqueous tissue, the absence of pharmacokinetic data for these purely adipose pixels was inconsequential.

Parametric Map Presentation

The pharmacokinetic information was integrated into the 3DSSMT images by modulating the hue of each 3DSSMT pixel as a function of the corresponding k₂₁ value according to the color scheme in ^,^,^,^,^,^,^,^,Figure 1d. Pixels lacking a k₂₁ value were assigned a neutral white hue. The intensity of each pixel was then made proportional to that in the original 3DSSMT image. This method of parametric map presentation was termed intensity-modulated parametric map display (^,^,^,^,^,^,^,^,Fig 1e). The selection of the color scheme for the k₂₁ map used by the intensity-modulated display method is critical. All colors (when fully saturated) must appear similar in intensity to each other and to white so that they do not confound the intensity information provided by the 3DSSMT data. In our ex-perience, colors ranging from cyan to green to yellow were suitable when reviewed on a color computer monitor. Other colors such as red, purple, and blue gave false impressions of lower signal intensity, even when fully saturated.

The success of the intensity-modulated parametric map display method relies on the coregis-tration of the contrast-enhanced high-resolution images and the dynamic data from which the pharmacokinetic parameters are calculated. Since the map display method highlights parametric information corresponding to high signal intensity, contrast enhancement and fat suppression in the 3DSSMT images are required to make the tumor more conspicuous.

In contrast, in the traditional overlaid map display (^,^,^,^,^,^,^,^,Fig 1f, 1g), pixels in the gray-scale anatomic image are replaced by fully saturated color-coded pixels representing only parametric information if they have a parametric value above a certain thresh-old. As a result, morphologic information is sacrificed for physiologic information. Although this trade-off is adequate for functional MR imaging of the brain, in which the areas of activation are small and morphologic assessment of the underlying tissue anatomy is not needed, it is less suitable for breast MR imaging, in which both pharmacokinetic parameters of contrast enhancement and lesion morphology must be assessed over the entire breast.

Radiologist Evaluation of Intensity-modulated Parametric Maps

The intensity-modulated parametric map display method was compared with standard overlaid map displays in a pilot study of radiologist preferences. Data from seven patients who had previously undergone the described protocol of gadolinium-enhanced dynamic and high-resolution MR imaging were processed and reviewed. All patients had given written informed consent, as approved by the human subjects panel of our institution. Subsequent surgical biopsy revealed eight lesions in these seven patients (five infiltrating ductal carcinomas, two infiltrating lobular carcinomas, one fibroadenoma). These eight lesions made up the study sample.

The eight lesions were independently evaluated by five radiologists (B.L.D., R.J.H., D.M.I., R.L.B., S.G.H.) using both the intensity-modulated display method and the traditional overlay display method. For each lesion, the radiologists rated each method on a scale of 1 to 10 (1 = worst, 10 = best) according to the 12 criteria listed in the ^,Table: ability to display morphologic details including shape, margins, internal features, and intensity of enhancement; ability to allow correlation of parametric values with lesion morphology; ability to display heterogeneity and distribution of parametric values within the lesion; ability to allow identification of lesion location within the breast; ability to allow characterization of features associated with malignancy, such as breast architectural distortion and skin or chest wall invasion; ability to demonstrate normal breast features such as blood vessels and enhancing abnormalities distinct from the lesion; and ability to suppress noise and spurious parametric data.

Rating differences between the two methods were analyzed for reader-independent significance by using a nested F test for nested or hierarchical classification (^,17) for each of the 12 criteria. An F value of 7.71 or higher indicated a statistically significant difference between the display methods (P < .05).

The results of the radiologists' assessment are shown in the ^,Table. The ratings showed a statistically significant preference for the intensity-modulated parametric map display method over the overlaid parametric display method for 10 of the 12 criteria. These benefits can be grouped into three areas: simultaneous visualization of pharmacokinetic information and lesion morphology, assessment of lesion extent, and suppression of noise in the parametric maps.

Simultaneous Visualization of Pharmacokinetics and Morphology

The ability of intensity-modulated parametric map display to allow simultaneous visualization of pharmacokinetic and morphologic information is illustrated in ^,^,^,^,^,^,^,^,Figures 1 and ^,^,^,^,^,2. The intensity-modulated parametric display allowed critical morphologic features of a lesion such as its margins to be assessed in concert with its pharmacokinetic features. It is hoped that such combined morphologic-pharmacokinetic assessment will lead to more accurate characterization of breast lesions, fewer false-positive results, and greater specificity (^,18). In addition, such assessment may enable correlation of the morphology and pharmacokinetics of specific foci within a heterogeneous breast lesion (^,^,^,Fig 3). This capability is particularly promising because vascular heterogeneity is a characteristic feature of breast cancer in animal models (^,19). Future studies are needed to determine the incremental specificity that this method adds over existing breast MR imaging techniques, such as the three time-point method (^,19).

Assessment of Lesion Extent

The intensity-modulated parametric map display also facilitated assessment of the extent of a lesion (^,^,^,^,^,Figs 2, ^,^,^,4). High-spatial-resolution details of the margins of a lesion remained visible even when accompanying pharmacokinetic data were acquired at lower resolution. Unlike overlaid mapping methods, which use a threshold to determine which parametric values are displayed, the intensity-modulated display does not require a threshold. Thus, the extent of an enhancing lesion can be determined by inspection of both the shape of the enhancing area and its pharmacokinetics rather than by means of a single predetermined pharmacokinetic threshold, which may obscure the shape of the enhancing lesion on the underlying morphologic image.

Noise Suppression

Spurious parametric values result when the pharmacokinetic curve-fitting algorithm incorrectly converges onto features due to signal noise in the time–signal intensity curve. This effect is especially problematic for pixels when there is little or no enhancement and noise is the dominant feature of the curve. Because the nonenhancing areas usually have much less signal on 3DSSMT images, the intensity-modulated parametric map display substantially deemphasized spurious parametric values due to noise (^,^,^,^,^,^,^,^,Figs 1, ^,^,^,4). This technique was more effective than the standard technique of an overlaid display with a high k₂₁ threshold (^,^,^,^,^,^,^,^,Fig 1). Intensity-modulated parametric map display also preserved depiction of the pharmacokinetic parameters of lesions with low k₂₁ values, which were neglected with overlay methods (^,^,^,^,^,^,^,^,Fig 1g). One limitation is that noisy pharmacokinetic values will not be deemphasized for nonenhancing lesions that have intrinsic high signal intensity on 3DSSMT images, such as cysts, ducts, or hematomas that contain aqueous material with a short T1.

Limitations

Implicit in the intensity-modulated parametric map display method is the assumption that all areas of suspicious enhancement have moderate to high signal intensity on the centric 3DSSMT images. In our experience, this assumption is valid, since even malignancies with early washout of contrast material remain enhanced for several minutes following injection of contrast material.

Patient motion is a potential problem because it can cause misregistration between different portions of the dynamic examination and with the 3DSSMT imaging. Since the pharmacokinetic analysis technique includes no specific means of detecting when significant motion has occurred, incorrect pharmacokinetic map values could occur. The raw data from all cases must be reviewed in a cine or paging fashion to visually exclude significant tissue motion. In our pilot study, the washout portion of the dynamic data set had to be ignored in two cases. In these cases, the pharmacokinetic calculations were performed on the wash-in data alone. In the future, automated motion correction algorithms may reduce misregistration effects.

Conclusions

Intensity-modulated parametric maps efficiently condense high-spatial-resolution and rapid dynamic MR imaging data into a single comprehensive image that is a visually intuitive tool for characterizing the morphology, pharmacokinetics, and extent of breast lesions. In the past, results of breast MR imaging were interpreted by reviewing anatomic images of enhancement without coregistered pharmacokinetic data and then selectively interrogating any suspicious focal regions with region-of-interest analysis of the time–signal intensity dynamic data. This shift from a retrospective interrogation strategy to a prospective review facilitates rapid assessment of both dynamic and morphologic features throughout the entire breast. It may make combined rapid dynamic and high-spatial-resolution breast MR imaging more practical as a screening examination by speeding interpretation and has the potential to highlight unsuspected abnormally enhancing foci with suspicious pharmacokinetic features. Future clinical studies are needed to determine the usefulness of this technique for screening applications (^,Table).

**Figure 1a.** Overview of the intensity-modulated parametric map display method in a patient with a biopsy-proved invasive ductal carcinoma (*IDC*) and a presumed fibroadenoma (FA). **(a)** Spiral MR image from the 18th acquisition of the wash-in dynamic 3D spiral series shows early enhancement of the carcinoma. **(b)** Corresponding high-resolution centric 3DSSMT image shows both the spiculated, rim-enhancing carcinoma and the smoothly marginated, enhancing fibroadenoma. **(c, d)** Pixel-by-pixel analyses of the dynamic time-signal intensity curves are used to generate a map of the k₂₁ pharmacokinetic parameter **(c)** according to the color scheme shown in d. Pixels with parametric values outside this range are assigned the yellow (k₂₁ = 0.04 sec^-1) or cyan (k₂₁ = 0.00 sec^-1) color of the closest maximum or minimum value, respectively. **(e)** The k₂₁ map is incorporated into the 3DSSMT image by using the intensity-modulated map display method. Morphologic features remain well demonstrated, including the spiculated margins and rim enhancement of the carcinoma and the smooth margin of the fibroadenoma. **(f, g)** Standard overlaid map displays obtained with minimum thresholds of k₂₁ = 0.018 sec^-1 and 0.036 sec^-1, respectively, are shown for comparison. At either threshold, the morphologic features of the carcinoma are obscured. At the low threshold **(f)**, significant noise is seen in the pharmacokinetic data from pixels where there is little enhancement, such as the fatty regions of the breast. This noise makes it difficult to identify the fibroadenoma, which has a similar k₂₁ value to that of the noise. In the intensity-modulated display **(e)**, the low signal intensity of the 3DSSMT image in the noisy regions of the k₂₁ map deemphasizes these values, rendering the fibroadenoma much easier to visualize. At the high threshold **(g)**, pharmacokinetic noise is reduced but still remains visible. Although the high threshold allows the fibroadenoma to be identified on the basis of its 3DSSMT gray-scale appearance, all pharmacokinetic data for the fibroadenoma have been lost.

**Figure 1b.** Overview of the intensity-modulated parametric map display method in a patient with a biopsy-proved invasive ductal carcinoma (*IDC*) and a presumed fibroadenoma (FA). **(a)** Spiral MR image from the 18th acquisition of the wash-in dynamic 3D spiral series shows early enhancement of the carcinoma. **(b)** Corresponding high-resolution centric 3DSSMT image shows both the spiculated, rim-enhancing carcinoma and the smoothly marginated, enhancing fibroadenoma. **(c, d)** Pixel-by-pixel analyses of the dynamic time-signal intensity curves are used to generate a map of the k₂₁ pharmacokinetic parameter **(c)** according to the color scheme shown in d. Pixels with parametric values outside this range are assigned the yellow (k₂₁ = 0.04 sec^-1) or cyan (k₂₁ = 0.00 sec^-1) color of the closest maximum or minimum value, respectively. **(e)** The k₂₁ map is incorporated into the 3DSSMT image by using the intensity-modulated map display method. Morphologic features remain well demonstrated, including the spiculated margins and rim enhancement of the carcinoma and the smooth margin of the fibroadenoma. **(f, g)** Standard overlaid map displays obtained with minimum thresholds of k₂₁ = 0.018 sec^-1 and 0.036 sec^-1, respectively, are shown for comparison. At either threshold, the morphologic features of the carcinoma are obscured. At the low threshold **(f)**, significant noise is seen in the pharmacokinetic data from pixels where there is little enhancement, such as the fatty regions of the breast. This noise makes it difficult to identify the fibroadenoma, which has a similar k₂₁ value to that of the noise. In the intensity-modulated display **(e)**, the low signal intensity of the 3DSSMT image in the noisy regions of the k₂₁ map deemphasizes these values, rendering the fibroadenoma much easier to visualize. At the high threshold **(g)**, pharmacokinetic noise is reduced but still remains visible. Although the high threshold allows the fibroadenoma to be identified on the basis of its 3DSSMT gray-scale appearance, all pharmacokinetic data for the fibroadenoma have been lost.

**Figure 1c.** Overview of the intensity-modulated parametric map display method in a patient with a biopsy-proved invasive ductal carcinoma (*IDC*) and a presumed fibroadenoma (FA). **(a)** Spiral MR image from the 18th acquisition of the wash-in dynamic 3D spiral series shows early enhancement of the carcinoma. **(b)** Corresponding high-resolution centric 3DSSMT image shows both the spiculated, rim-enhancing carcinoma and the smoothly marginated, enhancing fibroadenoma. **(c, d)** Pixel-by-pixel analyses of the dynamic time-signal intensity curves are used to generate a map of the k₂₁ pharmacokinetic parameter **(c)** according to the color scheme shown in d. Pixels with parametric values outside this range are assigned the yellow (k₂₁ = 0.04 sec^-1) or cyan (k₂₁ = 0.00 sec^-1) color of the closest maximum or minimum value, respectively. **(e)** The k₂₁ map is incorporated into the 3DSSMT image by using the intensity-modulated map display method. Morphologic features remain well demonstrated, including the spiculated margins and rim enhancement of the carcinoma and the smooth margin of the fibroadenoma. **(f, g)** Standard overlaid map displays obtained with minimum thresholds of k₂₁ = 0.018 sec^-1 and 0.036 sec^-1, respectively, are shown for comparison. At either threshold, the morphologic features of the carcinoma are obscured. At the low threshold **(f)**, significant noise is seen in the pharmacokinetic data from pixels where there is little enhancement, such as the fatty regions of the breast. This noise makes it difficult to identify the fibroadenoma, which has a similar k₂₁ value to that of the noise. In the intensity-modulated display **(e)**, the low signal intensity of the 3DSSMT image in the noisy regions of the k₂₁ map deemphasizes these values, rendering the fibroadenoma much easier to visualize. At the high threshold **(g)**, pharmacokinetic noise is reduced but still remains visible. Although the high threshold allows the fibroadenoma to be identified on the basis of its 3DSSMT gray-scale appearance, all pharmacokinetic data for the fibroadenoma have been lost.

**Figure 1d.** Overview of the intensity-modulated parametric map display method in a patient with a biopsy-proved invasive ductal carcinoma (*IDC*) and a presumed fibroadenoma (FA). **(a)** Spiral MR image from the 18th acquisition of the wash-in dynamic 3D spiral series shows early enhancement of the carcinoma. **(b)** Corresponding high-resolution centric 3DSSMT image shows both the spiculated, rim-enhancing carcinoma and the smoothly marginated, enhancing fibroadenoma. **(c, d)** Pixel-by-pixel analyses of the dynamic time-signal intensity curves are used to generate a map of the k₂₁ pharmacokinetic parameter **(c)** according to the color scheme shown in d. Pixels with parametric values outside this range are assigned the yellow (k₂₁ = 0.04 sec^-1) or cyan (k₂₁ = 0.00 sec^-1) color of the closest maximum or minimum value, respectively. **(e)** The k₂₁ map is incorporated into the 3DSSMT image by using the intensity-modulated map display method. Morphologic features remain well demonstrated, including the spiculated margins and rim enhancement of the carcinoma and the smooth margin of the fibroadenoma. **(f, g)** Standard overlaid map displays obtained with minimum thresholds of k₂₁ = 0.018 sec^-1 and 0.036 sec^-1, respectively, are shown for comparison. At either threshold, the morphologic features of the carcinoma are obscured. At the low threshold **(f)**, significant noise is seen in the pharmacokinetic data from pixels where there is little enhancement, such as the fatty regions of the breast. This noise makes it difficult to identify the fibroadenoma, which has a similar k₂₁ value to that of the noise. In the intensity-modulated display **(e)**, the low signal intensity of the 3DSSMT image in the noisy regions of the k₂₁ map deemphasizes these values, rendering the fibroadenoma much easier to visualize. At the high threshold **(g)**, pharmacokinetic noise is reduced but still remains visible. Although the high threshold allows the fibroadenoma to be identified on the basis of its 3DSSMT gray-scale appearance, all pharmacokinetic data for the fibroadenoma have been lost.

**Figure 1e.** Overview of the intensity-modulated parametric map display method in a patient with a biopsy-proved invasive ductal carcinoma (*IDC*) and a presumed fibroadenoma (FA). **(a)** Spiral MR image from the 18th acquisition of the wash-in dynamic 3D spiral series shows early enhancement of the carcinoma. **(b)** Corresponding high-resolution centric 3DSSMT image shows both the spiculated, rim-enhancing carcinoma and the smoothly marginated, enhancing fibroadenoma. **(c, d)** Pixel-by-pixel analyses of the dynamic time-signal intensity curves are used to generate a map of the k₂₁ pharmacokinetic parameter **(c)** according to the color scheme shown in d. Pixels with parametric values outside this range are assigned the yellow (k₂₁ = 0.04 sec^-1) or cyan (k₂₁ = 0.00 sec^-1) color of the closest maximum or minimum value, respectively. **(e)** The k₂₁ map is incorporated into the 3DSSMT image by using the intensity-modulated map display method. Morphologic features remain well demonstrated, including the spiculated margins and rim enhancement of the carcinoma and the smooth margin of the fibroadenoma. **(f, g)** Standard overlaid map displays obtained with minimum thresholds of k₂₁ = 0.018 sec^-1 and 0.036 sec^-1, respectively, are shown for comparison. At either threshold, the morphologic features of the carcinoma are obscured. At the low threshold **(f)**, significant noise is seen in the pharmacokinetic data from pixels where there is little enhancement, such as the fatty regions of the breast. This noise makes it difficult to identify the fibroadenoma, which has a similar k₂₁ value to that of the noise. In the intensity-modulated display **(e)**, the low signal intensity of the 3DSSMT image in the noisy regions of the k₂₁ map deemphasizes these values, rendering the fibroadenoma much easier to visualize. At the high threshold **(g)**, pharmacokinetic noise is reduced but still remains visible. Although the high threshold allows the fibroadenoma to be identified on the basis of its 3DSSMT gray-scale appearance, all pharmacokinetic data for the fibroadenoma have been lost.

**Figure 1f.** Overview of the intensity-modulated parametric map display method in a patient with a biopsy-proved invasive ductal carcinoma (*IDC*) and a presumed fibroadenoma (FA). **(a)** Spiral MR image from the 18th acquisition of the wash-in dynamic 3D spiral series shows early enhancement of the carcinoma. **(b)** Corresponding high-resolution centric 3DSSMT image shows both the spiculated, rim-enhancing carcinoma and the smoothly marginated, enhancing fibroadenoma. **(c, d)** Pixel-by-pixel analyses of the dynamic time-signal intensity curves are used to generate a map of the k₂₁ pharmacokinetic parameter **(c)** according to the color scheme shown in d. Pixels with parametric values outside this range are assigned the yellow (k₂₁ = 0.04 sec^-1) or cyan (k₂₁ = 0.00 sec^-1) color of the closest maximum or minimum value, respectively. **(e)** The k₂₁ map is incorporated into the 3DSSMT image by using the intensity-modulated map display method. Morphologic features remain well demonstrated, including the spiculated margins and rim enhancement of the carcinoma and the smooth margin of the fibroadenoma. **(f, g)** Standard overlaid map displays obtained with minimum thresholds of k₂₁ = 0.018 sec^-1 and 0.036 sec^-1, respectively, are shown for comparison. At either threshold, the morphologic features of the carcinoma are obscured. At the low threshold **(f)**, significant noise is seen in the pharmacokinetic data from pixels where there is little enhancement, such as the fatty regions of the breast. This noise makes it difficult to identify the fibroadenoma, which has a similar k₂₁ value to that of the noise. In the intensity-modulated display **(e)**, the low signal intensity of the 3DSSMT image in the noisy regions of the k₂₁ map deemphasizes these values, rendering the fibroadenoma much easier to visualize. At the high threshold **(g)**, pharmacokinetic noise is reduced but still remains visible. Although the high threshold allows the fibroadenoma to be identified on the basis of its 3DSSMT gray-scale appearance, all pharmacokinetic data for the fibroadenoma have been lost.

**Figure 1g.** Overview of the intensity-modulated parametric map display method in a patient with a biopsy-proved invasive ductal carcinoma (*IDC*) and a presumed fibroadenoma (FA). **(a)** Spiral MR image from the 18th acquisition of the wash-in dynamic 3D spiral series shows early enhancement of the carcinoma. **(b)** Corresponding high-resolution centric 3DSSMT image shows both the spiculated, rim-enhancing carcinoma and the smoothly marginated, enhancing fibroadenoma. **(c, d)** Pixel-by-pixel analyses of the dynamic time-signal intensity curves are used to generate a map of the k₂₁ pharmacokinetic parameter **(c)** according to the color scheme shown in d. Pixels with parametric values outside this range are assigned the yellow (k₂₁ = 0.04 sec^-1) or cyan (k₂₁ = 0.00 sec^-1) color of the closest maximum or minimum value, respectively. **(e)** The k₂₁ map is incorporated into the 3DSSMT image by using the intensity-modulated map display method. Morphologic features remain well demonstrated, including the spiculated margins and rim enhancement of the carcinoma and the smooth margin of the fibroadenoma. **(f, g)** Standard overlaid map displays obtained with minimum thresholds of k₂₁ = 0.018 sec^-1 and 0.036 sec^-1, respectively, are shown for comparison. At either threshold, the morphologic features of the carcinoma are obscured. At the low threshold **(f)**, significant noise is seen in the pharmacokinetic data from pixels where there is little enhancement, such as the fatty regions of the breast. This noise makes it difficult to identify the fibroadenoma, which has a similar k₂₁ value to that of the noise. In the intensity-modulated display **(e)**, the low signal intensity of the 3DSSMT image in the noisy regions of the k₂₁ map deemphasizes these values, rendering the fibroadenoma much easier to visualize. At the high threshold **(g)**, pharmacokinetic noise is reduced but still remains visible. Although the high threshold allows the fibroadenoma to be identified on the basis of its 3DSSMT gray-scale appearance, all pharmacokinetic data for the fibroadenoma have been lost.

**Figure 2a.** Simultaneous visualization of pharmacokinetic and morphologic information in a patient with an invasive ductal carcinoma (*IDC*) and a fibroadenoma (FA). Intensity-modulated map displays **(a, c)** and standard overlays of thresholded pharmacokinetic maps **(b, d)** show the intense yellow color of the carcinoma and the more blue-green color of the fibroadenoma, which indicate a suspiciously high k₂₁ value for the carcinoma and a normal k₂₁ value for the fibroadenoma. However, the intensity-modulated display more clearly reflects the irregular shape, spiculated margins, and heterogeneous internal features of the carcinoma **(a)** as well as the round shape, smooth margins, and homogeneous internal features of the fibroadenoma **(c)**. The intensity-modulated parametric display also correctly shows that the carcinoma does not invade the chest wall **(a)**. Chest wall invasion could not be excluded with the standard overlay display method because the anatomy of the tumor and chest wall was partially obscured.

**Figure 2b.** Simultaneous visualization of pharmacokinetic and morphologic information in a patient with an invasive ductal carcinoma (*IDC*) and a fibroadenoma (FA). Intensity-modulated map displays **(a, c)** and standard overlays of thresholded pharmacokinetic maps **(b, d)** show the intense yellow color of the carcinoma and the more blue-green color of the fibroadenoma, which indicate a suspiciously high k₂₁ value for the carcinoma and a normal k₂₁ value for the fibroadenoma. However, the intensity-modulated display more clearly reflects the irregular shape, spiculated margins, and heterogeneous internal features of the carcinoma **(a)** as well as the round shape, smooth margins, and homogeneous internal features of the fibroadenoma **(c)**. The intensity-modulated parametric display also correctly shows that the carcinoma does not invade the chest wall **(a)**. Chest wall invasion could not be excluded with the standard overlay display method because the anatomy of the tumor and chest wall was partially obscured.

**Figure 2c.** Simultaneous visualization of pharmacokinetic and morphologic information in a patient with an invasive ductal carcinoma (*IDC*) and a fibroadenoma (FA). Intensity-modulated map displays **(a, c)** and standard overlays of thresholded pharmacokinetic maps **(b, d)** show the intense yellow color of the carcinoma and the more blue-green color of the fibroadenoma, which indicate a suspiciously high k₂₁ value for the carcinoma and a normal k₂₁ value for the fibroadenoma. However, the intensity-modulated display more clearly reflects the irregular shape, spiculated margins, and heterogeneous internal features of the carcinoma **(a)** as well as the round shape, smooth margins, and homogeneous internal features of the fibroadenoma **(c)**. The intensity-modulated parametric display also correctly shows that the carcinoma does not invade the chest wall **(a)**. Chest wall invasion could not be excluded with the standard overlay display method because the anatomy of the tumor and chest wall was partially obscured.

**Figure 2d.** Simultaneous visualization of pharmacokinetic and morphologic information in a patient with an invasive ductal carcinoma (*IDC*) and a fibroadenoma (FA). Intensity-modulated map displays **(a, c)** and standard overlays of thresholded pharmacokinetic maps **(b, d)** show the intense yellow color of the carcinoma and the more blue-green color of the fibroadenoma, which indicate a suspiciously high k₂₁ value for the carcinoma and a normal k₂₁ value for the fibroadenoma. However, the intensity-modulated display more clearly reflects the irregular shape, spiculated margins, and heterogeneous internal features of the carcinoma **(a)** as well as the round shape, smooth margins, and homogeneous internal features of the fibroadenoma **(c)**. The intensity-modulated parametric display also correctly shows that the carcinoma does not invade the chest wall **(a)**. Chest wall invasion could not be excluded with the standard overlay display method because the anatomy of the tumor and chest wall was partially obscured.

**Figure 3a.** Assessment of pharmacokinetic heterogeneity of an invasive micropapillary carcinoma (*IMPC*). Intensity-modulated parametric map display **(a)** and standard overlay of a thresholded pharmacokinetic map **(b)** show high k₂₁ values (yellow) in the inferior aspect of the tumor and lower k₂₁ values (blue and green) in the superior aspect of the tumor. These k₂₁ values are more fully depicted with the intensity-modulated parametric map display **(a)**. Fine spiculations of the tumor margin also remain visible with this display method.

**Figure 3b.** Assessment of pharmacokinetic heterogeneity of an invasive micropapillary carcinoma (*IMPC*). Intensity-modulated parametric map display **(a)** and standard overlay of a thresholded pharmacokinetic map **(b)** show high k₂₁ values (yellow) in the inferior aspect of the tumor and lower k₂₁ values (blue and green) in the superior aspect of the tumor. These k₂₁ values are more fully depicted with the intensity-modulated parametric map display **(a)**. Fine spiculations of the tumor margin also remain visible with this display method.

**Figure 4a.** Assessment of the extent of an infiltrating lobular carcinoma. MR images displayed as an intensity-modulated map **(a)** and an overlaid map **(b)** show an infiltrating lobular carcinoma (*ILC*) with multiple satellite lesions. The tumor satellites are more conspicuous with the intensity-modulated display **(a)**. With the standard overlay method **(b)**, the morphologic details are obscured and true enhancing satellites are difficult to distinguish from spurious foci of high k₂₁ value caused by noise in the pharmacokinetic data.

**Figure 4b.** Assessment of the extent of an infiltrating lobular carcinoma. MR images displayed as an intensity-modulated map **(a)** and an overlaid map **(b)** show an infiltrating lobular carcinoma (*ILC*) with multiple satellite lesions. The tumor satellites are more conspicuous with the intensity-modulated display **(a)**. With the standard overlay method **(b)**, the morphologic details are obscured and true enhancing satellites are difficult to distinguish from spurious foci of high k₂₁ value caused by noise in the pharmacokinetic data.

**Results of Radiologist Evaluation of Display Methods**
Evaluation Criteria	Ratings for Intensity-modulated Display*	Ratings for Overlaid Map Display*	F Value	P < .05
*Values are means ± standard deviations; n = 40 (eight cases reviewed by five readers).
Shape	8.97 ± 1.07	3.77 ± 2.25	27.7	Yes
Margins	8.67 ± 1.40	2.82 ± 1.94	49.9	Yes
Internal features	8.57 ± 1.29	2.40 ± 1.97	50.6	Yes
Intensity of enhancement	9.10 ± 1.15	3.42 ± 2.40	24.5	Yes
Parametric-morphologic correlation	9.80 ± 0.51	7.32 ± 2.32	7.36	No
Parametric heterogeneity	8.27 ± 1.50	5.05 ± 2.93	5.32	No
Location	9.02 ± 1.00	3.15 ± 1.99	87.1	Yes
Architecture	8.96 ± 0.99	2.07 ± 1.36	95.0	Yes
Skin or chest wall invasion	9.10 ± 0.92	3.26 ± 1.93	23.8	Yes
Normal anatomy	9.52 ± 0.71	2.15 ± 2.20	229	Yes
Other enhancing lesions	8.50 ± 1.29	2.47 ± 1.67	33.1	Yes
Noise suppression	9.20 ± 0.79	2.05 ± 1.50	229	Yes

Abbreviations: 3D = three-dimensional, 3DSSMT = 3D spectral-spatial excitation magnetization transfer

References

1 Heywang SH, Wolf A, Pruss E, Hilbertz T, Eiermann W, Permanetter W. MR imaging of the breast with Gd-DTPA: use and limitations. Radiology 1989; 171:95-103. Link, Google Scholar
2 Kaiser WA, Zeitler E. MR imaging of the breast: fast imaging sequences with and without Gd-DTPA—preliminary observations. Radiology 1989; 170:681-686. Link, Google Scholar
3 Boetes C, Barentsz JO, Mus RD, et al. MR characterization of suspicious breast lesions with a gad-olinium-enhanced TurboFLASH subtraction technique. Radiology 1994; 193:777-781. Link, Google Scholar
4 Hulka CA, Smith BL, Sgroi DC, et al. Benign and malignant breast lesions: differentiation with echo-planar MR imaging. Radiology 1995; 197:33-38. Link, Google Scholar
5 Orel SG, Schnall MD, Powell CM, et al. Staging of suspected breast cancer: effect of MR imaging and MR-guided biopsy. Radiology 1995; 196:115-122. Link, Google Scholar
6 Kerslake RW, Carleton PJ, Fox JN, et al. Dynamic gradient-echo and fat-suppressed spin-echo contrast-enhanced MRI of the breast. Clin Radiol 1995; 50:440-454. Crossref, Medline, Google Scholar
7 Daniel BL, Yen YF, Glover GH, et al. Breast disease: dynamic spiral MR imaging. Radiology 1998; 209:499-509. Link, Google Scholar
8 Kuhl CK, Mielcareck P, Klaschik S, et al. Dynamic breast MR imaging: are signal intensity time course data useful for differential diagnosis of enhancing lesions? Radiology 1999; 211:101-110. Link, Google Scholar
9 Hoffmann U, Brix G, Knopp MV, Hess T, Lorenz WJ. Pharmacokinetic mapping of the breast: a new method for dynamic MR mammography. Magn Reson Med 1995; 33:506-514. Crossref, Medline, Google Scholar
10 Parker GJ, Suckling J, Tanner SF, et al. Probing tumor microvascularity by measurement, analysis, and display of contrast agent uptake kinetics. J Magn Reson Imaging 1997; 7:564-574. Crossref, Medline, Google Scholar
11 Parker GJ, Suckling J, Tanner SF, Padhani AR, Husband JE, Leach MO. MRIW: parametric analysis software for contrast-enhanced dynamic MR imaging in cancer. RadioGraphics 1998; 18:497-506. Link, Google Scholar
12 Nunes LW, Schnall MD, Siegelman ES, et al. Diagnostic performance characteristics of architectural features revealed by high spatial-resolution MR imaging of the breast. AJR Am J Roentgenol 1997; 169:409-415. Crossref, Medline, Google Scholar
13 Orel SG, Schnall MD, LiVolsi VA, Troupin RH. Suspicious breast lesions: MR imaging with radiologic-pathologic correlation. Radiology 1994; 190:485-493. Link, Google Scholar
14 Yen YF, Han K, Daniel BL, Heiss S, et al. Dynamic breast MRI with spiral trajectories: 3D versus 2D. J Magn Reson Imaging 2000; 11:351-359. Crossref, Medline, Google Scholar
15 Leong CS, Daniel BL, Herfkens RJ, et al. Characterization of breast lesion morphology with delayed 3DSSMT: an adjunct to dynamic breast MRI. J Magn Reson Imaging 2000; 11:87-96. Crossref, Medline, Google Scholar
16 Tofts PS, Brix G, Buckley DL, et al. Estimating kinetic parameters from dynamic contrast-enhanced T(1)-weighted MRI of a diffusable tracer: standardized quantities and symbols. J Magn Reson Imaging 1999; 10:223-232. Crossref, Medline, Google Scholar
17 Kanji GK. 100 statistical tests London, England: Sage Publications, 1993; 128-129. Google Scholar
18 Orel SG. Differentiating benign from malignant enhancing lesions identified at MR imaging of the breast: are time–signal intensity curves an accurate predictor? Radiology 1999; 211:5-7. Link, Google Scholar
19 Degani H, Gusis V, Weinstein D, Fields S, Strano S. Mapping pathophysiological features of breast tumors by MRI at high spatial resolution. Nat Med 1997; 3:780-782. Crossref, Medline, Google Scholar

Article History

Published in print: Jan 2001

Vol. 21, No. 1

Metrics

Downloaded 841 times

Altmetric Score