| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Editorial |
1 From the Department of Radiological Sciences, UCLA School of Medicine, University of California, Los Angeles, 10833 Le Conte Ave, Los Angeles, CA 90095-1721. Received May 26, 1999; accepted June 3. Address reprint requests to the author (e-mail: ckimmesmith@mednet.ucla.edu).
Index terms: Breast radiography, radiation dose, 00.99 • Breast radiography, technology, 00.12 • Breast radiography, utilization, 00.99 • Editorials • Physics
The article by Boone (1) in the current issue of Radiology describes the methods, assumptions, and verification of normalized glandular dose (DgN) tables that extend beyond those commonly used in screen-film mammography. While some women have a compressed breast thickness that is greater than the 8-cm thickness given in the glandular dose tables published by Wu et al (2,3), only rarely do we need this type of information, because we rarely need to know the DgN value for energies above 35 kVp or for tungsten anodes filtered with palladium or silver. So why do we need Boone's article? We need it because mammography is moving beyond the screen-film method, and medical physicists working in digital mammography need, or will need, these tables to help design optimal exposure techniques and, in some cases, refine the design of innovative mammography equipment. Radiologists will need to understand the trends and trade-offs of these designs to enable them to make intelligent decisions about equipment purchases and use.
Dosimetry became necessary for mammography when Bailar (4), in 1977, questioned the cost-benefit ratio of screening mammography. At that time, the average skin exposure per view was about 5 R (1.29 mC/kg), and xeromammography was the prevalent imaging modality, with tungsten anodes and heavy aluminum filtration. Because compression was very gentle and the peak energy across the x-ray tube was about 50 kVp, the dose to the middle of the breast was calculated to be about 20% of the dose to the skin. Use of filters to shape the general radiographic beam had been researched several years earlier (5), and molybdenum filtration began to be used, both with "dual-purpose" tungsten anodes and with molybdenum anodes for screen-film mammography. In early work from 1978 (6), it was suggested that mammography performed with rhodium anodes and tungsten anodes filtered with aluminum would result in lower doses than that performed with molybdenum anodes and filtration, but dose estimates could not be given because dosimetric methods had not yet been developed. Instead, the authors compared the exit exposures through Lucite phantoms to substantiate the reduced attenuation possible by means of hardening of the mammographic spectrum.
In 1979, Hammerstein and associates (7) at Memorial Sloan-Kettering Cancer Center (New York, NY) published an early study of midplane dose in mammography in which they used mastectomy specimens and thermoluminescent dosimeters. They also identified the atomic composition of breast tissues (fat, skin, and glandular) and the respective densities of these tissues; these parameters are still used for formulating tissue-substitute material today. They discovered that the midplane dose needed for screen-film mammography had been overestimated by about 25 fold and that the dose for xeromammography was three times lower than had previously been estimated. In addition, they identified the importance of beam spectral characteristics when estimating the f factor, which is used to convert skin exposure to average dose in the middle of the breast.
During this time, White et al (8) were developing a material that simulated various body tissues and that could be used for dosimetry. One of these materials, BR12, was formulated for mammography. BR12 simulated a 50% adipose, 50% glandular breast. This material was used by Stanton et al (9) to suggest a method for calculating mean glandular dose on the basis of skin exposure as measured with thermoluminescent dosimeters or an ion chamber. Stanton et al also suggested the model of the compressed breast that was used until Boone (1) pointed out that this did not model scatter from the chest wall and changed the model for his present study.
Before the work of Stanton et al (9), Dance (10) used a Monte Carlo technique to simulate the photon interactions in the breast and thus derive a midplane average dose. Dance's data formed the basis of many of the subsequent studies (2,3,11) that have produced the DgN tables we have used since 1991. These have been widely used to estimate the dose-saving potential of rhodium anodes and rhodium filtration. In addition, knowledge of the changes in DgN during magnification (11) has helped supply values of total mammographic dose to young patients who are concerned about their cumulative radiation dose during diagnostic examinations.
Because Siemens Medical Systems (Erlangen, Germany) has introduced tungsten anode combinations, we have been without adequate tables to calculate DgN values for patients who obtain mammograms with such equipment. Boone's tables now rectify this problem, but they fulfill a much greater need for those of us working with digital mammography of the whole breast. At present, there are three manufacturers of whole breast digital receptors that are associated with x-ray–generating equipment. In addition, the Fuji 9000–series computed radiography systems (Fuji Medical Systems, Stamford, Conn) can be used with any mammography system.
Each manufacturer has taken a different design path for its digital image receptor. The SenoScan (Fischer Imaging, Denver, Colo) is a slot scanner with cesium-iodide phosphor, fiberoptics, and charge-coupled devices under the inferior slot. Because of tube loading, a tungsten anode and aluminum filtration are used (see fig 11 in Boone's article). While interpolation to 30 kVp is necessary for this application, it is the first opportunity we have had to estimate the mean glandular dose from published DgN factors for this system. The TREX Digital Mammography System (TREX Medical, Danbury, Conn) is designed with a three-by-four charge-coupled device array, fiberoptics, and cesium iodide phosphor, while the Senographe 2000 D (GE Medical Systems, Milwaukee, Wis) consists of cesium-iodide in contact with an amorphous silicon array. These systems are all undergoing clinical trials for 510(k) approval from the U.S. Food and Drug Administration.
For the GE Medical Systems digital unit and the TREX Medical digital system, screen-film doses are used for the matching digital studies. Because large breasts cannot fit on either of these receptors, there has been no problem calculating doses similar to the screen-film doses for the digital studies. However, new clinical trials are about to begin in which optimized digital techniques established on the basis of the contrast-to-noise ratio and the mean glandular dose will be used. For the Fischer Imaging system, testing to obtain correct mean glandular dose is now facilitated by these DgN tables. Furthermore, because phantom studies show an improved contrast-to-noise ratio at an increased kilovolt peak level and because DgN increases with increased kilovolt peak level, it may be necessary to increase the kilovolt peak energy to beyond 35 kVp (the limiting peak energy level for the tables in Wu et al [2,3]) for these studies. We can now do this by using the tables in Boone's article (1).
Several investigators (12–16) have studied the beneficial effects of increasing the energy spectrum for digital mammography. Johns and Yaffe (12) obtained early data on which later results have been based. They examined 14 samples of fibrous (glandular) breast tissue and 12 infiltrating ductal carcinomas at monochromatic energies of 18–110 keV to calculate the mean linear attenuation coefficients for each tissue type. These results were used in later models of breast tissue surrounding a mass in investigations of the effects of higher peak energy levels on lesion visibility on digital mammographic images.
Fahrig and Yaffe (13) then used these results to calculate the best anode material for particular breast thicknesses and lesions (masses or calcifications). They made the point that for digital mammography, display can be independent of the method of acquiring an attenuation map of the breast; therefore, very small subject contrast differences can be amplified for display in digital mammography, as long as noise is kept low in comparison with signal. In a companion article, Fahrig and Yaffe (14) described the model used for these calculations. In this case, signal is the difference in contrast between the lesion and a mix of 50% glandular tissue and 50% adipose tissue. Fahrig and Yaffe (13,14) studied a range of breast thicknesses (2–8 cm) and peak energy levels (20–31 kVp) for molybdenum anodes and tungsten anodes filtered with a range of k-edge filters. For each breast thickness, dose was kept constant, so that higher kilovolt peak levels and higher k-edge filters are needed for larger breasts. However, the model shows that reasonable signal-to-noise ratios (of 8 and larger) can be achieved with tungsten anodes and 30–31 kVp spectra, and, in fact, these have superior signal-to-noise ratios in comparison with those achieved with molybdenum anodes at these peak energy levels. Fahrig et al (15) later compared the fiberoptic systems modeled earlier (similar to the TREX Medical and Fischer Imaging systems) to an amorphous selenium–based system. However, as modeled previously, the dose in all cases was limited to 60 mrad (0.6 Gy), so an increased kilovolt peak level with tungsten anodes was necessary for larger breasts.
A series of phantom measurements obtained over several months with a TREX Digital Mammography System showed an increase in the contrast-to-noise ratio, for the same 4-cm-thick phantom, as the kilovolt peak level was increased when a molybdenum anode–molybdenum filter combination was used; when a rhodium filter was used, however, the contrast-to-noise ratio decreased as the kilovolt peak level increased (16). However, because DgN factors were available only for energies up to 35 kVp, these investigators did not obtain results beyond that value. Exposures were selected to produce the same exit exposure from the phantom. As long as a minimum threshold of subject contrast between glandular tissue and carcinoma is preserved and noise is kept low, contrast enhancement can restore the contrast lost due to an increased peak energy level.
The ability to reduce dose by increasing the kilovolt peak level may also allow a reduction in compression for digital mammography. For example, a breast that is compressed to 4 cm and exposed to 802 mR (0.21 mC/kg) at 25 kVp for a screen-film radiograph might be compressed to 5 cm for a digital mammographic radiograph; if mammography is then performed at 30 kVp, the skin exposure and mean glandular dose might be the same as for a screen-film examination (see table 1 in the article by Boone [1]), but the exposure to the image receptor would be much higher because of less attenuation in the breast at the higher kilovolt peak level. This will result in a higher signal-to-noise ratio, while the contrast-to-noise ratio will be dependent on the object and background selected for the measurement. Of course, to implement such a change in the compression method would require clinical trials to ensure the diagnostic equivalence of less compression in comparison with imaging with more compression. Although sufficient compression to spread tissue and prevent motion will always be necessary, a 20% decrease in compression may be beneficial in some women. This will also lead to an increase in the use of Boone's tables for breast thicknesses greater than 8 cm.
In addition to conventional uses of digital mammography, Johns and associates (17,18) have been experimenting with dual-energy mammography for many years. To test the theoretic hypothesis that calcifications would be more visible if background tissue contrast could be removed, Johns et al (17) used a precursor of the Fischer Imaging digital mammography unit to image mastectomy specimens with superimposed calcifications. This algorithm requires two images, one obtained at 50 kVp (19 keV) and the other obtained at 110 kVp (68 keV). To ensure that the signal-to-noise ratio is sufficient on the final "calcium" image, the 68-keV mean glandular dose should be 3.2 times that of the 19-keV mean glandular dose. Therefore, DgN factors for polyenergetic spectra must be known at these energy levels if the tissue is to be sufficiently exposed.
It is interesting to note that one of the steps for analyzing breast calcifications with most computer-aided detection methods is to filter out or remove background texture that surrounds calcifications. This often results in removal from the image of some of the smaller calcifications, which have less contrast, although these are the most important calcifications for differentiation of benign from malignant masses. When digital mammography becomes mature, perhaps dual-energy mammography will be used in diagnostic examinations as a preprocessing operation for computer-aided detection of calcifications normally obscured by glandular tissue. The more accurate calculation of DgN will help implement effective dual-energy mammography.
Monoenergetic beams should increase contrast and reduce dose and will also assist dual-energy mammography. Several methods to generate these partially monoenergetic beams have been developed, and at least one method will probably be available for clinical trials in the immediate future (19–21). Boone's article (1) gives us methods for calculating DgN values for these energy spectra. Note that although DgN values for monoenergetic spectra are higher than those for polyenergetic spectra, the skin exposure will be much less. Because these methods require a high density of photons to generate sufficient monoenergetic flux to image a 4–6-cm-thick compressed breast in a reasonable time, their first use may be for specimen radiography. Because core biopsy and localization of very early cancers are making specimen radiography more difficult, the use of digital methods combined with a monochromatic beam would allow precise identification of specimens more quickly and with fewer errors than with the current analog method. Fischer Imaging has started this trend by supplying a core sample holder that can be positioned on their digital stereotactic biopsy unit for magnified specimen imaging.
Two other research programs currently are under investigation. Use of angiography to investigate breast tumor angiogenesis with digital mammography is being studied; it requires kilovolt peak values higher than the k edge for iodine, so Boone's tables will be most welcome for dose calculations. Other digital mammography research is focused on the three-dimensional representation of the breast. Early work has been performed by Maidment et al (22), who used digital breast stereotactic projections of calcifications. Currently, conventional energy spectra are being used, but it is certain that higher kilovolt peak methods and dual-energy subtraction would be valuable when calcifications alone are projected in a rotating three-dimensional display. The ability to see whether calcifications line up in a ductlike organization or whether they are the result of fibroadenoma absorption would be helpful for diagnosis. The clinical implementation of multiple-projection radiography will require a careful analysis of the mean glandular dose required.
Finally, I hope radiologists will note the large differences in DgN factors between primarily adipose and primarily glandular breasts. Because exposure increases as glandular content increases, differences in mean glandular dose between adipose and glandular breasts are not so disparate. However, radiologists should realize that medical physicists who are estimating an individual patient's radiation dose must also estimate the ratio of adipose to glandular tissue so that the correct DgN factor will be used, and this decreases the accuracy of such calculations. In addition, radiologists may note that small changes in the assumed skin thickness (figure 8 in Boone's article [1]) affect DgN values by up to 15% when a molybdenum anode–molybdenum filter energy spectrum is assumed. Because patient differences in skin thickness can easily range over the 2–6-mm thickness studied, reasonable error approximations due to patient differences approach 10%.
While it is important to respect the concerns of the patients and of the Mammography Quality Standards Act, or MQSA, inspectors who wish to contain mammography dose within reasonable limits, Boone's tables illustrate that dose calculations have an inherent error, which limits their use for fine tuning of mammographic exposures. We must not lose sight of the reason for mammography and so should insist on diagnostic content over lowered radiation dose, if such reductions will lower the radiologist's ability to establish an accurate diagnosis. Changes in mammographic techniques that have proven track records, in terms of diagnostic accuracy, so that doses at mammography can be reduced by 10% or 15% should be acceptable only if the results of clinical trials show substantial equivalence between the two methods. Moreover, because a 10% savings in dose can be realized by changing the DgN algorithm (by changing the approximations of skin thickness or glandular content), such improvements may not have a measurable effect on decreasing patient carcinogenesis.
Articles such as Boone's can provide medical physicists with the tools to design or modify digital mammography equipment without affecting mean glandular dose and can give radiologists insights about the complex interactions that affect patient dose estimates.
Footnotes
See also the article by Boone (pp 23–37 ) in this issue.
References
Related Article
This article has been cited by other articles:
|
|
 |
|
  F. E. Carroll Tunable Monochromatic X Rays: A New Paradigm in Medicine Am. J. Roentgenol., September 1, 2002; 179(3): 583 - 590. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
