European Radiology

, Volume 23, Issue 6, pp 1537–1545

Breast dynamic contrast-enhanced examinations with fat suppression: Are contrast-agent uptake curves affected by magnetic field inhomogeneity?

Authors

    • Royal Marsden NHS Foundation Trust and Institute of Cancer Research
  • M. Borri
    • Royal Marsden NHS Foundation Trust and Institute of Cancer Research
  • E. Scurr
    • Royal Marsden NHS Foundation Trust and Institute of Cancer Research
  • G. Ertas
    • Royal Marsden NHS Foundation Trust and Institute of Cancer Research
  • G. Payne
    • Royal Marsden NHS Foundation Trust and Institute of Cancer Research
  • E. O’Flynn
    • Royal Marsden NHS Foundation Trust and Institute of Cancer Research
  • N. deSouza
    • Royal Marsden NHS Foundation Trust and Institute of Cancer Research
  • M. O. Leach
    • Royal Marsden NHS Foundation Trust and Institute of Cancer Research
Breast

DOI: 10.1007/s00330-012-2735-4

Cite this article as:
Schmidt, M.A., Borri, M., Scurr, E. et al. Eur Radiol (2013) 23: 1537. doi:10.1007/s00330-012-2735-4

Abstract

Objectives

To investigate the effect of magnetic field heterogeneity in breast dynamic contrast-enhanced examinations with fat saturation (DCE-FS).

Methods

The magnetic field was mapped over the breasts in ten patients. DCE-FS was undertaken at 1.5 T with fast spoiled gradient echoes and spectrally selective fat saturation. Signal intensity was calculated for T1 values 25–1,200 ms both on and off resonance, and results were verified with a test object. Clinical examinations were evaluated for the predicted effects of field heterogeneity.

Results

Magnetic field was found to vary by 3.6 ± 1.2 ppm over the central transaxial slice and 5.1 ± 1.5 over the whole breast volume (mean ± standard deviation). Computer simulations predict a reduction in the dynamic range if field heterogeneity leads to unintended water suppression, and distortion to CA uptake curves due to fat suppression failure (for fat containing pixels). A compromise between dynamic range and fat saturation performance is required. Both water suppression and fat suppression failure are apparent in clinical examinations.

Conclusion

Magnetic field heterogeneity is likely to reduce the sensitivity of DCE-FS by distorting the CA uptake curves because of fat suppression failure (for fat containing pixels) and by reducing the dynamic range because of unintended water suppression.

Key Points

Magnetic field heterogeneity is significant in breast magnetic resonance.

Contrast-agent uptake curves are distorted by a non-uniform magnetic field.

Radiologist must be aware of possibility of distortion to interpret uptake curves correctly.

Compromise between fat suppression and dynamic range is required.

Keywords

Magnetic resonance imagingBreastDiagnostic techniques and proceduresNeoplasmsGadolinium compound

Introduction

Breast MRI has become established in diagnostic radiology [15] and has been recommended as an additional examination for young women at high risk of developing breast cancer [610]. The Breast Imaging Reporting and Data System (BI-RADS) classification of lesions requires high spatial resolution imaging for morphological assessment and also dynamic contrast-enhanced (DCE) imaging for analysis of contrast agent (CA) uptake patterns [11]. CA uptake reduces T1, and the classification of CA uptake curves presumes that the image intensity is proportional to the concentration of CA. For this reason DCE protocols must be designed to provide image intensity inversely proportional to T1 over the range of expected T1 values associated with CA uptake.

In early clinical studies of breast DCE, data were acquired with fast 3D T1-weighted pulse sequences based on spoiled gradient echoes in the steady-state with the echo-time (TE) chosen to keep fat and water signals in phase [1217]. In these examinations performed without fat suppression, the linear relationship between signal intensity and CA concentration is ensured by making an appropriate choice of flip angle (FA) and repetition time (TR). Lesion detection relied on subtraction from baseline images to eliminate background signals, which was problematic when subject motion occurred and led to use of DCE pulse sequences with fat suppression [1821]. Recent breast DCE examinations have benefited from technical advances in phased-array and parallel imaging techniques; current systems are expected to achieve higher spatial and temporal resolution than that reported in early studies. The introduction of fat suppression removes the steady-state condition, breaking the data acquisition process into a number of segments, each one containing many echoes. Because the signal intensity cannot be kept constant for all echoes within a segment over the wide range of T1s associated with CA uptake, the relationship between image intensity and CA concentration also becomes dependent on how the echoes in a segment are assigned to k-space positions; differences in the screening sensitivity of different systems and protocols as reported by Jensen et al. [22] are therefore expected. In addition, very short TEs are often used in breast DCE with fat suppression; if fat and water signals are not in phase, failure of fat suppression may lead to complete or partial cancellation of water signals. Spectrally selective adiabatic fast passage (AFP) inversion pulses for fat saturation can be very effective in DCE protocols, but because the effects of this pulse depend on both T1 and resonant frequency, good B0 field homogeneity is required. It is therefore unfortunate that shimming over the breasts is known to be difficult because of their shape [23], resulting in a position-dependent resonant frequency.

In this article we evaluate B0 field homogeneity over the breasts and investigate whether it can affect CA uptake curves in breast DCE examinations with fat saturation (DCE-FS). We examine both the possibility of unintended water suppression and the possibility of fat suppression failure, and consider how the relationship between image intensity and CA concentration, presumed to be approximately linear, depends on the resonant frequency. For this purpose we calculate the image intensity as a function of T1 and resonant frequency for the clinical protocol employed, validate this modelling by imaging test objects, and evaluate clinical breast DCE-FS for the effects predicted by the calculations. These methods allow us to assess whether spatial variations in the static magnetic field B0 can affect the sensitivity of breast DCE-FS.

Materials and methods

Measurement of B0 field homogeneity over breasts

The B0 field homogeneity (after shimming) was evaluated in ten bilateral clinical examinations with approval from the Research Ethics Committee. Phase images were acquired with fat and water in phase and out of phase (TE1/TE2 = 2.3/4.6 ms, TR 100 ms, FA 25°) covering the whole breast volume in 5-mm transaxial slices. Before evaluating field heterogeneity, linear phase drifts along the readout direction were measured and removed, as they may simply be caused by non-centred echoes and eddy-current related distortion of the readout waveforms (and not by variations in magnetic field). Following this processing, the range of magnetic field values over the breast was estimated over the central slice and over the whole breast. The volume of interest includes the anterior portion of the axilla, but excludes posterior regions subject to artefacts either from cardiac motion or parallel imaging.

Calculation of image intensity as a function of T1 on and off resonance

In DCE-FS, fat saturation is achieved by introducing a preparatory inversion pulse between segments of rapid gradient-echo acquisitions, and signal behaviour is usually modelled by solving the Block equations numerically [24, 25]. In the context of breast DCE-FS the results can be generalised as typical sequences employ minimum TR and TE, and the signal intensity depends mainly on the inversion time (chosen to minimise fat signals), the flip angle and the number of echoes in each segment (usually 40–100).

Using the parameters of the clinical breast DCE-FS provided below, the image intensity was calculated as a function of T1 and resonant frequency for T1 values from 25 to 1,200 ms (unenhanced blood) (in-house software, IDL 7.1, Boulder, CO, USA). Signal intensity was calculated for the echoes assigned to the centre of the k-space acquired in the beginning of each segment, disregarding T2* decay, and assuming, for simplicity, that the AFP inversion pulse produced a linear variation in the longitudinal magnetisation within the transition region as shown in Fig. 1.
https://static-content.springer.com/image/art%3A10.1007%2Fs00330-012-2735-4/MediaObjects/330_2012_2735_Fig1_HTML.gif
Fig. 1

Function employed to model the effect of the adiabatic fast passage inversion pulse, with transition regions indicated. Spins with resonant frequency in the adiabatic region are inverted by the AFP inversion pulse. If the resonant frequencies for a given position coincide with the nominal resonant frequencies, fat signals are inverted at the fat resonant frequency (ffat), and water signals remain unaffected at the water resonant frequency (fwater). Vertical lines below show the water resonant frequencies at 25-Hz intervals for which image intensity was calculated as a function of 1/T1 when considering unintended suppression of water signals (Fig. 2). The same frequencies are the fat resonant frequencies for which the image intensity was calculated when considering fat suppression failure (Fig. 3)

Unintended water suppression

Signal intensity was calculated as a function of T1 for frequency offsets from the nominal central water frequency to a frequency 3.5 ppm (225 Hz at 1.5 T) below the nominal water central frequency at 25-Hz intervals. The signal linewidth was estimated from a two-point T2* measurement (TE1/TE2 = 4.6/46 ms, TR = 100 ms, FA = 25° and 5 mm slices, in-house software IDL 7.1, Boulder, CO, USA); T2* values were measured in normal breast parenchyma and in breast tumour pre-contrast on five patients with large lesions who underwent MRI before mastectomy (with approval from the Research Ethics Committee and informed consent). Regions of normal breast parenchyma and breast tumour were outlined with reference to DCE images. The simulation was also performed for a range of FA values (0 to 40°) because FA is expected to affect the dynamic range of the investigation and also because the B1 radiofrequency excitation was shown not to be uniform in breast examinations [26].

Fat suppression failure

Frequency offsets up to 3.5 ppm above the nominal water resonant frequency were considered, also in 25-Hz steps. As the AFP inversion pulse gradually loses its efficacy as the frequency rises, the image intensity was calculated for a voxel containing 50 % fat and 50 % water signals, assuming that the CA uptake in fat is negligible.

In our institution DCE breast examinations are routinely undertaken at 1.5 T (Philips Intera and Achieva, Best, The Netherlands) using spectrally selective fat suppression and a fast segmented spoiled gradient-echo sequence with the shortest possible TE and TR (TE/TR = 1.8/3.9 ms), FA 18°, 60 echoes per segment, parallel imaging factor 2 (direction right/left). Data were acquired for a 3D volume 240 × 216 × 150 mm, pixel size 1.25 × 1.25 × 2.0 mm (overcontiguous slices). The readout gradient direction is anterior/posterior, minimising the presence of cardiac motion artefacts over the breasts. Echoes from each segment of the sequence are associated with k-space positions in the coronal plane, starting from the k-space centre (for effective fat suppression) and progressing towards higher spatial frequencies (either in a radial pattern or in a linear pattern). No preparation pulses are employed, FA is kept constant for all echoes, and no echoes are discarded in the beginning of the segment. Fat suppression uses a hyperbolic secant adiabatic fast passage (AFP) inversion pulse (bandwidth 339 Hz at FWHM and 386 Hz at 99 %), with inversion time 90 ms. Each time frame is acquired in just under 1 min, in agreement with national breast screening guidelines. A single dose of CA (DOTAREM, Guerbet, France) is injected after the acquisition of the first time frame at 3 ml/s (MedRad, USA). Previous quantitative measurements in breast DCE estimate the shortest T1 value in the breast following a single dose injection examination to be of the order of 100 ms [12].

Measurement of image intensity as a function of T1 (on and off resonance)

A test object containing solutions of Gd-DOTA (DOTAREM, Guerbet, France) with concentrations from 0.1 to 10 mM was examined by MRI using the clinical breast DCE protocol in order to verify the signal intensity as a function of T1 at different offsets from the central frequency. In order to achieve that, the central frequency was shifted over a range of 400 Hz in steps of 10 to 25 Hz. Therefore, the AFP intended to suppress fat was used over a range of frequencies from the fat central frequency to the water central frequency. Image intensity was plotted as a function of 1/T1 for different frequency offsets from the nominal water central frequency. T1 was measured using standard inversion recovery techniques for the solutions used.

Analysis of clinical examinations

Breast DCE examinations of ten clinical patients undertaken with the protocol described above over a 2-year period were anonymised and analysed by one physicist for evidence of fat suppression failure or unintended water suppression. This group includes the ten clinical DCE breast examinations for which a field map is available and other examinations of the same patients that occurred within the 2-year period. Non-fat suppressed T1-weighted images were used as a reference. Fat suppression failure was detected by the presence of bright areas known to be fat and also by the presence of dark pixels at the interface between breast fat and breast parenchyma attributed to the presence of enough fat signal to cancel the parenchymal water signal. Water suppression was detected as extensive signal loss in areas where no paucity of signal is present in anatomical images acquired without fat suppression.

Results

Measurement of B0 field homogeneity over breasts

The range of magnetic field values measured over the central transaxial slice in bilateral breast examinations was 3.6 ± 1.2 ppm (mean ± standard deviation); the best field homogeneity achieved was 2.0 ppm and the worst 6.0 ppm. Considering the whole breast, the range of magnetic field values measured was 5.1 ± 1.5 (mean ± standard deviation). The best field homogeneity was 2.8 ppm and the worst 6.9 ppm.

Calculation of image intensity as a function of T1

Unintended water suppression

Considering all patients, T2* values of 50 ± 46 ms and 25 ± 8 ms were found for breast lesions and normal parenchyma, respectively (mean ± standard deviation). The average T2* found in breast lesions was used for unenhanced tissue in computer simulations. Figure 2 shows image intensity as a function of 1/T1 for FAs 18° and 12° (Fig. 2a and b, respectively). On water resonance the image intensity has an approximately linear relationship with 1/T1 over the range of T1 down to 100 ms for the 18° FA employed. The linear relationship is preserved for off-resonance frequencies but with a smaller relative change in image intensity associated with a given CA uptake (i.e., with a reduction in the dynamic range). For the 12° FA the relationship between CA concentration and signal intensity cannot be considered linear for the same range of T1 values.
https://static-content.springer.com/image/art%3A10.1007%2Fs00330-012-2735-4/MediaObjects/330_2012_2735_Fig2_HTML.gif
Fig. 2

Image intensity (arbitrary units) as a function of 1/T1 for the sequence used clinically (a), and for the same sequence, but with a lower flip angle (b). The gray region corresponds to T1 values between 1,200 ms (unenhanced blood) and 100 ms (estimated maximum expected enhancement). T1 = 300 ms (fat) is also indicated. The top curve relates to non-suppressed spins and the bottom curve to suppressed signals inverted by the AFP pulse (both black). The gray curve in the centre is a frequency half-way between fat and water, where fat suppression is partial, and all other curves are the intermediate values indicated in Fig. 1. Curves not visible coincide with either the top or bottom black curves. c Pairs of curves of same colour show non-supressed and suppressed signals, respectively, for a given FA. d Image intensity as a function of FA for breast tissue (T1 = 700 ms) and for fat (suppressed or unsupressed, T1 = 300 ms). FA in the range 10°–12° produces optimal fat suppression, but with a reduced dynamic range

Figure 2c shows that the dynamic range of the examination is reduced when FA is reduced, i.e. the linear relationship between image intensity and 1/T1 is no longer preserved for the required range of T1s expected in breast examinations with a single dose, not even on resonance. However, the separation between the curves for suppressed and unsupressed signals is larger for lower FA. For the fat signals (at T1 approximately 300 ms at 1.5 T [27]), the relative reduction signal intensity associated with fat suppression is greatest with FAs of the order of 10°–12° (Fig. 2d). It is therefore possible to improve fat suppression performance by reducing FA at the price of reducing the dynamic range of the examination. A compromise between fat suppression efficiency and the dynamic range of the examination is therefore implicit in the design of the pulse sequences with integral fat suppression.

Non-suppression of fat signals

With TE = 1.8 ms the chemical shift difference between water and fat causes their signals to be 150° out of phase. Considering again 300 ms as breast fat T1, the linear relationship between signal intensity and 1/T1 of water is not preserved for voxels containing 50 % fat if fat suppression fails (Fig. 3a, 18° FA). This is due to partial cancellation of water signals by the out of phase fat signals. For a 12° FA the uptake curve is more severely distorted (Fig. 3b).
https://static-content.springer.com/image/art%3A10.1007%2Fs00330-012-2735-4/MediaObjects/330_2012_2735_Fig3_HTML.gif
Fig. 3

Image intensity (arbitrary units) as a function of 1/T1 for the clinically used sequence (a) and for the same sequence, but with a lower flip angle (b), considering a voxel containing both water and fat. The gray region corresponds to the range of T1 values of interest (1,200 ms to 100 ms). T1 = 300 ms (fat) is also indicated. Top black curve corresponds to correct fat suppression (i.e. fat and water resonant frequencies are identical to the nominal ones). Fat saturation gets progressively worse in the other curves and fails completely on the curve showing the lowest image intensity. Image intensity is only inversely proportional to T1 for pixels containing both fat and water if fat is suppressed

Test object experiments

Figure 4a shows images of the test object as a function of frequency, demonstrating gradual suppression of the water signals. The experimental results (Fig. 4b and c) are in agreement with the simulation and suggest that the actual FA may be lower than the nominal 18° in this experiment. This is probably associated with poor coil loading with this test object, not comparable to the typical patient load. The small reduction in image intensity for the shortest T1 values is probably due to T2* relaxation, considered to be negligible in the simulations.
https://static-content.springer.com/image/art%3A10.1007%2Fs00330-012-2735-4/MediaObjects/330_2012_2735_Fig4_HTML.gif
Fig. 4

a Test object images acquired for different resonant frequencies (higher to lower) showing increasing suppression of water signal. Image intensity as a function of T1 is shown for FA = 18° (b) and FA = 12° (c). Highest signal intensity curve (top) is associated with non-suppressed signals and the lowest signal intensity curve (bottom) to complete suppression. Measurements shown are 20 Hz apart

Analysis of clinical examinations

Figure 5 shows examples of fat saturation failure. Subtracted phase images show a localised rise in the magnetic field in locations where fat saturation fails. This feature is relatively common but rarely seen over large volumes. In areas of poor fat saturaton a decrease in image intensity is visible at the interface between breast parenchyma/adipose tissue and adipose tissue/skin; interface pixels contain both fat and water, and those signals cancel. Computer simulation predicts distortion of the CA uptake curve for those pixels if the CA causes the shortening of water T1.
https://static-content.springer.com/image/art%3A10.1007%2Fs00330-012-2735-4/MediaObjects/330_2012_2735_Fig5_HTML.gif
Fig. 5

Fat suppression failure (arrows, top row) matches areas of localised rise in the magnetic field (subtracted phase images, TE1/TE2 = 2.3/4.6 ms, bottom row). Phase maps show 180° phase shift between water-dominant and fat-dominant areas. The interface between breast parenchyma/adipose tissue and adipose tissue/skin has lost signal intensity as interface pixels contain both fat and water that are out of phase

Figure 6 shows transitions between areas of good fat saturation and poor fat saturation within the same transaxial section. The image intensity over breast adipose tissue becomes lower as the resonant frequency rises and fat saturation efficiency is reduced. This paradoxical effect is due to the fact breast adipose tissue contains some water [27], which is out of phase with the fat signals. Therefore, the correctly fat-suppressed breast adipose tissue has a visible signal intensity associated with the water content. Incorrectly suppressed fat can cancel out this water component, leading to a lower image intensity (box in Fig. 6a). In Fig. 6b, a vessel surrounded by fat penetrates a region where fat suppression is poor and an area of low signal intensity is apparent at the interface between fat and the vessel, as those pixels contain both fat and blood/vascular endothelium. The vessel appears to become wider as it gets enhanced, but this enlargement is artefactual and only present in areas of poor fat suppression. For those boundary pixels containing both water and fat, a decrease in signal intensity is associated with CA uptake.
https://static-content.springer.com/image/art%3A10.1007%2Fs00330-012-2735-4/MediaObjects/330_2012_2735_Fig6_HTML.gif
Fig. 6

Transitions between areas of good fat suppression and poor fat suppression within the same transaxial section. a Image intensity over breast adipose tissue becomes lower as the resonant frequency rises and fat suppression efficiency is reduced (area indicated within box). This paradoxical effect is due to the fact breast adipose tissue contains some water, which is out of phase with the fat signals. b Vessel surrounded by fat in the region of poor fat suppression: the area of low signal intensity is apparent at the interface between fat and the vessel, comprising pixels containing both fat and blood. The vessel appears to become artefactually wider as it gets enhanced

Figure 7 shows two examples of signal loss over the breast in areas where non-suppressed images show enough signal. In Fig. 7a areas of low signal intensity are likely to be associated with partial fat saturation failure: signals from fat cancel the water signal within the adipose tissue, but vessels and skin remain unaffected. In Fig. 7b, unusual breast positioning created an angled indentation and unintended water saturation is the most likely cause of signal loss as skin signals are saturated.
https://static-content.springer.com/image/art%3A10.1007%2Fs00330-012-2735-4/MediaObjects/330_2012_2735_Fig7_HTML.gif
Fig. 7

Post-contrast images show signal loss over the breast in areas where signal is present in non-suppressed images. a Areas of low signal intensity are likely to be associated with partial fat suppression failure: signals from fat cancels the water signal within the adipose tissue but vessels and skin remain unaffected. b Unusual breast positioning created an angled indentation and unintended water suppression is the most likely cause of signal loss, as skin signals are suppressed

Discussion

The clinical DCE protocol employed automated standard shimming routines, typical of current clinical practice. The field heterogeneity found in clinical breast examinations over the volume comprising the breast and the anterior part of the axilla is significant in comparison with the fat-water chemical shift, and therefore both unintended water suppression and failed fat suppression can be expected to be common in breast DCE examinations. Our results are in broad agreement with those presented by Maril et al. [23].

The characteristics of the transition between areas of correct fat suppression and either failed fat suppression or unintended water suppression depend on spatial variations of the magnetic field, design of the AFP pulse and T2*. Therefore, quantification of fat suppression failure in clinical examinations is a very challenging subject, as its severity varies spatially and does not necessarily compromise the clinical examination. In this work we simply verified the relationship between the field inhomogeneity map and the effects we predicted in breast DCE examinations. Our objective was to clarify technical issues in breast DCE in order to consider the implications of field heterogeneity for the sensitivity of breast DCE examinations.

This article highlights two specific limitations of DCE protocols with fat suppression currently widely used: (1) reduction of the dynamic range in areas of unintended water suppression and (2) distortion of CA uptake curves in areas of poor fat suppression for voxels containing both water and fat. The latter is particularly relevant for the adipose tissue within the breast, known to contain some water [27]. Correct understanding of those limitations is essential for correct interpretation of CA uptake curves and the classification of breast lesions, as they tend to reduce the overall sensitivity of breast DCE.

Field heterogeneity also causes sequences with fat suppression to be unsuitable for any quantitative pharmacokinetic analysis. In quantitative non-fat suppressed DCE the flip angle can be optimised toward improving the accuracy of pharmaco-kinetic parameters [28, 29]; in qualitative DCE-FS a compromise between correct dynamic range and fat suppression performance is implicit in the choice of FA, and the optimal compromise is open to discussion. In examinations where priority is given to ensuring the dynamic range is sufficient (by using higher FAs) the fat suppression failure is more likely to be observed, but a greater level of confidence can be expected for the classification of CA uptake curves for pixels without fat or with fat correctly suppressed. Conversely, if priority is given to ensuring good fat suppression (by using lower FAs), the reduction in dynamic range may cause a prevalence of types I and II curves, as it will be more difficult to detect the peak CA concentration in a type III curve. This is of particular concern at high magnetic fields as the FA may be position dependent [26], causing malignant lesions to be less conspicuous in certain positions. If reliable CA uptake curves are essential, for example in detection and diagnosis of small lesions, it is then advisable to ensure that the FA is high enough to maintain image intensity proportional to CA concentration over the range of T1 values expected despite increasing the probability of fat suppression failure.

In conclusion, spatial variations in magnetic field are a significant issue in breast DCE examinations with fat saturation: CA uptake curves can be distorted by fat suppression failure (for fat containing pixels) and by unintended water suppression. Quantitative pharmacokinetic modelling using DCE examinations with integral fat suppression is therefore ruled out as intrinsically inaccurate. In addition, CA uptake curves must be interpreted taking into account the compromise between dynamic range and fat suppression efficiency defined by the choice of FA in the DCE protocol.

Acknowledgement

This article presents independent research partially funded by the CR-UK and EPSRC Cancer Imaging Centre (C1060/A10334) and carried out at the National Institute for Health Research (NIHR) Royal Marsden Clinical Research Facility in association with the NIHR Biomedical Research Centre. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health.

Copyright information

© European Society of Radiology 2012