Magnetic Resonance Imaging of Breast Cancer and Magnetic Resonance Imaging–Guided Breast Biopsy

Precontrast T1- and T2-Weighted Imaging

Conventional breast MRI begins with T1-weighted images, a so-called localizer or scout, to define the position and anatomy of the breast. T1-weighted images using the signal from the “body coil” rather than the breast coil enable basic evaluation of the axillae, anterior mediastinum, chest wall, and supraclavicular fossa for enlarged regional lymph nodes. Thereafter, the dedicated breast coil should be used to perform all subsequent sequences.

T2-weighted noncontrast fast spin-echo (FSE) images, or other sequences that shows fluid as high signal, are then obtained to characterize the breast and any lesions. T2-weighted scans using an FSE, turbo spin-echo (TSE), or rapid acquisition with relaxation inhibition technique, with effective echo time (TE) 80 to 100 ms and repetition time (TR) at least 3000 ms, produce high-quality images within reasonable scan times of 5 to 6 minutes using 3- to 4-mm-thick slices and 256 × 192 matrix or higher for small FOV sagittal images.

High fat signal on T2-weighted FSE images can be prevented with fat suppression. Proper shimming of the magnetic field and choice of center frequency are essential to ensure adequate fat suppression. Bilateral high-spatial resolution imaging is possible on most scanners because of the development of hardware and software required for bilateral shimming and because coverage is possible with reasonable scan times. However, despite optimal shimming, the breast anatomy near the nipple, axilla, chest wall, and interior breast may cause unavoidable variations in B0 field homogeneity and fat suppression failures, resulting in artifacts. Evaluation of T1-weighted scan allows the radiologist to evaluate shimming of the magnetic field and helps determine whether any abnormally bright areas on fat-suppressed T2-weighted images are caused by poor fat signal suppression.

Dixon imaging (also known as two-point Dixon, three-point Dixon, IDEAL, or Flex) is another approach to water-specific imaging that uses the differential evolution of water and fat spin phase on nonfat-suppressed images obtained with different echo times to generate water- and fat-specific images, but may take longer than fat suppression, unless performed with high acceleration factors.

Diffusion-Weighted Imaging

Diffusion-weighted imaging (DWI) is an unenhanced breast MRI image sequence sensitive to the variability of water molecule mobility in vivo. The amount of movement or diffusion of a water molecule depends on the random molecule mobility called Brownian motion. In vivo, it is also affected by structural or biological factors, such as cell density, membrane integrity, vascular structures, and extracellular matrix composition. Higher signal intensity on DWI indicates increased restriction of diffusion. However, high DWI signal is occasionally seen in regions in which diffusion is not restricted; this is known as T2 shine-through phenomenon. B -values have not been standardized yet, but most investigators use 500 to 1000 with good results. Parallel imaging may reduce distortions in echo planar imaging.

ADC is a measure of the magnitude of diffusion, and is calculated from two or more DWI scans obtained at different b -values. A lower ADC value represents increased restriction of diffusion. The ADC map is more accurate in visualizing the diffusion variability than DWI, without misleading artifacts from the T2 shine-through phenomenon. A reliable estimation of the ADC requires perfect registration of high-quality source images at different b -values without motion artifacts.

Dynamic Contrast-Enhancement Study

Significant variation exists worldwide in methods used for the contrast-enhanced portion of the examination. Most investigators agree that both the time course of enhancement provided by dynamic contrast-enhancement (DCE) scanning and lesion morphology on high-spatial resolution scanning provide distinct and useful information about the risk of malignancy in enhancing lesions. However, commercially available MRI pulse sequences necessitate a compromise between the dynamic and high-spatial resolution approaches. Currently, most scans are done with repeated T1-weighted, fat-saturated, 3D spoiled gradient-echo (SPGR) scans (TR ≤6 ms; flip angle [FA] 10°–15° for T1 weighting). SPGRs provide adequate T1-weighted images with faster scanning than spin-echo sequences, but must be used with optimized flip or tip angles, depending on field strength and repetition time, for proper T1 weighting. Three-dimensional imaging is most often used to avoid interslice gaps, to image faster than 2D multislice imaging, to provide improved reformatted images caused by reconstruction of overlapping slice locations, and to avoid SAR limits at 3 T. Slice thickness and resolution are selected to give maximum resolution and bilateral whole-breast coverage within a 60- to 90-second scan duration. Parallel imaging, fractional k -space (1/2 NEX, etc.), and intermittent/partial fat saturation substantially speed up imaging, allowing much higher resolution within the same scan time. This sequence is repeated as rapidly as possible before, during, and for approximately 5 to 7 minutes after administration of a rapid IV bolus of gadolinium contrast agent. Most investigators use axial or coronal images with a rectangular FOV to maximize efficiency.

According to ACR guidelines ( http://www.acr.org/∼/media/ACR/Documents/PGTS/guidelines/MRI_Breast.pdf ), dynamic images should be obtained at intervals separated by 4 minutes or less. In the authors’ experience, a shorter time interval than 4 minutes is advised to capture the features of the dynamic curves. Contrast material should be administered as an IV bolus dose of 0.1 mmol/kg bodyweight, followed by a flush of at least 10 mL of saline, so that the entire bolus rapidly reaches the systemic circulation. A poor bolus may result in poor image quality or false-negative DCE curves. Although the ACR does not specifically recommend a specific rate for contrast injection, many studies achieve satisfactory results with 2 mL/s. All standard low molecular weight gadolinium-based contrast materials (gadoteridol, gadopentetate dimeglumine, gadobutrol, etc.) have been used by the vast majority of studies with very high-quality results, without clear differences in performance. Gadobutrol (Gadavist) is specifically labeled for use for breast MRI in the United States.

Subtraction images describes the images of contrast enhancement derived from a subtraction of precontrast images from postcontrast images so that only brightly enhancing findings are seen, such as cancer. Subtraction processing suppresses signal from bright fat because adipose tissue does not enhance significantly. However, motion artifact may plague subtraction images, resulting in grainy or noisy images with altered shapes or kinetics. Some software packages allow spatial smoothing of adjacent voxels, reducing noise and artifacts in color mapping, but at the risk of obscuring some intrinsic spatial heterogeneity and features within tumors. For this reason fat suppression is more often used so that high spatial resolution scans may reveal subtle features of cancer such as spiculated margins, and fat suppression can also reduce motion artifacts during subtraction processing. Fat suppression pulses applied intermittently can achieve good fat suppression with little effect on scan time; adiabatic spectrally selective inversion recovery fat suppression is the most efficient method that yields robust, rapid, high-quality images regardless of field strength from 1 T to 3 T.

Maximum intensity projections (MIPs) present the entire breast in a 3D display rather than a single slice to show the entire extent of enhancing findings from the first postcontrast scan. recently showed that a rapid 3-minute breast MRI first postcontrast subtracted T1-weighted image (FAST) scan displayed as an MIP showed 11 breast cancers (7 invasive and 4 ductal carcinoma in situ [DCIS]) in 443 women undergoing 603 screening rounds. MIP analysis showed a sensitivity of 98.9% increasing to 100% with the FAST images; this is a very promising technique for breast cancer screening. MIPs may be reconstructed to sagittal and axial projections resembling x-ray mammograms or may be manipulated to show either volume renderings or thin slabs to display extensive multifocal tumors or extent of nonmass enhancement, while limiting overlapping signal from other regions in the breasts. However, cross-sectional thin-slice MRI planes display fine details of lesions, whereas the MIP may obscure some internal features. Thus the MIP alone should not be used to diagnose tumors. On the other hand, multiplanar MIPs may be reformatted to display tumor distances to the nipple, skin, and chest wall; there are software programs that measure these distances automatically or semiautomatically.

A DCE curve (kinetic curve, or a time intensity curve) shows a lesion’s signal intensity as a function of time during the bolus IV administration of contrast material and over the postcontrast scan time periods. This is usually done by using commercially available programs or computer workstations with the user drawing a region of interest (ROI) greater than 3 pixels large over the enhancing finding.

A variety of image-processing techniques have been developed to automate analysis of dynamic images throughout the breast. Most commercially available programs generate color maps and kinetic curves on a pixel-by-pixel basis using the following steps. First, the program calculates the percentage of enhancement or relative enhancement ( RE ) to the baseline/background signal intensity [SI] value at the operator-defined time point of the early postinjection phase using the formula (SI1 − SI0)/SI0 × 100 (%), where SI0 and SI1 are signal intensities at the baseline and at the operator-defined early postinjection time, respectively. The early postinjection time point is usually defined as between 60 and 90 seconds or at least within the first 2 minutes after a bolus injection. The program then identifies voxels that enhance more than a specified threshold, which can be set by the operator (frequently an RE of 50%–100%, but may vary depending on the intrinsic T1-weighting of the scan protocol, which depends on field strength, FA, and scan repetition rate), at the early postinjection time point, and colorizes these voxels according to the subsequent dynamic kinetics during the delayed phase. The late postinjection time point is usually defined at 4 to 5 minutes after the bolus injection or later, because washout is generally seen at 4 to 5 minutes postinjection. Although no specific color has been set by ACR for the color of the lesion, most vendors determine the color using the following thresholds: (1) increase or decrease by 10% of the early-phase enhancement (Breast Imaging Reporting and Data System [BI-RADS] 2013) or (2) signal enhancement ratio (SER) of 0.9 and 1.1 ( ). SER is the ratio of SI at the early postinjection time point relative to SI at the late postinjection time point expressed by (SI1 − SI0)/ (SI2 − SI0), where SI0, SI1, and SI2 are signal intensities at the precontrast, the early postinjection, and the late postinjection time points, respectively ( ). For example, voxels with persistent enhancement, characterized by increase of relative SI by at least 10% or defined as SER <0.9, is colored in blue, and those with washout enhancement, defined as decease of relative SI by at least 10% or SER >1.1, in red. The others, which belong to plateau kinetics, are colored in green or yellow. Suspicious lesions, that usually have plateau or washout kinetics, stand out against normally gradually enhancing background parenchymal enhancement (BPE).

Some software packages automatically calculate and segment breast tumor volumes by detecting suspicious enhancement kinetics (visually these would be based on their color). Automated volume measurements are less accurate in malignancies that are heterogeneous, enhance slowly or poorly, and may produce spurious results in the face of uncorrected motion artifact. Strong spatial variations in signal intensity caused by variations in surface coil sensitivity, and B1 + transmit inhomogeneity may also produce inaccurate breast tumor results.

Pharmacokinetic scans or physiologic scans are scans that superimpose physiologic maps, such as enhancement information, on morphologic images, combining both types of information into one format. This type of scan usually shows the morphologic appearance of a lesion with the physiologic image superimposed in color.

MRI Spectroscopy

MRI spectroscopy can be performed as an optional study to obtain information about the chemical content of breast lesions. Several studies have shown that choline-containing compounds (tCho) peak measured with hydrogen ( ¹ H)-spectroscopy were more often elevated in malignant breast lesions than in benign lesions or normal breast tissues ( ), suggesting the potential utility of MRI spectroscopy in the diagnosis of breast cancer. For single-voxel choline spectroscopy, a minimum voxel size of 1 × 1 × 1 cm is recommended for adequate signal-to-noise. Localized shimming and high-quality spatial saturation pulses, fat suppression, and partial water suppression improve quality of spectra. It has been recommended to avoid negatively charged gadolinium agents if spectroscopy is performed after contrast enhancement study, because they may reduce choline signal ( ).

Common Artifacts

Common artifacts and pitfalls seen in breast MRI include line(s) of noise, ghosting artifact, blurring, misregistration artifacts, wraparound artifact ( aliasing ), susceptibility artifact, poor fat saturation, and chemical shift artifact ( Table 7.2 ). They can make interpretation of breast MRI challenging and lead to misdiagnosis.

Lines of noise can be generated when there is electronic noise, room shielding is poor, scan room door opens, or there is a failure of MRI equipment.

Physiologic movement, including cardiac pulsation, respiration, or gastrointestinal peristalsis, as well as patient motion, can cause a few types of artifact. Physiologically moving objects, such as blood in the heart and vessels, pleural fluid, or bowel fluid, can become blurred and produce duplicated high signal intensity in the phase-encoding direction. This is called ghosting ( Fig. 7.4 ). Ghosting occurs always in the phase-encoding direction, regardless of the direction of the motion. Ghosting can be prevented from obscuring breast tissue by careful selection of phase- and frequency-encoding directions, and by placing a saturation band over the moving structure. Patient motion can result in blurring of the entire image, so it is especially important that the patient hold still and breathe quietly during scanning. Specifically on subtraction images, patient motion can cause alternating bright and dark bands at structure interfaces (such as fat–skin or fat–glandular interfaces) called misregistration artifact, when the patient moves between the images to be subtracted ( Fig. 7.5 ).

One other cause of noise, including unstructured noise or structured ghosting, arises from problems with parallel image reconstruction. Parallel imaging is an image-processing technique that uses spatial information contained in the component coils of an array to partially replace spatial encoding, which would normally be performed using gradients. There have been various parallel imaging techniques developed, such as ASSET, SENSE, mSENSE, GRAPPA, and ARC. The parallel imaging has advantages, including significant reduction in imaging acquisition time and reduction in susceptibility artifacts, but also has disadvantages, including reduction in SNR and increased artifacts caused by the additional imaging processing. These parallel imaging-specific artifacts happen if the acceleration factor and/or coil sensitivity mapping (obtained by the calibration scan) is not correct for the type of breast coil used and are caused by the shape and size of the patient.

Wraparound artifact (also called aliasing or phase wrap ) is caused by aliasing from tissue outside the prescribed imaging FOV that is excited during the image acquisition process. In this artifact, the excited tissue outside the imaging FOV is misregistered and is wrongly superimposed on structures within the FOV in the reconstructed image ( Fig. 7.6 ). Wrapped around artifact is mostly encountered in the phase-encoding direction, because wrapped around artifact in the frequency-encoding direction is usually suppressed by oversampling or the use of a frequency filter. To minimize wrapped around artifact in the phase-encoding direction, increasing the imaging FOV or oversampling in the phase-encoding direction can be used, although these techniques have several disadvantages, such as increasing imaging time and reducing the temporal resolution.

Susceptibility artifact is caused by metallic objects, which induce inhomogeneity in the magnetic field, and is shown as local signal dropout, bright spots, or tissue distortion ( Fig. 7.7 ). Ferromagnetic metals (such as iron, nickel, and cobalt) cause more severe inhomogeneity compared with nonferromagnetic metals (such as titanium). Susceptibility artifacts appear more prominent on gradient-echo images than on FSE images, and at higher field strength (on 3 T than on 1.5 T). These are common around marker clips.

Poor or inhomogeneous fat suppression is usually caused by poor shimming or incorrect choice of the excitation center frequency (usually 220 Hz at 1.5 T), especially in patients with silicone implants or non–MRI-compatible objects, such as metallic BB skin markers, magnetic tissue expanders, metallic scar markers, and metal infusion ports, in or near the breast. Even with optimal shimming, the breast anatomy near the nipple, axilla, chest wall, and interior breast may cause unavoidable variations in B0 field homogeneity and fat suppression failures, resulting in artifacts. Poor fat suppression is easily identified as hyperintense fat on fat-suppressed images ( Fig. 7.8 ).

Chemical shift artifact can be common on nonfat suppressed gradient-echo images. It occurs at the interface of fat:water pixels. Because the fat signal is shifted a few pixels in the read-out direction relative to water, there can be a piling up of combined fat:water signal at some interfaces, and lack of signal at others. This may cause artificial appearance of asymmetric skin thickness, and bright/dark signal around biopsy clips especially on spin-echo images.

Pitfalls in DCE Studies

Only a good bolus of contrast and timely scanning will translate into breast cancer enhancing rapidly on MRI scans. Failure in DCE studies, including slow bolus, delayed bolus, or late scanning, may result in a nondiagnostic examination and may cause misinterpretation of kinetic curves. Before interpreting dynamic curves, it is important to assess if the technical factors producing the dynamic curves were optimal, resulting in dynamic data good enough to be evaluated.

Cardiovascular kinetic curve obtained with a ROI over large vessels (aorta) or the heart is a useful indicator of failure in DCE studies. The healthy heart/aorta usually shows normal rapid, avid (fast) initial enhancement and rapid late washout ( Figs. 7.9 and 7.10A ). Alternation of the normal cardiovascular dynamic curve pattern tells the radiologist that the DCE study is suboptimal ( Fig. 7.10 B–D ).

Slow or poor bolus of IV contrast may occur when the IV line is injected slowly or may even become detached from the vein, the connector may leak, the catheter may be kinked, there may be a poor saline flush, or heart function is impaired. The contrast either enters slowly, rather than in a bolus, or never enters the patient. Insufficient delivery of contrast media may cause the decrease or absence of the peak of cardiovascular dynamic curve ( Fig. 7.10B ), resulting in poor enhancement of breast lesions and leading to misinterpretation of the washout kinetics as plateau or persistent patterns.

Delayed bolus of IV contrast or poor circulation may cause the delay of the cardiovascular enhancement with or without decrease of the peak enhancement ( Fig. 7.10C ). The first postcontrast scan, which is usually performed to detect rapid peak enhancement, can be obtained earlier than expected, and thus cannot detect a cancer with rapid initial enhancement as different from background.

In late MRI scanning, the early postcontrast scan can miss the peak enhancement ( Fig. 7.10D ). If a cancer has been already washed out at the start of MRI scan, a bright signal from the cancer during the early phase will be missed ( Fig. 7.11 ). Late scanning is caused by an operator error and an incorrect protocol, or when a patient interrupts the scan. Cancers may be obscured by BPE with late scans.

When the cardiovascular dynamic curve pattern is altered, these MRI scans cannot be trusted to show cancer in the breast, and further investigation would be needed to determine the causes of the failed bolus or the late scanning. It is sometimes difficult even for trained technologists to notice the failure in contrast enhancement, thus it is important for radiologists to pay attention to the adequacy of contrast enhancement bolus whenever they assess DCE MRI studies.

Motion artifact is another cause that leads spurious results in DCE analysis. Patient motion can generate faulty kinetic curves if the ROI includes different parts of the lesion or if it includes the adjacent breast parenchyma instead of the target between dynamic images ( ). Particular attention should be paid to identify any patient motion.