Radiol Clin N Am 42 (2004) 871 884

Digital mammography
John M. Lewin, MDa,*, Carl J. DOrsi, MDb, R. Edward Hendrick, PhDc
a Diversified Radiology of Colorado, P.C., 938 Bannock Street, Suite 300, Denver, CO 80204, USA Breast Imaging Center, Winship Cancer Institute, Department of Radiology, Emory University Hospital, 1701 Uppergate Drive, Suite C1104, Atlanta, GA 30322, USA c Lynn Sage Comprehensive Breast Center and Department of Radiology, Northwestern University, Feinberg School of Medicine, Galter Pavilion, 13th Floor, 251 East Huron Street, lChicago, IL 60611, USA b

Traditional film mammography is the single best breast cancer screening test to date and has been shown to reduce mortality from breast cancer in large randomized trials [1 3]. Mammography, however, is far from perfect. Using the common definition of a missed cancer as one that becomes clinically evident within a year of a negative mammogram, screening mammography is only about 70% to 75% sensitive in current clinical practice [4 6]. About 10% of cancers are mammographically occult even after they are palpable [7,8]. At a National Cancer Institute sponsored workshop in 1991, an expert panel reviewed all the potential breast cancer screening technologies on the horizon. They concluded that, of all the technologies presented, full-field digital mammography (FFDM) held the greatest promise to improve breast cancer detection [9]. The first commercial prototype of a FFDM unit was tested by Fischer Imaging Corporation (Denver, Colorado) on employee volunteers in 1995. Over the next 2 years, Fischer, General Electric, and Trex Medical (which bought Lorad and Bennett) each produced a small number of prototypes for clinical trials. In January, 2000, General Electric became the first manufacturer with a Food and Drug Administration (FDA) approved device (the Senographe 2000D), followed by Fischer Imaging (Senoscan) in September, 2001, and Hologic/Lorad (Bedford, Massachusetts) in March 2002. As of this writing, there are several more manufac-

* Corresponding author. E-mail address: [email protected] (J.M. Lewin).

turers (eg, Fuji, Siemens, Sectra, and Instrumentarium) with digital units approved for use in other countries, which have not yet completed the FDA approval process. Most FFDM devices are similar to screen-film mammography (SFM) units, with screen-film cassette used to record the image replaced by a digital detector (Fig. 1). The most widely used system, made by General Electric, uses a stationary flat-panel detector composed of cesium iodide (to absorb x-rays and convert them to visible light), behind which is an array of 4.4 million 100-mm photodiodes with the same dimension as the resulting image. The detector size is 19 23 cm, roughly the size of a smaller mammography screen-film cassette, with a 100 mm limiting spatial resolution. A system with a larger detector, about 24 30 cm, the size of the larger mammography screen-film cassette, is being developed. The Fischer system is unique among the clinically available systems in that it uses a 1-cm collimated x-ray beam that sweeps across the breast in synchrony with a 1-cm wide slot array of charge-coupled device (CCD) chips similar to those used in digital cameras. To accommodate this scanning procedure, the Fischer system has wider x-ray tube housing and a curved platform for the breast (Fig. 2). To sustain the 5.2-second scan needed to cover the entire breast, the tube target material is tungsten, filtered with aluminum, as in conventional x-ray tubes. This system produces an image that is 21 29 cm, containing 4096 5624 pixels, at a spatial resolution of 54 mm. Lorad (Hologic) originally used a panel of 12 fiberoptic reducers into 12 CCD arrays in a mo-

0033-8389/04/$ see front matter D 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.rcl.2004.06.004

872

J.M. Lewin et al / Radiol Clin N Am 42 (2004) 871 884

Fig. 1. The Lorad Selenia full-field digital mammography system is based on the popular M-IV film mammography device. The digital detector sits where the film cassette holder is on the M-IV, but otherwise the appearance is identical to a film unit. (Courtesy of Hologic, Bedford, MA; with permission.)

puter radiography systems read the emitted blue light from only one side of the screen. By making the base of the screen transparent and using a double-sided computer radiography reader, Fuji can make the photostimulable phosphor thinner and still get adequate signal emitted for a mammography exposure. This has allowed them to bring the nominal spatial resolution of their double-sided computer radiography system from 100 to 150 mm down to between 50 and 70 mm. All the techniques described use an x-ray absorbing screen or crystal that transforms the x-ray energy into light, which is then directed to the CCD array or flat panel diode array, or a photostimulable phosphor that requires scanning to produce an electronic image. This type of image capture is termed bindirectQ because there is a second stage used to read out the pattern of x-ray absorption before storage of the image by computer. Another form of indirect image capture is to use an x-ray absorber similar to that used in xeroradiography, amorphous selenium. An amorphous selenium panel is a good absorber of x-rays and a good capacitor. By keeping the panel fixed in

saic or tiled pattern held within a cassette to achieve FFDM. Each of the 12 CCD arrays was 3 3 cm, and 1500 1500 pixels. Because of a 2:1 fiberoptic reducer in front of each array, each 6 6 cm area was covered by one array, yielding 40-mm pixels. The difficulty with this design was the cost of the fiberoptic CCD components: over $20,000 each, making the system too expensive for widespread commercial use. The Fuji Corporation has had a computed radiography system for chest examinations but with a spatial resolution believed to be inadequate for mammography (150 mm). This system stores the latent image on a photostimulable phosphor, which stores x-ray energy rather than emitting it as light. The latent image is read from the phosphor by placing the computer radiography cassette in a reader, which scans the phosphor point by point with a laser. The low energy (red) laser light stimulates the emission of higher energy (blue) light in proportion to prior x-ray exposure. The emitted blue light is measured at each point by light guides and photomultiplier tubes, amplified, and converted from an analog to digital signal, producing an electronic image that can be stored by computer and later displayed. Earlier com-

Fig. 2. Fischer Imagings Senoscan full-field digital mammography system scans the breast over a 5-second period. The appearance varies from that of a film unit in that it has a curved detector and breast platform. This curved design is necessary to keep the distance between the x-ray tube and the detector constant as the x-ray beam sweeps across the breast. The housing for the x-ray tube is also wider to accommodate the motion of the tube. (Courtesy of Fischer Imaging Corporation, Denver, CO; with permission.)

J.M. Lewin et al / Radiol Clin N Am 42 (2004) 871 884

873

the mammography unit, placing a bias voltage across the panel, and reading the charge collected on the bottom side of the plate, which collects ionizing electrons, a digital image can be read out. This requires a second stage, a silicon diode array, to be used to measure the charge pixel by pixel under the selenium plate. This eliminates the need to convert x-ray energy to light, and has the potential to preserve more information for a given exposure to the breast (higher detective quantum efficiency). Lorad (Hologic) has received FDA approval for this technology and is currently marketing a unit with this detector type. The Lorad (Hologic) unit has 23 30 cm field of view and 70 mm limiting spatial resolution. Lorad has abandoned the tiled fiberoptic CCD approach. Other manufacturers, in particular Sectra Imaging (Finland), are using direct digital detectors that measure each absorbed electron created by ionization in the detector material. This has the potential to increase signal ever further for the same dose to the breast, or to reduce dose for comparable image quality. This detector is being used at research sites in Europe, but is not yet approved by the FDA for use in the United States. The signs of cancer on FFDM are essentially the same as those on SFM (Fig. 3). Once stored digitally by computer, a digital mammogram can be printed on film with a fixed window level and window width, reproducing an image similar to SFM, which can be

interpreted in the same way, making the transition from film to digital easier. To realize the full benefits of the digital image, however, digital mammograms are best interpreted in softcopy (monitor display) using a workstation specifically designed for the requirements of digital mammography. These requirements include the ability to deal with the large number of pixels (a digital mammogram has between 4.4 and 27 million pixels, 4 11 times as many pixels as a typical digital chest radiograph) and the rapid throughput required for screening mammography. To meet these requirements, the workstation must be more powerful and have higher-resolution monitors than a standard digital radiology workstation and the software must be optimized for the task. Additionally, to detect the subtle findings, including microcalcifications, the monitors must be brighter than those used in the rest of radiology; just as mammographic light boxes are brighter than those used to view other types of imaging studies.

Technical advantages of digital mammography Unlike SFM, where the image is captured, displayed, and stored on film, digital mammography decouples these three tasks. The image is captured by the digital detector but is displayed on a monitor or film, usually after mathematic processing by a computer. By separating these tasks, the typical tradeoff between dynamic range, the overall range of shadesof-gray that can be imaged by the system, and contrast resolution, the ability to distinguish between small differences in shades-of-gray, can be avoided and both can be optimized. The result is a detector with both higher contrast resolution and equal or better dynamic range than the combination of film and phosphorescent screen used in SFM [10 15]. Additionally, because the digital detector captures more of the incoming x-ray photons than the screenfilm combination and because digital imaging eliminates the film processor, the image has lower noise. With both digital and film mammography, the amount of noise depends primarily on the level of exposure: more x-rays means less noise. Unlike film, where underexposure results in a light image (and low contrast) and overexposure results in a dark image (and low contrast), the amount of exposure in a digital system is independent of image brightness (and contrast). For a wide range of exposures (within the dynamic range of the digital detector), underexposure or overexposure affects only image noise, not image contrast. Although SFM can be evaluated based on the optical densities of the film, digital mammog-

Fig. 3. Cancers on digital mammography have the same features has on standard film mammography. Here a 1-cm invasive ductal carcinoma presents on a digital mammogram as a spiculated mass with adjacent architectural distortion (arrow).

874

J.M. Lewin et al / Radiol Clin N Am 42 (2004) 871 884

Fig. 4. Because the brightness of a digital mammogram is independent of the exposure, an underexposed image may not be recognized by the technologist. (A) Digital mammogram is grossly underexposed. Image is displayed with acceptable contrast and brightness, but excessive image noise obscures image detail. (B) Repeat view at the correct exposure has much less image noise allowing the tissue to be better seen.

raphy must be evaluated based on noise or signal-tonoise ratio. For this reason, observers must be careful to evaluate an image for excessive noise, which could indicate underexposure (Fig. 4). As a display device, a video monitors suffer from a tradeoff between dynamic range and contrast resolution. In fact, mammography film can display more shades-of-gray over a wider range than can a monitor. Because of the digital nature of the image, however, the entire range of a digital image can be displayed with high contrast simply by mathematically processing the image. At its simplest level this involves interactively changing the window and level of the image to vary its displayed brightness and contrast. This process is familiar to anyone who has looked at both soft tissue and bone windows on a CT scan. This process does not allow the user optimally to view the entire image at once, but does provide a method of separately viewing the dense and fatty portions of the breast by optimizing the window and level parameters for each. Krupinski et al [16] demonstrated the effect of luminance in reading efficiency for mammograms that were negative or had subtle masses or calcifications. Four viewing parameters were used, two with SFM on a bright and intermediate luminance view box and two on monitors with relatively high and low luminance. Although the difference in accuracy was not significant, the difference in viewing time taken to detect a mass from the brightest view box with

SFM to the lowest luminance monitor with FFDM was 25.5 seconds longer for FFDM. The time for calcification detection was 26.2 seconds longer for FFDM and, interestingly, the time to declare a mammogram as normal demonstrated a difference of 30.4 seconds longer for FFDM. Another solution, shown later, is to use more advanced mathematics to process the image so that the entire range of brightness levels is reduced but without reducing the contrast between adjacent structures in the image. An example of this type of processing is illustrated in Fig. 5. One area in which SFM surpasses digital is in spatial resolution. Spatial resolution is measured in terms of the smallest high-contrast objects that can be resolved as distinct. A SFM cassette can resolve objects that are about 25 mm in size if the object is placed and imaged directly atop the cassette. In contact mammography, SFM resolves between 12 and 15 line pairs per millimeter, which is equivalent to 42 to 30 mm pixels, respectively. The digital systems spatial resolution ranges from 100 to about 50 mm in standard whole breast mode, or 5 to 10 line pairs per millimeter, respectively. All digital systems have the ability to resolve smaller objects using magnification or a high-resolution mode. It is unclear how important spatial resolution below 100 mm is, given that objects, such as calcifications, become fainter as they become smaller, making contrast resolution more important at that size. It is known that the

J.M. Lewin et al / Radiol Clin N Am 42 (2004) 871 884

875

Fig. 5. Digital mammograms must be processed to see the entire breast, including the skin, at a single window or level setting. (A) Unprocessed digital mammogram shows detail at the center of the breast but the skin and subcutaneous tissue are lost. (B) Thickness compensation brightens the portion of the breast near the skin, allowing it to be seen without losing contrast. (C) Adaptive histogram equalization (called premium view on the GE system) modifies the entire image to allow the brightest and darkest parts of the breast to be seen at a single window or level setting while increasing contrast throughout the image.

smallest microcalcifications detected in contact SFM are between 200 and 250 mm in largest dimension.

Clinical trials comparing film and digital mammography Two trials comparing digital mammography with film mammography for breast cancer screening have

been completed. The Colorado-Massachusetts screening trial was performed between 1997 and 2000 using a prototype unit made by General Electric [17,18]. Each of the 6768 subjects in this trial underwent both a film mammogram and a digital mammogram. The resulting images were then read independently by a different radiologist, with each radiologist reading an equal number of film and digital examinations. Findings detected on either film or digital were

876

J.M. Lewin et al / Radiol Clin N Am 42 (2004) 871 884

Fig. 6. Case from the Colorado-Massachusetts digital mammography screening trial. A 1-cm invasive ductal carcinoma detected only on digital mammography. On film mammography (left) the cancer (arrow) is an indistinct density that is similar to normal tissue, whereas on the digital mammogram (right) the cancer (arrow) appears as a dense, irregular mass.

worked-up to resolution with additional imaging and, if indicated, biopsy. Despite the expectation that digital is superior to film, this trial failed to find any difference between the two types of mammograms in terms of breast cancer detection. In fact, film mammography detected more cancers than did digital mammography (34 versus 27), although this difference did not reach statistical significance. Each type of mammography detected cancers missed by the other. Film detected 16 cancers that were not detected by digital, whereas digital detected nine cancers missed on film. Eighteen cancers were detected on

both mammograms. Figs. 6 and 7 show examples of cancers detected on only one device. The only bright spot for digital was that a significantly smaller number of women without cancer were recalled for additional work-ups from the digital study than from the film study. In practice, this translates into less patient anxiety and fewer medical dollars spent on screening. The trial included an analysis, conducted prospectively during the accrual period, of reasons for differences in interpretation of the two sets of mammograms. To perform this task, the two radiolo-

Fig. 7. Case from the Colorado-Massachusetts digital mammography screening trial. Ductal carcinoma in situ is equally visible on both film mammography (left) and digital mammography (right), but was detected only on film mammography. Readers at the time determined the causes of the miss to be an error in detection and poor use of the digital workstation controls during interpretation of the digital mammogram.

J.M. Lewin et al / Radiol Clin N Am 42 (2004) 871 884

877

gists who gave the differing readings evaluated both images side-by-side and decided, by consensus, the major reason for the difference. Along with random differences in the appearance of overlapping normal structures in the breast, caused by slight differences in positioning, the largest causes of discordant readings were because of interpretation factors, as opposed to factors relating to the visibility or conspicuity of the finding on the image. The largest such factor was a minor difference of opinion between the radiologists, but errors of perception, errors of interpretation, and poor use of the digital workstation were also cited. Several of the cancers that were detected only on film were judged to have been missed on digital because of factors relating to interpretation, including errors of perception and poor use of the workstation. Because of this, it is postulated that deficiencies in the workstation included with the prototype system might have contributed to the unexpectedly mediocre clinical performance, considering its technical advantages, of digital mammography in this trial. Similar results, however, were obtained by the more recently completed Oslo I trial, despite the use of a much higher quality commercial digital mammography interpretation workstation [19]. This trial, which enrolled 3683 subjects from the Oslo screening program, was similar to the Colorado-Massachusetts trial in that each subject underwent both film and digital mammography and each of those examinations was interpreted independently of the other. In this trial, however, two radiologists independently interpreted each of the film and digital examinations. As in the Colorado-Massachusetts trial, no statistically significant difference was detected in the cancer detection rate of the two types of mammograms with film mammography again detecting a few more cancers, 28 versus 23 for digital. Again, both film and digital detected cancers missed by the other modality with seven detected only on film and two detected only on digital. No significant difference in recall rate was observed, however, in this trial. After the Colorado-Massachusetts trial failed to show a difference between film and digital mammography, a much larger trial was proposed by a newly formed cooperative group, the American College of Radiology Imaging Network. This trial, modeled on the Colorado-Massachusetts trial and having a similar protocol, began in October, 2001, and finished enrollment of its target of 49,500 subjects in November, 2003. The total cost of the trial is about $26 million. The trial includes four different types of digital mammography devices, but the number of subjects imaged on each type of machine is far from equal and the trial was not designed to distinguish

among them. As in the earlier trials, each subject will be followed for an additional year to ascertain the total number of false-negative mammograms. The earliest results of the trial are not anticipated to be released until the middle of 2005.

Clinical advantages of digital Most of the benefits of digital mammography perceived by its early adopters are not those reported in the literature discussed previously. These include both operational advantages and true advantages in diagnostic ability or confidence. Although not easily measured, these positively impact the patient and her physicians. Digital mammography, like other digital modalities, allows for digital storage and transmission of each study, eliminating lost films and eventually eliminating the need for a film library. Images can be sent electronically to several treating physicians simultaneously, or given to the patient, without any loss of quality. This is an important operational change. At present, however, images cannot generally be sent electronically between institutions because of incompatible systems and security firewalls, so centers still must print out films when patients change to another location and request their prior studies. Even more important to mammography than the operational benefits of a digital image is the elimination of film artifacts, such as dust and the structured noise caused by film processing. Digital also reduces the variability in contrast, density, dose, and exposure time associated with film emulsion and processing. The film processor, in particular, is a major source of variability and requires daily quality assurance to monitor and correct changes. Insufficient contrast is not an issue with digital images and, because the detector is sealed, there are no dust artifacts. From a patients perspective, the biggest advantage of digital is speed. This is especially true in diagnostic studies during which the radiologist reviews each image or pair of images before deciding the next step in the work-up. By not having to wait for films to develop, the length of a diagnostic mammography examination is shortened. Wire localizations are affected more dramatically, because for this procedure the patient must stay in compression while the last image taken is processed and then reviewed by the radiologist. By cutting down the time between exposure and image display to 10 to 20 seconds (from about 3 minutes for film processing including carrying the films to and from the processor), both total procedure time and patient discomfort are

878

J.M. Lewin et al / Radiol Clin N Am 42 (2004) 871 884

markedly decreased. Decreasing examination time has positive effects for the physician, technologist, and radiology administrator, the last benefiting from the efficiencies in terms of technologist and room costs, because examinations can be scheduled more frequently on a digital unit. In most cases, these savings do not completely offset the increased price of the digital equipment, but along with decreased film and film library costs, they do provide a partial justification for going all-digital. Although it has not met expectations in terms of microcalcification detection at screening, digital mammography is ideally suited for microcalcification characterization with magnification or high-resolution views because of the low image noise, the ability to handle a wide variety of exposure settings, and the ability to magnify the image on the screen [12,15]. Two small clinical studies have shown a benefit to digital in terms of depicting and characterizing microcalcifications [20,21]. Low noise is the most important feature for this task because image noise is what most hides subtle calcifications, making them difficult to assess. Wide dynamic range and high contrast allow the image to be acquired at a higher voltage but a shorter exposure time, reducing patient motion. The one disadvantage of digital mammography, that it has lower spatial resolution, is countered by the use of geometric magnification (accomplished by moving the breast away from the detector), which enlarges the projected image of the calcifications above the lower limit of even the lowest spatial resolution digital system. An important point realized early in the dissemination of the technology is that magnification of the image on the monitor alone is no substitute for true geometric magnification or imaging in a special high-resolution mode. Although it had been hoped that digital mammography would eliminate the need to recall patients for special magnification views, this has not turned out to be the case. Because of the large dynamic range, digital mammography is ideal for imaging implants. A single exposure can be displayed to show optimally the details of the implant itself, at the expense of the surrounding breast tissue, or by adjusting the window and level settings, can optimally show the breast tissue, at the expense of seeing into or through the implant. Fig. 8 shows an example of this effect. For the same reason, large dynamic range digital mammography is also ideal for imaging the skin and tissues immediately deep to the skin. This tissue is typically blackened on a well-exposed film mammogram, requiring a special hot light partially to recover the information. Processing of the digital image allows the skin to be evaluated routinely without

Fig. 8. The ability to vary the display parameters, combined with the large dynamic range of the digital detector, allows the same image to be used to examine both the breast implant and the surrounding tissue. (A) Digital mammogram is displayed with optimal parameters to see the breast tissue surrounding the retroglandular silicone implant. (B) The same image can be displayed with a different window and level settings to darken the image and show the detail of the implant. Here we can see that the capsule is extensively calcified and a portion of the implant has herniated through the capsule (arrow).

any extra effort, such as additional windowing or leveling. Although the skin is usually of no importance, it can be thickened in some disease processes, including inflammatory carcinoma. Additionally, special views are sometimes performed to localize indeterminate calcifications to the skin, to prove that they are benign. Digital mammography is ideal for these skin views because it shows the skin and because the multistep process used to obtain these views is similar to that used for wire localization and is accomplished much faster using digital (Fig. 9).

Advanced applications of digital mammography Although digital mammography itself, used in the same manner as film mammography, will likely not revolutionize breast cancer detection and diagnosis, there are advanced applications made possible by the use of digital images that have shown promise in early studies.

J.M. Lewin et al / Radiol Clin N Am 42 (2004) 871 884

879

Computer-aided detection One of the first motivations behind the development of digital mammography was to allow the convenient use of computer-aided detection and computer-aided diagnosis. These technologies help the radiologist find and classify, respectively, lesions on mammograms. Although computer-aided diagnosis has not yet found clinical application, computeraided detection is available commercially and is rapidly disseminating for use with both film and digital mammography. When used with film mammography, the process requires that each film be run through a digitizer before interpretation. This process adds time and expense, because personnel must be hired to perform the digitization. With digital mammography, computer-aided detection can be available at the touch of a button, without any added ancillary expenses. When using a computer-aided detection system, the radiologist first interprets the entire image and then activates the computer-aided detection, which marks areas of suspicion (Fig. 10). The radiologist then examines the marked areas more closely to make sure that nothing was missed. The hope is that computer-aided detection can reduce the

Fig. 9. Calcifications are proved to be contained within the skin on this digital spot compression tangential image. To obtain this view, the mammogram is taken tangentially to a metal bead that has been placed over the calcifications in a standard projection. Digital mammography is ideal for evaluating the skin because of large dynamic range and theability to vary the viewing parameters and magnify the image on the display workstation. Additionally, the procedure of locating the calcifications on the standard view involves several images and is performed much more rapidly with digital mammography.

Fig. 10. Computer-aided detection may be used with either film mammography or digital mammography, but is more convenient with digital because films do not need to be digitized before computer analysis. (A) Digital mediolateral oblique view marked by a commercial computer-aided detection system that uses asterisks to mark masses and triangles to mark calcifications. A 2-cm invasive ductal carcinoma has been correctly marked with an asterisk. An additional calcifications marker (black triangle) is a false-positive because there are no calcifications at that site. (B) On the corresponding craniocaudal view the computer-aided detection system has correctly marked the cancer with no false-positives. The white triangle on both images is a metal skin marker placed to mark the palpable abnormality caused by this cancer. (Courtesy of Kelly McDonough, MD, Breast Imaging of Oklahoma, Edmond, OK.)

880

J.M. Lewin et al / Radiol Clin N Am 42 (2004) 871 884

rate of cancers missed because of errors of observation. The downside is that computer-aided detection systems all have high false-positive rates, marking between one and two normal areas on every film. It is up to the radiologist to dismiss these marks. There are several companies selling FDA-approved devices. A few studies provide indirect evidence of the efficacy of computer-aided detection [22,23] in

increasing mammographic sensitivity. Only one published study was conducted in an actual clinical screening setting. In this trial, involving two radiologists at a high-volume screening setting, computeraided detection increased the number of cancers detected in a year at their center from 41 to 49, a 19.5% increase [24]. Of the eight additional cancers detected by computer-aided detection in this trial,

Fig. 11. (A) Diagrammatic depiction of breast tomosynthesis. Multiple low-dose exposures are taken at different angles. These images are then processed to make tomographic slices. (B) Photograph of General Electric tomosynthesis prototype. (Courtesy of CAPT Jerry Thomas, Uniformed Services University of the Health Sciences, Bethesda, MD.)

J.M. Lewin et al / Radiol Clin N Am 42 (2004) 871 884

881

seven presented as calcifications, consistent with common experience that calcification detection is more susceptible to observational error than is mass detection. As expected, computer-aided detection did increase the recall rate in the practice, but the increase was in proportion to the number of additional cancers detected and was thought to be reasonable. A more recent clinical study, however, compared recall rates

and cancer detection rates for a group of 24 academic radiologists before and after the introduction of computer-aided detection and found no statistically significant difference [25]. Although double-reading is another known way of decreasing missed cancers [26,27], computer-aided detection is thought to be almost as good [28], and is much less expensive. In addition, although Medicare has always refused

Fig. 12. (A) Digital craniocaudal view shows 3-cm and 1-cm masses with partially obscured margins (arrows). (B) Selected slices from a tomosynthesis study better demonstrate two circumscribed masses (arrows), later proved to be cysts by ultrasound. (Courtesy of Hologic, Bedford, MA; with permission.)

882

J.M. Lewin et al / Radiol Clin N Am 42 (2004) 871 884

to pay extra for double reading of a mammogram, it is willing to pay extra for the use of computeraided detection. Tomosynthesis Tomosynthesis involves the mathematic processing of a set of planar images to create tomographic slices, similar in appearance to the linear tomographic images used to evaluate the kidney as part of an excretory urogram. For breast tomosynthesis, multiple mammographic images are acquired at various angles (Fig. 11). Each image is acquired with a fraction of the typical x-ray dose of a mammogram, so that the total radiation dose is equivalent to that of a standard mammogram. The detector can remain stationary or move under the breast; the breast must stay immobile. Breast compression has both advantages and disadvantages with this technique; it is unclear at this point how that trade-off will sort out. The concept of tomosynthesis has been around for decades, but only with the advent of digital mammography has it been able to be applied to the breast [29,30]. There is much excitement about the technique. Because overlapping tissue can both hide a cancer and simulate one, it is a source of both falsepositive and false-negative mammograms. It is hoped that tomosynthesis will be able greatly to reduce these errors. A prototype system made by General Electric and based on their digital detector is being tested on patients at the Massachusetts General Hospital with encouraging (but so far unpublished) early results (D. Kopans, personal communication, 2003). Other manufacturers are developing systems of their own (Fig. 12). One small field-of-view tomographic system is already FDA-approved and commercially available from Instrumentarium (Milwaukee, Wisconsin) for use in diagnostic evaluation of mammographic abnormalities detected on standard mammography [31].

raphy has not had a large impact on its performance, the use of a contrast agent has been proposed to improve its sensitivity to cancers not presenting with calcifications. Because the sensitivity of mammography, or any projection radiography technique, to iodinated contrast agent is small, some sort of subtraction must be used to eliminate unenhanced normal tissue and thereby better show enhancing cancers. Two subtraction methods have been tried: temporal subtraction, in which a postcontrast image is subtracted from a precontrast image; and dual-energy subtraction, in which a low-energy image and a high-energy image, both obtained after iodine injection, are mathematically combined in such a way that unenhanced breast tissue is eliminated but iodine is well seen.. This is possible because of the different x-ray absorption spectra of iodine versus breast tissue. Both temporal subtraction [32] and dual-energy subtraction [33] have been tested in small pilot studies on volunteers with suspicious mammographic or palpable lesions. Both studies used a standard nonionic iodinated contrast agent such as is used clinically for CT. Both techniques performed well demonstrating cancers, with few false-positives. Fig. 13 shows an example

Contrast-enhanced digital mammography Although some cancers do not show up well on mammography because of obscuration by overlying tissue, others are missed even though they are quite visible, simply because they look like normal tissue. MR imaging, with its very high tissue contrast and three-dimensional properties, is superb at showing breast tissue, yet it was learned early on that breast MR imaging performs best when used with an intravenous contrast agent. Similarly, given that the improved contrast resolution of digital mammog-

Fig. 13. A 12-mm invasive lobular carcinoma. (A) Digital mammogram is normal. A metal bead marks a palpable abnormality. (B) Dual-energy CEDSM image shows the cancer as a round enhancing mass in the superior breast (arrow). (From Jong RA, Yaffe MJ, Skarpathiotakis M, Shumak RS, Danjoux NM, Gunesekara A, et al. Contrastenhanced digital mammography: initial clinical experience. Radiology 2003;228:842 50; with permission.)

J.M. Lewin et al / Radiol Clin N Am 42 (2004) 871 884

883

gies; more powerful and better-designed interpretation workstations; and, perhaps most exciting of all, novel advanced applications, such as tomosynthesis and contrast-enhanced mammography, that are not possible with standard film mammography.

References
[1] Fletcher SW, Black W, Harris R, Rimer BK, Shapiro S. Report of the International Workshop on Screening for Breast Cancer. J Natl Cancer Inst 1993;85: 1644 56. [2] Nystrom L, Rutqvist LE, Wall S, et al. Breast cancer screening with mammography: overview of Swedish randomised trials. Lancet 1993;341:973 8. [3] Tabar L, Fagerberg G, Duffy SW. Update of the Swedish two-county program of mammographic screening for breast cancer. Radiol Clin North Am 1992;30:187 210. [4] Poplack SP, Tosteson AN, Grove MR, Wells WA, Carney PA. Mammography in 53,803 women from the New Hampshire mammography network. Radiology 2000;217:832 40. [5] Linver MN, Paster SB, Rosenberg RD, Key CR, Stidley CA, King WV. Improvement in mammography interpretation skills in a community radiology practice after dedicated teaching courses: 2-year medical audit of 38,633 cases. Radiology 1992;184:39 43. [6] Rosenberg RD, Hunt WC, Williamson MR, et al. Effects of age, breast density, ethnicity, and estrogen replacement therapy on screening mammographic sensitivity and cancer stage at diagnosis: review of 183,134 screening mammograms in Albuquerque, New Mexico. Radiology 1998;209:511 8. [7] Baker LH. Breast Cancer Detection Demonstration Project: five-year summary report. CA Cancer J Clin 1982;32:194 225. [8] Bird RE. Low-cost screening mammography: report on finances and review of 21,716 consecutive cases. Radiology 1989;171:87 90. [9] Shtern F. Digital mammography and related technologies: a perspective from the National Cancer Institute. Radiology 1992;183:629 30. [10] Suryanarayanan S, Karellas A, Vedantham S, Ved H, Baker SP, DOrsi CJ. Flat-panel digital mammography system: contrast-detail comparison between screenfilm radiographs and hard-copy images. Radiology 2002;225:801 7. [11] Cooper III VN, Oshiro T, Cagnon CH, Bassett LW, McLeod-Stockmann TM, Bezrukiy NV. Evaluation of detector dynamic range in the x-ray exposure domain in mammography: a comparison between film-screen and flat panel detector systems. Med Phys 2003;30: 2614 21. [12] Hermann KP, Obenauer S, Funke M, Grabbe EH. Magnification mammography: a comparison of fullfield digital mammography and screen-film mammog-

Fig. 14. A 61-year-old-woman with multicentric invasive lobular carcinoma demonstrated with dual-energy contrastenhanced digital subtraction mammography in two projections. (A) Mediolateral oblique view. (B) Craniocaudal view. High-density masses are foci of cancer throughout the breast. Nonenhancing oval mass (*) is a large cyst. The cancer (arrowheads) surrounds the cyst in the lateral breast, as shown on the mediolateral oblique view, but is also present in the posterocentral and medial breast, as shown on the craniocaudal view (arrowheads).

of a mammographically occult cancer demonstrated with dual-energy contrast-enhanced digital subtraction mammography. The advantage of the dualenergy technique is that the breast does not have to remain immobilized during and after the contrast agent injection. This allows multiple projections to be obtained (Fig. 14) versus the single projection possible with temporal subtraction.

Summary Despite its technical advantages, early clinical trials comparing digital mammography with film mammography for screening have been somewhat disappointing. This result is caused in part by the many years of experience in optimizing and interpreting film mammography. Digital mammography, in comparison, is in its infancy and can be expected to improve more rapidly than film mammography. Some areas of improvement being observed now include the development of new detector technolo-

884

J.M. Lewin et al / Radiol Clin N Am 42 (2004) 871 884 raphy for the detection of simulated small masses and microcalcifications. Eur Radiol 2002;12:2188 91. Undrill PE, OKane AD, Gilbert FJ. A comparison of digital and screen-film mammography using quality control phantoms. Clin Radiol 2000;55:782 90. Berns EA, Hendrick RE, Cutter GR. Performance comparison of full-field digital mammography to screen-film mammography in clinical practice. Med Phys 2002;29:830 4. Rong XJ, Shaw CC, Johnston DA, Lemacks MR, Liu X, Whitman GJ, et al. Microcalcification detectability for four mammographic detectors: flat-panel, CCD, CR, and screen/film. Med Phys 2002;29:2052 61. Krupinski E, Roehrig H, Furukawa T. Influence of film and monitor display luminance on observer performance and visual search. Acad Radiol 1999;6(7):411 8. Lewin JM, Hendrick RE, DOrsi CJ, Isaacs PK, Moss LJ, Karellas A, et al. Comparison of full-field digital mammography to screen-film mammography for cancer detection: results of 4945 paired examinations. Radiology 2001;218:873 80. Lewin JM, DOrsi CJ, Hendrick RE, Moss LJ, Isaacs PK, Karellas A, et al. Clinical comparison of full-field digital mammography to screen-film mammography for breast cancer detection. AJR Am J Roentgenol 2002;179:671 7. Skaane P, Young K, Skjennald A. Population-based mammography screening: comparison of screen-film and full-field digital mammography with soft-copy readingOslo I study. Radiology 2003;229:877 84. Fischer U, Baum F, Obenauer S, Luftner-Nagel S, von Heyden D, Vosshenrich R, et al. Comparative study in patients with microcalcifications: full-field digital mammography vs screen-film mammography. Eur Radiol 2002;12:2679 83. Diekmann S, Bick U, von Heyden H, Diekmann F. Visualization of microcalcifications on mammographies obtained by digital full-field mammography in comparison to conventional film-screen mammography. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 2003;175:775 9. Warren Burhenne LJ, Wood SA, DOrsi CJ, Feig SA, Kopans DB, OShaughnessy KF, et al. Potential contribution of computer-aided detection to the sensitivity of screening mammography. Radiology 2000;215: 554 62 [erratum in: Radiology 2000;216:306]. [23] Brem RF, Baum J, Lechner M, Kaplan S, Souders S, Naul LG, et al. Improvement in sensitivity of screening mammography with computer-aided detection: a multiinstitutional trial. AJR Am J Roentgenol 2003;181: 687 93. [24] Freer TW, Ulissey MJ. Screening mammography with computer-aided detection: prospective study of 12,860 patients in a community breast center. Radiology 2001;220:781 6. [25] Gur D, Sumkin JH, Rockette HE, Ganott M, Hakim C, Hardesty L, et al. Changes in breast cancer detection and mammography recall rates after the introduction of a computer-aided detection system. J Natl Cancer Inst 2004;96:185 90. [26] Warren RM, Duffy SW. Comparison of single reading with double reading of mammograms, and change in effectiveness with experience. Br J Radiol 1995;68: 958 62. [27] Thurfjell E, Lernevall K, Taube A. Benefit of independent double reading in a population-based mammography screening program. Radiology 1994; 191:241 4. [28] Karssemeijer N, Otten JD, Verbeek AL, Groenewoud JH, de Koning HJ, Hendriks JH, et al. Computer-aided detection versus independent double reading of masses on mammograms. Radiology 2003;227:192 200. [29] Niklason LT, Christian BT, Niklason LE, Kopans DB, Castleberry DE, Opsahl-Ong BH, et al. Digital tomosynthesis in breast imaging. Radiology 1997; 205:399 406. [30] Suryanarayanan S, Karellas A, Vedantham S, Glick SJ, DOrsi CJ, Baker SP, et al. Comparison of tomosynthesis methods used with digital mammography. Acad Radiol 2000;7:1085 97. [31] Lehtimaki M, Pamilo M, Raulisto L, et al. Diagnostic clinical benefits of digita1 spot and digital 3D mammography following analysis of screening findings. SPIE Medical Imaging 2003;5029:698 706. [32] Jong RA, Yaffe MJ, Skarpathiotakis M, Shumak RS, Danjoux NM, Gunesekara A, et al. Contrast-enhanced digital mammography: initial clinical experience. Radiology 2003;228:842 50. [33] Lewin JM, Isaacs PK, Vance V, Larke FJ. Dual-energy contrast-enhanced digital subtraction mammography: feasibility. Radiology 2003;229:261 8.

[13]

[14]

[15]

[16]

[17]

[18]

[19]

[20]

[21]

[22]