Mammography fails to demonstrate all breast cancers, including many that are already palpable. It is hoped that full-field digital mammography will surpass standard screen-film mammography in detecting breast cancers because of its superior contrast resolution, but unfortunately, clinical studies have not yet borne out this hope.

The true benefit of digital mammography may lie in its use in advanced applications. One idea that has been proposed is to use intravenous iodinated contrast to enhance cancers that would be otherwise mammographically occult; this technique is termed contrast-enhanced digital subtraction mammography (CEDSM). Four years after the introduction of the first full-field digital mammography prototypes, CEDSM is just beginning to show signs of success.

Breast cancers enhance with iodinated contrast on CT, and, of course, with gadolinium contrast on MRI. Both CT and MRI, however, have far greater contrast resolution than even digital mammography or any projection radiography technique. Except in the kidneys, where iodinated contrast is naturally concentrated, plain films have not proved useful for depicting contrast enhancement of tissue.

If contrast enhancement is to be seen on a mammogram, acquisition parameters must be optimized and subtraction of the normal breast parenchyma is desirable. Digital mammography allows the use of digital subtraction techniques, as commonly used in digital angiography, providing greater flexibility and ease of use than film subtraction. Advanced algorithms and image registration techniques can be employed as needed.

One drawback of mammography is that compression of the breast applies a pressure greater than that of venous return and seems to decrease the flow of intravenous contrast to the breast tissue. This implies that the breast should not be fully compressed during (or immediately after) contrast administration. It is not possible, however, to take a compressed mammogram, release compression for contrast injection and then recompress the breast for another mammogram in exactly the same position. This poses a problem for performing temporal subtraction, or subtraction of the precontrast mammogram from the postcontrast mammogram.

It is possible to overcome this problem by taking the precontrast mammogram with limited or no compression and having the patient remain in that position during the injection and postcontrast images. This approach requires balancing the negative effect of compression on contrast uptake with its positive effects in terms of immobilization, decrease of breast thickness (to decrease x-ray scatter), and separation of normal overlapping structures. Using linear or nonlinear digital image registration techniques prior to image subtraction partially compensates for the problem of small changes in position during the contrast injection.

At our center we are developing and testing an alternate method for performing contrast-enhanced digital mammography. Called dual-energy CEDSM, the method uses the technique of dual-energy subtraction imaging to increase the conspicuity of the contrast enhancement. Images are acquired at energies straddling the k-edge of iodine (33.2 keV) and are then subtracted to produce an image that minimizes the attenuation of breast tissue, allowing better detection of the iodine attenuation. Because the technique does not require subtracting the precontrast images from the postcontrast images, the breast can be uncompressed during contrast administration but imaged while compressed without adversely affecting the subtraction process. The energy of the x-ray beam is modified between the low- and high-energy exposures by changing the kVp and applying additional external filtration.

Early Results

This work, funded by a grant from the Susan Komen Foundation, began in January 2000. About 25 patients will be studied during a two-year period. A variety of modifications to the technique are being tested during this pilot study, including x-ray technical factors, contrast dose and rate, and the timing of images. We have also experimented with performing the technique with no compression. Although the noncompressed images often depict a cancer, the compressed images are superior. This is likely because there is less overlapping breast tissue and less x-ray scatter.

Ten malignant and eight benign cases have been studied so far. Each invasive cancer has shown enhancement and most have strongly enhanced. Figure 1 shows a 12-mm invasive lobular carcinoma that was palpable but mammographically occult. A metal BB marks the lesion on all images. The cancer appears as a dense mass on a bland background on the postcontrast dual-energy image (Figure 1C). With dual-energy subtraction, it is not necessary to perform imaging prior to contrast administration, but for comparison, the precontrast dual-energy image was obtained (Figure 1B), showing only the bland image of the breast resulting from subtraction of the breast tissue. During the same injection, uncompressed images were taken (Figures 2A, B, and C). While the enhancing cancer is still visible (2C), the contrast is not as great as in the compressed images. A surprising finding is that the tumor is visible on the uncompressed precontrast mammogram (2A), even though it is occult on the standard compressed mammogram (1A).

Figure 3 illustrates how the process works and shows the benefits of subtraction in more subtle cases. This patient entered the study after being recommended for biopsy because of new, highly suspicious microcalcifications. Figure 3A is the unenhanced digital mammogram. Figures 3B, C, and D are the postcontrast low-energy, postcontrast high-energy, and subtracted images, respectively. The enhancing areas on the subtracted image correlated with the areas of invasive ductal carcinoma at biopsy. Although the enhancement can be easily seen on the subtracted image, it is not appreciable on the unsubtracted postcontrast images. Note that the areas of ductal carcinoma in situ (DCIS), indicated by the calcifications, did not enhance. This correlates with findings from breast MRI, for which sensitivity for DCIS is much lower than for invasive cancer.

Both breast MRI performed with gadolinium contrast and contrast-enhanced digital mammography make use of the fact that breast cancers take up more contrast agent than does normal tissue. This is presumably due to tumor angiogenesis, which results in the formation of a high density of abnormally permeable microvessels in the tumor. Breast MRI suffers from a lack a specificity because of the enhancement of benign fibroglandular tissue and benign masses.

It is likely that CEDSM will also show enhancement of benign lesions, but none has enhanced thus far. One case (Figure 4) showed diffuse enhancement of normal tissues in a postmenopausal 46-year-old woman. She presented with an approximately 2-cm palpable abnormality in the superolateral right breast. Ultrasound imaging of the area did not reveal the cause of the palpable abnormality, but a 5-mm hypoechoic shadowing mass, suspicious for cancer, was demonstrated adjacent to it. CEDSM was performed, demonstrating what later proved to be a 5-mm tubular carcinoma (Figure 4C, arrow), a low-grade type of invasive ductal carcinoma. The 1.5-cm fibroadenoma, found at surgery to be the cause of the palpable abnormality, did not enhance.

Discussion

Dual-energy CEDSM has performed well in demonstrating invasive cancers in early results. If further study confirms this apparent high sensitivity, CEDSM may be useful as a problem-solving technique, as breast MRI is today. CEDSM has an advantage over MRI in that it could easily be used in preoperative needle localization for surgical biopsy or, with minor modification of the equipment, in stereotactic biopsy.

For a new technique to have a significant impact on breast cancer mortality, however, it needs to be employed for screening, and both CEDSM and MRI must overcome obvious hurdles before such use could become practical. Equipment cost, lack of specificity, and difficulty in performing biopsy are generally cited as the major barriers to screening with MRI.

The equipment needed for CEDSM-a full-field digital mammography unit-costs more than a screen-film unit but less than an MR scanner, and it may well be a standard part of the all-digital department of the future. The purchase and administration of the contrast media will also contribute to the cost.

The specificity of CEDSM is not yet known, but our early results are promising and indicate that it is likely to be more specific than standard screening mammography. A large fraction of positive screening mammograms are due to densities caused by overlapping normal tissue; these are resolved by recalling the patient and performing additional mammographic views. While these cost far less than a biopsy, they are a considerable part of the total cost of mammographic screening. Overlapping tissue would not be a source of false-positive findings in CEDSM, nor would islands of normal tissue, another source of recalls and negative biopsies. If CEDSM can be performed with substantially greater specificity than standard mammography, the cost savings from fewer recalls and biopsies could offset the increased costs of the initial screening test.

The increased invasiveness of the technique, including IV placement and, especially, the risk of a contrast reaction, might be more difficult to overcome. While some physically invasive screening exams, such as the Pap smear, sigmoidoscopy, and serum cholesterol, are already in common use, none combines its invasiveness with the level of risk involved in injecting contrast.

According to the American College of Radiology contrast manual, the risk of a life-threatening contrast reaction is about one in 100,000 to one in 200,000 for low-osmolar iodinated contrast. The risk of gadolinium contrast is similar. While it is tolerable for use in diagnostic studies such as CT or MRI of the head or body, such a risk is likely too great for a screening test. Eliminating women with a history of allergy or heart disease (performing standard mammography instead) would likely lower these risks, but by how much is unknown.

Because some of the risk is dose-related, reducing the contrast dose would decrease the risk somewhat. If the dose could be brought down below about 20 cc, the need for an IV placement would be eliminated.

An advantage of the dual-energy technique for CEDSM is that the patient does not need to be immobilized during injection and could be injected outside the mammography room prior to the exam.

To reduce the contrast agent dose while maintaining a good contrast-to-noise ratio for enhanced cancers, improvements in techniques must be sought through research. At this early stage, all aspects of both dual-energy and temporal subtraction are ripe for improvement. Areas with the greatest potential include:

• optimization of injection rate and timing of imaging;
• optimization of the x-ray spectrum; and
• novel digital image processing techniques for combining images.

Developing safer and possibly more concentrated contrast agents would likely require a substantial investment by contrast agent manufacturers. We certainly cannot count on this happening, but with an estimated 48 million screening mammograms performed around the world every year, it is a market that is likely to gain their attention. In the meantime, much work remains to be done by the breast imaging community in determining the sensitivity and specificity of this promising technique and optimizing it for diagnostic use.