B₁ transmission-field inhomogeneity and enhancement ratio errors in dynamic contrast-enhanced MRI (DCE-MRI) of the breast at 3T

DYNAMIC CONTRAST-ENHANCED breast MRI (DCE-MRI) is considered to be the most sensitive imaging technique in the diagnosis of breast cancer (1, 2). The technique can also be used to characterize breast tissue by monitoring T₁-weighted signal enhancement after a bolus administration of an intravenous low molecular-weight paramagnetic contrast agent (CA). Breast cancers have abnormal microvascular properties because of their increased vascular capillary density and permeability so that they show increased signal enhancement compared to normal tissue (3, 4).

Generally, two quantitative diagnostic approaches are used to evaluate signal enhancement in DCE-MRI, one of which is based on pharmacokinetic modeling (5, 6), the other is enhancement curve analysis (7, 8). For routine clinical applications, the latter technique is normally used. In this approach lesions are characterized by measuring the signal at the regions of suspected lesion before (S_PRE) and after (S_POST) the arrival of CA in the tissue. This information is then used to produce time-intensity curves and to calculate signal enhancement indices. A common index used for this purpose is enhancement ratio (ER), where:

(1)

S_POST is often calculated at 90 seconds postinfusion and also at the maximum observed focal value (9).

In the United Kingdom, breast DCE-MRI is typically performed at 1.5T using a 3D T₁-weighted spoiled gradient echo sequence (e.g., flip angle [α] = 35°, repetition time [TR] = 10–16 msec, echo time [TE] = 4.2 msec [water and fat signals in-phase], and imaging time per repetition = 90 seconds) (10). 3T clinical MRI machines have been introduced. The main advantage of a higher-field MRI scanner is the increase in signal-to-noise ratio (SNR). This increase can be traded off to improve the spatial and/or temporal resolution of a dynamic imaging study. The improvement in morphological detail and temporal resolution can be manipulated to increase both the sensitivity and specificity of DCE-MRI. This may lead to an increase in the accuracy of breast MRI cancer diagnosis and reduce the number of unnecessary biopsies (11). Furthermore, acquisition of functional information at 3T with smaller voxels may improve the reproducibility of analysis techniques.

However, a significant drawback at high field is the increase in B₁ transmission-field inhomogeneity across the field-of-view (FOV), which is particularly marked in breast imaging (12). This problem is caused by the off-center positioning of the patient's torso within the transmitting whole-body radiofrequency (RF) birdcage coil giving rise to unequal loading effects, and by a standing RF wave effect accentuated by the short RF quarter-wavelength at 3T. Both of these effects are more pronounced at higher field strengths (13, 14).

Kuhl et al (12) have shown that the signal enhancement of breast lesions is lower at 3T relative to the enhancement measured at 1.5T. A significant inhomogeneity of the B₁-field across the two breasts was observed (a right-to-left difference of a factor of 2 was reported). They associate the resulting errors in flip angle (due to the B₁ inhomogeneity) to a reduced T₁ weighting and hence to reduced signal enhancement. Kuhl et al (11, 12) suggest that in locations of very low B₁-field there would be no enhancement observed at all. As such enhancement measurements of breast lesions obtained at 3T may be insufficient to establish a useful diagnosis for some patients and result in malignant lesions being overlooked. This issue is a significant setback to the use of clinical breast MRI at 3T.

The aim of this study was to quantify B₁ transmission-field inhomogeneity in breast imaging of normal volunteers at 3T using a 3D T₁-weighted spoiled gradient echo pulse sequence and to assess the resulting errors in ERs in DCE-MRI studies. We employed this sequence because it is commonly used for breast DCE-MRI in the United Kingdom (10). In this work, ERs in the presence of B₁-field inhomogeneity were assessed using computer simulation and experiments on T₁ gel phantoms.

MATERIALS AND METHODS

Simulation of B₁ Transmission-Field Inhomogeneity and Enhancement Ratio

A set of gel phantoms were developed to simulate pre- and postcontrast breast tissue T₁ values. The phantoms were constructed using methods proposed by Walker et al (15) and Waiter and Foster (16). T₁ values were measured using the 2D inversion-recovery (IR) sequence (inversion times (TI) = 50, 200, 400, 800, and 1500 msec; TR = 7000 msec; TE = 17 msec; matrix size = 256 × 256; FOV = 160 × 160 mm; and slice thickness = 4 mm) using an Achieva X series 3T scanner and quadrature transmit-receive birdcage head coil (Philips Healthcare, Best, the Netherlands). T₁ values were calculated by fitting Eq. [ 2] for T₁ and S₀ to plots of signal intensity (S) vs. TI obtained using a very long TR (TR > 5T₁).

(2)

where S₀ is the maximum signal obtainable from the system.

To simulate the effect of B₁ transmission-field inhomogeneity on ER, an adjustable and uniform α across a FOV was required, so that all gel phantoms were subject to the same condition. Hence, 3D T₁-weighted spoiled gradient echo imaging (α = 35°, TR/TE = 10/2.3 msec, matrix size = 256 × 256, FOV = 200 × 200 mm, slice thickness = 2 mm, and number of slices = 30) was performed using a quadrature transmit-receive head coil, which produces relatively the most homogeneous B₁ transmission-field across an FOV. The nominal α of 35° was chosen because it is a commonly used flip angle in clinical DCE-MRI in the United Kingdom for the range of TR used (10) and it produces a good linear contrast response and sufficient SNR over the range of T₁ values found in breast tissues.

To simulate the effect of B₁-field inhomogeneity, α was changed to 17°, 26°, 44°, and 53°, which represent 49%, 74%, 126%, and 151% of the desired B₁. These values were chosen to cover a wide range of B₁ inhomogeneity. The lower limit corresponds with the minimum value expected for this scanner and field strength (12). Assuming that the native T₁ (T₁₀) of ductal tissue is approximately 1300 msec at 3T (17), which matches the T₁ value of one of our gel phantoms, the ER was calculated for each phantom at each α using Eq. [ 1]. By assuming that the gel with the longest T₁ represents the precontrast tissue and that the postcontrast tissue is represented by of each of the shorter T₁ phantoms, the difference in relaxation rate (ΔR1) (ΔR1 = [1/T₁] – [1/T₁₀]), which is assumed to be directly proportional to CA uptake (18), was calculated. This approach was also used by Hoffman et al (19).

The effect of B₁-field inhomogeneity on ER was also evaluated using computer simulation. The signal intensity was calculated using Eq. [ 3] (TE was assumed ≪ T2*) and ER using Eq. 1. Computer simulations were performed using Matlab R2008 (Mathworks, Inc., Natick, MA, USA). The same αs that simulate different levels of B₁-field inhomogeneity as used in experiments were employed.

(3)

where S₀ is maximum signal obtainable from the system.

B₁-field maps (TR₁/TR₂ = 30/150 msec, TE = 2.4 msec, α = 60°, matrix size = 256 × 256, FOV = 200 × 200 mm, slice thickness = 1.5 mm, and number of slices = 20) were performed as a check on the homogeneity of B₁ transmission-field for gel phantom experiment.

RESULTS

In Vivo B₁ Transmission-Field Mapping

An example of a typical B₁ transmission-field map and profile produced on a volunteer is shown in Fig. 1a and b. It can be observed that the B₁-field is reduced to about one-half that of the right breast.

The average B₁-field measured on the left and right breast on 25 volunteers is shown in Fig. 2a. It was found that the B₁ is reduced significantly in the right breast for all the volunteers. The median B₁-field is around 60% of the nominal B₁. For some volunteers the amplitude was reduced by more than 50%. Minimal alteration was observed for the left breast but the amplitude is reduced toward the center of the body as shown in Fig. 1b.

A scatter-plot of the difference between B₁ transmission-field measured on the left and right breast vs. BMI is shown in Fig. 2b. In general, larger reductions of the B₁-field field on the right breast were observed for volunteers with higher BMI. The linear regression line and 95% confidence limits are shown (Pearson's r² = 0.352; P = 0.002). The relationship between B₁ field and cross sectional area of the thorax and breast is shown in Fig. 2c. A linear regression gives Pearson's r² = 0.03 and P = 0.48. There is a larger B₁-field variation between left breast axilla and right breast close to the sternum than across the middle of both of the breasts. This was measured to be 47% (median).

Simulation of B₁ Transmission-Field Inhomogeneity and ER

The T₁ values (mean ± SD) measured using the IR sequence on the gel phantoms developed are: 91 ± 1, 185 ± 3, 378 ± 6, 577 ± 7, 822 ± 8, 988 ± 6, and 1301 ± 11 msec.

B₁-field map and profile measured on gel phantoms are shown in Fig. 3a and b, respectively. By using a transmit-receive head coil, an acceptable homogeneous B₁-field in the area occupied by the gel phantoms was produced. Hence, it provides a sufficient accuracy to simulate B₁-inhomogeneity using different α on gel phantoms.

ER vs. ΔR1 for five simulated B₁-fields is shown in Fig. 4. Here, the gel with the longest T₁ was used as the precontrast baseline (hence T₁₀ = 1301 msec). From computer simulation and experiment, it can be observed that ER significantly decreases with the reduction of B₁-field and increases (with a lesser extent) with B₁-field higher than the optimal value. Figure 4 also shows that the error in ER increases with ΔR1 and thus CA uptake (ΔR1 was assumed to be directly proportional to CA uptake).

DISCUSSION

We have shown that the B₁ transmission-field is always reduced toward the right side of the body when 3D T1-weighted spoiled gradient echo images were obtained using the standard manufacturer settings of our 3T MR scanner (Philips Achieva X series) and patient positioning technique (prone and head-first). Therefore, the right breast will always receive less B₁-field compared to the left breast. This problem will cause reduced intensity on the MR images obtained. It was shown that for some subjects, the B₁-field could be reduced to less than 50% of the expected level. This result supports the observations of Kuhl et al (12) who reported the same right-to-left B₁-field reduction on 2D T1-weighted spoiled gradient echo. The authors found that B₁-field is consistently reduced in the right breast to about one-half of the field in the left breast. However, our results show that the degree of reduction is dependent on the patient's body size—the B₁-field inhomogeneity effect becomes more prominent in subjects with large body size (increased BMI). However, the B₁-field difference is only weakly correlated with BMI (r² = 0.352; P = 0.002) (Fig. 2b) and with the body cross-sectional area (Pearson's r² = 0.03; P = 0.48) (Fig. 2c).

In order to analytically investigate the claim made by Kuhl et al (12) that ER can be reduced by a variable degree caused by B₁-field inhomogeneity at 3T, we simulated the effects of the inhomogeneity on ER using computer simulation and experiment on T₁ gel phantoms. We demonstrate that ER is influenced by the B₁-field inhomogeneity where a reduction in B₁-field reduces ER. It was also found that this effect increases with both B₁-field error and CA uptake (represented by ΔR1).

It has been reported that by using 3D spoiled gradient echo at 1T, the range of ΔR1 for malignant tissue can be up to 7.5 seconds⁻¹ (18). In the presence of a 50% inhomogeneity at this ΔR1, the ER will be reduced by 50%. Higher B₁-field reduction and ΔR1 are possible in clinical breast DCE-MRI at 3T. This might explain the claim made by Kuhl et al (12) that no enhancement may be measured in an area of very low B₁-field where contrast is determined by proton-density, not by T₁ weighting.

Our results also suggest that the use of α larger than 35° (with TR of 10 msec) does not greatly alter the ER. However while the use of a larger α might be useful to reduce the errors in ER, it may not be practical at 3T due to a resulting excessive RF specific absorption rate (SAR). Furthermore, for bilateral tumors and staging of the other breast direct comparison of signal intensity and signal enhancement in the left and right breast is required.

In addition to the simple enhancement index used in this article, other indices have been proposed (18, 20). The variation in enhancement index with B₁ transmit inhomogeneity arises because of the nonlinearity of the signal response curves with R1, which varies with pulse angle, so that errors are expected with any index of enhancement.

In conclusion, 3T MRI is rapidly becoming the new standard in routine clinical imaging. High-field systems offer a great potential in improving the sensitivity and specificity of breast cancer imaging. However, increasing the magnetic field to 3T will increase the B₁ transmission-field inhomogeneity effects, which will produce errors in any method that attempts to quantify the enhancement of a tumor including ER measurement. These errors increase with both B₁-field error and CA uptake. Therefore, further studies must be done to find a practical and effective way to minimize this problem so that 3T MR scanners can be used efficiently in clinical breast DCE-MRI.