Parallel processing gives boost to mammography

December 2, 2004

Strong evidence suggests that MR imaging is the most sensitive of all breast imaging techniques, including state-of-the-art mammography and high-frequency breast ultrasound. In contrast to mammography, the sensitivity and specificity of MR are not impaired by dense breast tissue, fibrocystic disease, or therapeutic interventions such as postsurgical scars or radiotherapy-induced fibrosis.

Strong evidence suggests that MR imaging is the most sensitive of all breast imaging techniques, including state-of-the-art mammography and high-frequency breast ultrasound.1-11 In contrast to mammography, the sensitivity and specificity of MR are not impaired by dense breast tissue, fibrocystic disease, or therapeutic interventions such as postsurgical scars or radiotherapy-induced fibrosis.

Sensitivity for invasive breast cancers is virtually 100% with state-of-the-art MR scanners. A nonenhancing breast cancer is such a rare finding that it remains worthy of a case report. Sensitivity for ductal carcinoma in situ is lower, presumably because of the inconsistent angiogenic activity of preinvasive cancer, which translates into an inconsistent enhancement pattern. With adequate diagnostic criteria, however, sensitivity for DCIS is about 90%.

Current data suggest that breast MR and mammography have complementary roles in DCIS. Mammography helps detect the 5% to 10% of DCIS cases that do not enhance on MR and another 30% that reveal imaging characteristics overlapping benign lesions, while MR helps identify additional cases of DCIS that are mammographically occult because of the absence of microcalcifications.8

MR of the breast has been advocated mainly as a second-line imaging modality that can noninvasively clarify equivocal findings on conventional imaging, particularly in women who have undergone breast conservation therapy for breast cancer. Unlike mammography or breast ultrasound, MR imaging allows the ad hoc distinction of scar and recurrent tumor with high sensitivity and specificity. But breast MR should be performed even when the diagnosis of breast cancer is clear, based on conventional imaging.

In this setting, MR imaging is indicated for preoperative local staging to identify additional conventionally occult breast cancers in the same or contralateral breast before breast conservation therapy is initiated. When MR is used in this way, additional cancers are identifiable in up to 16% to 24% of patients with a seemingly solitary breast cancer diagnosed on mammography and breast ultrasound.3,4 In patients with implants inserted for cosmetic or reconstructive reasons, MR is the only breast imaging technique that allows detection of recurrent breast cancer arising behind the prosthesis.

It has become increasingly evident in recent years that MR imaging is also useful as a first-line imaging modality. With growing evidence of its superior diagnostic accuracy compared with conventional imaging methods, several large-scale trials are under way to evaluate the use of MR in screening specific subsets of patients.12-15 Despite the higher direct costs of MR over conventional methods, using the less sensitive imaging methods as gatekeepers for the method with higher sensitivity may not be sensible.

Why should a radiologist have to wait until a cancer is visible on a mammogram to be "allowed" to use breast MR? A paradigm shift would be in accordance with the concept of individualizing screening efforts so that not all women undergo the same protocol of yearly mammographic screening starting at age 40. Screening efforts could be tailored to the individual risk profile so that intensified screening protocols, possibly including MR, would be offered to women who are at increased risk.

Women with an increased risk of breast cancer include those already diagnosed with breast cancer, who have a high risk of recurrent ipsilateral, synchronous, or metachronous contralateral breast cancer; women with a history of borderline tissue diagnosis such as lobular carcinoma in situ, atypical ductal hyperplasia, or radial scars; women with a strong family history of breast cancer, particularly early-onset breast cancer; and those with presumed or proven mutation in a breast-cancer susceptibility gene, resulting in a condition called hereditary or familial breast cancer.


An important issue in breast MR is its limited specificity. Contrast enhancement is not confined to malignant lesions but is associated with a huge variety of benign disease states. Even normal fibroglandular tissue exhibits variable enhancement, especially in younger women under the influence of endogenous ovarian hormones and postmenopausal women using hormonal replacement therapy. Accordingly, the first and most difficult task for a radiologist performing breast MR is not to detect breast cancers but to differentiate benign from malignant enhancement.

The most important approach to distinguishing benign or spontaneous enhancement from a malignant lesion is to assess the lesion's morphologic features and enhancement behavior. The angiogenic activity of breast cancers leads to an increased local vessel density and capillary permeability. This translates into relatively characteristic enhancement kinetics of breast cancers compared with benign lesions. Breast cancers typically exhibit rapid and strong enhancement after a gadolinium-DTPA bolus, followed by a more or less pronounced contrast washout after the first two to three minutes. This washout of contrast agent from the tumor is one of the most important diagnostic features used for differential diagnosis of enhancing lesions.

To track the rapid signal intensity changes that occur after contrast, and in particular to track the contrast washout, it is necessary to image in a dynamic mode, i.e., repetitively, with a temporal resolution tailored to the contrast kinetics one wishes to observe. Rapid (dynamic) imaging is vital not only for tracking lesion kinetics but also for assessing lesion morphology. The contrast between malignant lesions and surrounding enhancing fibroglandular tissue is optimal for only about two to three minutes, and lesion-to-parenchyma contrast makes a detailed analysis of morphologic features feasible only during this short time. Washout of contrast in the lesion and the concomitant progressive enhancement in the surrounding parenchyma leads to a cancellation of lesion contrast after this period. A temporal resolution of less than two minutes, and ideally about one minute, per dynamic scan is desirable for contrast-enhanced breast MR imaging.

But at the same time, high spatial resolution is an indispensable prerequisite for all breast imaging techniques, and MR is no exception. High spatial resolution is needed to depict morphologic details of tumor boundaries and internal architecture. An in-plane and through-plane pixel size of


The challenges of MR imaging of the breast are comparable to those of contrast-enhanced MR angiography: In both settings, rapid imaging must be combined with high spatial resolution.

Earlier concepts of rapid imaging proved unsuitable for breast imaging, mainly because the inherent image artifacts or the changes of lesion contrast enhancement kinetics were unacceptable; this holds true for turbo spin-echo, echo-planar, and keyhole imaging techniques. The majority of breast imaging protocols, therefore, still use a true "dinosaur" pulse sequence: a simple 2D or 3D T1-weighted gradient-echo technique. Simply increasing gradient strength is not an effective way to accelerate image acquisition in breast MR, because echo times are dictated by in-phase settings for fat and water.

But increasing gradient strength was, until recently, the only strategy available to meet the expanding demands of advanced diagnostic imaging applications. This technique is limited by physical, economical, and medical considerations, however. Increasing gradient strength is technically difficult, is associated with significant hardware costs, and involves a risk of inducing unwanted side effects such as peripheral neurostimulation.

Parallel imaging techniques such as SENSE can overcome this dilemma. In parallel imaging, phase-encoding steps are replaced by a method that exploits the spatial information inherent to the spatially variable sensitivity of an array of surface coils.

Several MR system vendors have developed parallel imaging techniques (ASSET, IPAT, SENSE). All enable a reduction of phase-encoding steps by a factor of at least two to six.16-20 By reducing the number of phase-encoding steps necessary for image generation, parallel imaging has become an efficient tool for speeding image acquisition. The resulting gain can be traded for improved spatial and/or temporal resolution in any pulse sequence.

Any increase in spatial resolution is associated with a strong penalty in the signal-to-noise ratio; the use of SENSE is accompanied by a reduction in SNR of about 30%. Most SENSE applications at 1.5T are used to improve temporal resolution. If increased spatial resolution is desired, the use of SENSE is usually reserved for pulse sequences with a very high inherent SNR, such as MR angiography. SENSE will be even more useful at 3T, due to the higher SNR-not only because it helps reduce acquisition time, but particularly because it helps reduce radio-frequency deposition as a result of the reduced number of phase-encoding steps. As specific absorption rate limitations are reached more quickly at 3T than at 1.5T, reducing RF deposition is extremely important to fully exploiting the SNR advantage at 3T.

Using SENSE in breast MR solves-or at least soothes-the high-spatial-versus-high-temporal dilemma that exists with current breast MR imaging protocols, and it does so without affecting image contrast.

At the University of Bonn, we use a SENSE factor of 2 to speed up the acquisition time of a 2D gradient-echo dynamic pulse sequence. With conventional sequential k-space sampling and at a temporal resolution of one minute per dynamic scan, we can accommodate 29 sections with a 256 x 256 imaging matrix (TR 280/TE 4.6/flip angle 90 degrees ). With SENSE, we have increased the spatial resolution to a 400 x 512 imaging matrix, keeping the temporal resolution at one minute and without changing the contrast-determining image parameters.

Our clinical impression is that the yield in spatial resolution translates to an almost proportional increase in diagnostic specificity. With the increased spatial resolution provided by SENSE imaging, we were able to identify spicules in breast cancers that had seemingly smooth borders in our regular imaging protocol (see figure). In turn, lesions that had been rated equivocal in our conventional protocol could be correctly classified as definitively benign in the SENSE-enhanced protocol because internal septations (a finding with a high specificity for fibroadenoma) became visible at high spatial resolution.

The reduction of SNR that is associated with the use of SENSE and a high acquisition matrix was perceivable on the subtracted postcontrast images. But the high anatomic detail and crisp arterial-phase image contrast that SENSE provides are much more important clinically, and the quality of the SENSE image was consistently rated higher compared with the sequential imaging protocol.

Apart from economic considerations, limited specificity is the single major reason that MR is not used clinically on a much broader scale. Its unprecedented sensitivity should recommend it as the best breast cancer screening tool. We expect that the advent of SENSE will have a positive impact on diagnostic specificity, which may greatly enhance the clinical use of breast MR imaging, including its acceptance as a screening tool.


1. Kuhl CK. MRI of breast tumors. Eur Radiol 2000;10:46-58.

2. Kuhl CK, Schild HH. Dynamic image interpretation of MRI of the breast. J Magn Reson Imaging 2000;12:965-974.

3. Fischer U, Kopka L, Grabbe E. Breast carcinoma: effect of preoperative contrast-enhanced MR imaging on the therapeutic approach. Radiology 1999;213:881-888.

4. Liberman L, Morris EA, Kim CM, et al. MR imaging findings in the contralateral breast of women with recently diagnosed breast cancer. AJR 2003;180:333-341.

5. Olson JA Jr, Morris EA, Van Zee KJ, et al. Magnetic resonance imaging facilitates breast conservation for occult breast cancer. Ann Surg Oncol 2000;7:411-415.

6. Morris EA, Schwartz LH, Drotman MB, et al. Evaluation of pectoralis major muscle in patients with posterior breast tumors on breast MR images: early experience. Radiology 2000;214:67-72.

7. Morris EA, Schwartz LH, Dershaw DD, et al. MR imaging of the breast in patients with occult primary breast carcinoma. Radiology 1997;205:437-440.

8. Lee SG, Orel SG, Woo IJ, et al. MR imaging screening of the contralateral breast in patients with newly diagnosed breast cancer: preliminary results. Radiology 2003;226:773-778.

9. Tillman GF, Orel SG, Schnall MD, et al. Effect of breast magnetic resonance imaging on the clinical management of women with early-stage breast carcinoma. J Clin Oncol 2002;20:3413-3423.

10. Weinstein SP, Orel SG, Heller R, et al. MR imaging of the breast in patients with invasive lobular carcinoma. AJR 2001;176:399-406.

11. Harms SE. Integration of breast MRI in clinical trials. J Magn Reson Imaging 2001;13:830-836.

12. Kuhl CK, Schmutzler R, Leutner C, et al. Breast MR imaging screening in 192 women proved or suspected to be carriers of a breast cancer susceptibility gene: preliminary results. Radiology 2000;215:267-279.

13. Cilotti A, Caligo MA, Cipollini G, et al. Breast MR imaging screening in eight women proved or suspected to be carriers of BRCA1&2 gene mutations. J Exp Clin Cancer Res 2002;21(3 suppl):137-140.

14. Podo F, Sardanelli F, Canese R, et al. The Italian multi-centre project on evaluation of MRI and other imaging modalities in early detection of breast cancer in subjects at high genetic risk. J Exp Clin Cancer Res 2002;21(3 suppl):115-124.

15. Stoutjesdijk MJ, Boetes C, Jager GJ, et al. Magnetic resonance imaging and mammography in women with a hereditary risk of breast cancer. J Natl Cancer Inst 2001;93:1754-1755.

16. Pruessmann KP, Weiger M, Scheidegger MB, Boesiger P. SENSE: sensitivity encoding for fast MRI. Magn Reson Med 1999;42:952-962.

17. Griswold MA, Jakob PM, Heidemann RM, et al. Generalized autocalibrating partially parallel acquisitions (GRAPPA). Magn Reson Med 2002;47:1202-1210.

18. Griswold MA, Jakob PM, Nittka M, et al. Partially parallel imaging with localized sensitivities (PILS). Magn Reson Med 2000;44:602-609.

19. Sodickson DK, Manning WJ. Simultaneous acquisition of spatial harmonics (SMASH): fast imaging with radiofrequency coil arrays. Magn Reson Med 1997;38:591-603.

20. Sodickson DK. Tailored SMASH image reconstructions for robust in vivo parallel MR imaging. Magn Reson Med 2000;44:243-251.