European Radiology

, Volume 17, Issue 8, pp 1931–1942

Digital mammography: what do we and what don’t we know?

Open Access
Breast

Abstract

High-quality full-field digital mammography has been available now for several years and is increasingly used for both diagnostic and screening mammography. A number of different detector technologies exist, which all have their specific advantages and disadvantages. Diagnostic accuracy of digital mammography has been shown to be at least equivalent to film-screen mammography in a general screening population. Digital mammography is superior to screen-film mammography in younger women with dense breasts due to its ability to selectively optimize contrast in areas of dense parenchyma. This advantage is especially important in women with a genetic predisposition for breast cancer, where intensified early detection programs may have to start from 25 to 30 years of age. Tailored image processing and computer-aided diagnosis hold the potential to further improve the early detection of breast cancer. However, at present no consensus exists among radiologists on which processing is optimal for digital mammograms. Image processing may also vary significantly among vendors with so far limited interoperability. This review aims to summarize the available information regarding the impact of digital mammography on workflow and breast cancer diagnosis.

Keywords

Digital mammography Breast cancer screening Image processing Workflow Quality assurance 

Introduction

The concept of digital mammography with all its advantages including easier image storage and tailored image processing is not new. As early as 1967, Fred Winsberg from the University of Chicago proposed an algorithm for computer-aided breast cancer detection using digitized film mammograms [1]. Although film digitization can produce high-quality digital mammograms, this process is not only labor intensive and expensive, but primarily does not solve the limitations inherent to film mammography, mainly the narrow dynamic range caused by the non-linear characteristic curve of film. Initial experiments with digital mammography based on computed radiography (CR) started in the late 1980s [2]. However, image quality was inferior to state-of-the-art film-screen mammography at comparable dose due to the poor detective quantum efficiency (DQE) of these early CR systems [3]. For several years now, newer dedicated mammography CR systems as well as a number of different integrated full-field digital mammography systems have become available [4, 5, 6], many of which have received regulatory approval (Table 1). Increasingly these systems are replacing film-screen mammography both for screening as well as for diagnostic mammography. This review will summarize the current evidence on the advantages and disadvantages of digital mammography.
Table 1

Digital mammography systems

Description

Detector type

Detector material

Signal generation

Pixel size

Image area

Pixel matrix

Name

Manufacturer (distributors)

Comments

FDA approvala

CCD mosaic

Integrated (area)

CsI:Tl phosphorscintillator

CCD array(3 × 4 mosaic)

41 μm

18.6  ×  24.8 cm

6,400 × 4,800

Digital Breast Imager

Trex (Lorad, Bennett)b

No longer available

15.3.2002

CCD slot scanning

Integrated (scanning)

CsI:Tl phosphorscintillator

1.4-cm wide array of 4 CCDs

54 μmc

22.1 × 30.4 cm

4,096 × 5,625

Senoscan

Fischer (Philips)d

No grid necessary

25.9.2001

Phosphor flat panel

Integrated (area)

CsI:Tl phosphorscintillator

Array of photo diodes/TFT

100 μm

19.2 × 23 cm

1,914 × 2,294

Senographe 2000D

General Electric

Images can be obtained in rapid sequence

28.1.2000

Senographe DS

General Electric

19.2.2004

24 × 30.7 cm

2,394 × 3,062

Senographe Essential

General Electric

11.4.2006

Selenium flat panel

Integrated (area)

Amorphous selenium

Array of electrode pads/TFT

70 μm

25 × 29 cm

3,584 × 4,096

Selenia

Lorad/Hologic(Agfa)

Direct conversion of X-ray photons to electric charge

2.10.2002

Novation

Siemens (Agfa)

20.8.2004

85 μm

17.4 × 23.9 cm or 23.9 × 30.5 cm

2,016 × 2,816 or 2,816 × 3,584

Giotto IMAGE

IMS

 

Nuance

Planmed

 

Photon counter

Integrated (scanning)

Silicon strip

Array of X-ray photon counters

50 μm

24 × 26 cm

4,800 × 5,200

Micro-Dose Mammography

Sectra

Very high DQE, currently no AEC

 

Computed radiography (CR)

Cassette-based(area)

Photostimulable phosphor

Laser scanning

50 μm

18 × 24 cm or 24 × 30 cm

Approx. 1,800 × 2,400 or 2,400 × 3,000

FCR Profect CSFCR 5000 MA

Fuji(Siemens, Philips)

Dual-sided readout

10.7.2006

CR 75.0 CR 85-X

Agfa

  

CR 975

Kodak

  

REGIUS 190 Mammo

Konica Minolta

  

Cs:Tl = thallium-activated cesium iodide; TFT = thin film transistor; CCD = charge-coupled device, AEC = automatic exposure control

aInformation current as of 25 September 2006 [7]

bNo longer produced, all installations replaced by Lorad/Hologic Selenia systems

cAdditional high-resolution mode with pixel size of 27 μm available with limited field of coverage

dIntellectual property bought by Hologic in 2005, existing installations remain in service, however, no new systems are produced or distributed

eThese systems are sometimes also called amorphous silicon systems, because the detector is mounted on an amorphous silicon substrate. However, this name is misleading, since this is true also for the selenium flat-panel systems.

fExact pixel matrix varies slightly among vendors

Measuring image quality in digital mammography

Image quality of conventional film-screen systems can be fairly accurately described by three parameters: (1) the characteristic curve of a film-screen system, (2) the sensitivity or “speed” and (3) the high-contrast line-pair resolution. All three concepts do not apply to digital imaging.

Digital mammography detectors have a linear relationship between detector dose and signal intensity, and no fixed characteristic curve as in film-screen mammography exists. Translation of detector signal intensities into monitor brightness is achieved by specific window settings and non-linear look-up tables, which can be modified to optimize the contrast in a certain image area of interest [8, 9].

While the sensitivity of a film-screen system defines the amount of dose required to reach a certain optical density of the developed film, there is no single optimal detector dose in digital mammography. With decreasing detector dose, image noise increases in the digital image and vice versa. By changing the X-ray beam energy spectrum, image noise can be exchanged against image contrast while keeping the parenchymal dose to the patient constant. In digital mammography, it is often beneficial to move to a higher energy spectrum than with film-screen mammography, since image noise is lower and the resulting loss in image contrast can be compensated for by adjusting the window setting [10, 11].

Due to the continuous course of the modulation transfer function (MTF) in film-screen mammography, the high-contrast line-pair resolution can be used to accurately predict the performance of the system at lower frequencies. At least with modern flat-panel digital mammography, the MTF abruptly declines at the nyquist limiting frequency defined by the pixel size of the detector [12, 13]. This results in a nominally lower spatial (line-pair) resolution, although the MTF at lower (clinically more relevant) frequencies may be significantly higher than with film-screen mammography. For digital mammography, line-pair resolution is therefore meaningless, and the performance of the system is better described by the so-called contrast-detail resolution, the ability to visualize object details of a certain size and radiation contrast. Contrast-detail resolution for a given dose is in turn determined by the DQE of the digital detector. A system with a higher DQE will reach the same contrast-detail resolution at a lower dose than an otherwise similar system with a lower DQE (Fig. 1). This can be used in clinical practice, where integrated digital mammography systems with a high DQE are usually operated at a mean glandular dose 20–30% lower than that of film-screen mammography [16, 17, 18]. Since the DQE of a digital system is difficult to measure in a standardized manner and may depend on a variety of factors such as X-ray beam quality and detector dose [19], acceptance testing for digital mammography systems is usually achieved by assessing the contrast-detail resolution of a system with certain phantoms such as the CDMAM phantom [14, 20]. However, contrast-detail phantoms such as the CDMAM phantom showing objects on a uniform background may not be ideal to predict the performance of a system in clinical practice. Such phantoms tend to favor digital systems by overestimating the detection performance for larger low-contrast objects due to unrealistically narrow window settings used when viewing the digital images, made possible by the lack of background structure [21]. By adding a structured background to the CDMAM phantom, Grosjean and Muller were able to show that while visibility of small details <0.4 mm was still limited by noise sources related to the image acquisition process, detection of larger low-contrast objects was mainly determined by the structured background [22]. This is in keeping with results from earlier experimental work based on digitized film [23, 24] and matches the experience from clinical practice. While adequate detector dose and low quantum noise levels are necessary to adequately show microcalcifications, visibility of larger masses is much less affected by image noise associated with lower dose or inferior DQE of the detector (Fig. 2). This effect can be used to significantly reduce the parenchymal dose for additional localization views, e.g., as part of interventional procedures in which the presence of a certain abnormality is already known (Fig. 3) [25, 26].
Fig. 1

Relationship between mean glandular dose at 6-cm compressed breast thickness and radiation threshold contrast for 0.1-mm details with a CR (Fuji Profect) and a flat panel (Siemens Novation) digital mammography system. Data points outside the grey acceptable/achievable area do not fulfill the minimum criteria for contrast-detail resolution or are associated with a mean glandular dose above the maximum acceptable dose level as defined by the European guidelines for quality assurance in breast cancer screening and diagnosis [14]. Results are averaged from three human readers each scoring four CDMAM phantom images. Threshold contrast values shown are nominal at 28 kV Mo/Mo, actual exposure parameters were 26 kV Mo/Rh for the CR system and 28 kV W/Rh for the flat-panel system. Compared to the CR system, the flat-panel system can alternatively be operated at the same dose with higher image quality or at the same image quality with lower dose. Source: Young et al. [15]

Fig. 2

Influence of various degrees of superimposed random Gaussian noise (increasing from left to right) on visibility of objects with different size. The large low-contrast object (top row) simulating a mass is almost unaffected by the image noise and is well seen on all images. The five smaller objects (bottom row) simulating microcalcifications are rapidly lost with increasing image noise

Fig. 3

Example of a breast interventional procedure using reduced dose mammographic images. A 76-year-old patient with bifocal invasive-ductal carcinoma surrounded by high-grade DCIS. Normal-dose mammographic image (a) and needle localization image at 50% reduced dose (b) with enlarged area of microcalcifications (c,d) in the vicinity of the main tumor. Both masses are equally well seen on the reduced dose image during the localization procedure. However, the individual microcalcifications (arrows) are less well depicted on the lower dose image (d) than on the normal dose image (c) due to a slightly higher amount of image noise in the lower dose image

Digital mammography systems

There are now several different types of digital mammography systems available, which all are capable of producing high-quality digital mammograms, but all have specific advantages and disadvantages. Digital mammography systems can be grouped according to detector material, whether they are integrated or cassette-based systems, or whether they use an area detector or slot scanning technique (Table 1). Integrated systems usually allow for a higher throughput than cassette-based CR systems, but are more expensive. Slot scanning systems often can operate at a lower dose, since the slot collimation is effective in reducing scatter radiation, thus obliviating the need for an additional anti-scatter grid. Disadvantages of the slot scanning systems include longer scan times, high tube strain and the need for exact mechanical registration of the moving collimation slot and the detector. There are two different basic types of integrated area detectors, one on the bases of a phospor scintillator combined with an array of photo diodes capturing the light generated by the phosphor layer and the second type using an amorphous selenium layer with direct conversion of the X-ray photons to an electric charge. Both systems use a TFT array mounted on a amorphous silicon base for signal read-out. Detector elements in the phosphor flat-panel systems with a pixel size of 100 μm are usually slightly larger than in the selenium-based systems (70–85 μm). Reduction in detector element size in phosphor flat-panel systems is difficult, since with a decreasing size of the detector elements, the relative portion of the active detector area would decrease rapidly compared to the relatively fixed inactive portion of the detector related to the signal read-out, resulting in a lower DQE and higher parenchymal dose for the patient. An advantage of phosphor flat-panel systems is that images can be obtained in relatively short sequence, which is useful for patient throughput and advanced applications such as tomosynthesis and contrast-enhanced mammography [27, 28]. A disadvantage of selenium-based systems compared to phosphor flat-panel systems is the higher amount of image lag (image signal carried over from a previous to a subsequent exposure) and ghosting (temporary change in sensitivity base on prior exposure history). However, detector development in this area is ongoing, and both lag and ghosting have been reduced significantly in newer clinical selenium-based systems [29].

The value of CR mammography systems compared to integrated full-field systems has recently been under intense discussion. One major selling point of CR systems is the lower investment cost, especially if existing mammography equipment can be used for acquiring the mammographic images. This cost advantage, however, is significantly smaller when considering a new installation. One argument brought forward against CR mammography is that the dose necessary to operate CR systems at acceptable image quality levels is higher than that of integrated full-field systems. Although there is no doubt that CR systems have a slightly lower DQE than integrated full-field systems, part of this dose disadvantage may be explained by other factors. Integrated systems usually optimize the entire imaging chain including choice of the exposure parameters such as kVp and the anode/filter combination. It has been shown that the major dose savings with digital mammography systems are achieved in patients with larger breasts by switching to a higher energy beam spectrum earlier than with conventional film-screen systems [16, 30]. Since CR mammography systems are used with standard mammography equipment traditionally designed for film-screen mammography, this optimization of exposure parameters often does not occur. Another common problem with CR mammography is that imaging processing algorithms developed for other radiographic exams (e.g., chest films) are used. Image noise with digital images is higher in areas of lower detector dose, e.g., the mediastinum. Since this noise may be perceived as disturbing, special processing algorithms have been developed for CR images to suppress noise in bright (underexposed) image areas [31]. In mammography, such algorithms will lead to impaired visibility of microcalcifications in areas of dense parenchyma and should therefore not be used.

Clinical comparison of digital and film-screen mammography

Early clinical studies comparing digital mammography with film-screen mammography were inconclusive (Table 2). None of the clinical trials so far has demonstrated significant differences in detection performance in a general screening population between film-screen and digital mammography. While in the study of Lewin et al. [33], the recall rate with digital mammography was significantly lower than with film-screen mammography, both the Oslo I and II studies found a higher recall with digital mammography [34, 36]. These results are difficult to compare, since the recall rates in the US are in general much higher than in European screening programs (Table 2).
Table 2

Prospective clinical screening trials comparing film-screen and digital mammography

 

Study design

Number of sites

Digital system

Age(years)

 

Number of exams

Recall rate

Cancer detection rate

ppv

Lewin [32, 33]

Paired, single-reading

2

GE phosphorflat panel prototypea

>40

Film-screen

6,736

14.9%*

4.9‰

3.3%

Digital

6,736

11.8%*

4.0‰

3.4%

OSLO I [34, 35]

Paired, double-reading with consensus

1

[GE Senographe 2000D]

50–69

Film-screen

3,683

3.5%

7.6‰

21.9%

Digital

3,683

4.6%

6.2‰

13.7%

OSLO II [36]

Randomized,double-reading with consensus

1

GE Senographe 2000D

50–69

Film-screen

10,304

2.5%*

5.4‰

22.1%

Digital

3,985

3.8%*

8.3‰

21.6%

45–49

Film-screen

7,607

3.0%*

2.2‰

7.4%

Digital

3,012

3.7%*

2.7‰

7.1%

DMIST [37, 38]

Paired, single-reading

33

GE Senographe 2000 D (45%)Fischer Senoscan (23%)Fuji FCR (22%)Lorad Digital Breast Imager and Hologic Selenia(together around 10%)b

all

Film-screen

42,745

8.4%

4.0‰

5%

Digital

42,570

8.4%

4.3‰

5%

< 50

Film-screen

14,355

10%

2.2‰*

2%

Digital

14,355

10%

3.3‰*

3%

*Differences statistically significant

aPrototype predecessor of the Senographe 2000D (General Electric) using the same phosphor flat-panel detector.

bDuring the course of the trial, the Lorad/Trex Digital Breast Imager units were all replaced by Lorad/Hologic Selenia systems. Exam numbers for both systems are not specified separately

One reason for the variable results of clinical mammography trials is that differences in positioning and reader performance far outweigh any difference in the acquisition technique, be it between screen-film and digital mammography or between different digital mammography systems [39]. This is easily demonstrated by the fact that in paired screening trials with two separate mammographic exams obtained at the same time (one film-screen and one digital) the number of detected cancers increases by 30% and more (Table 3), just by obtaining a second set of mammographic images independently read by one or more additional radiologists, while differences in cancer detection between digital and film-screen mammography on the whole are negligible.
Table 3

Impact of double examination on cancer detection in paired screening trials

 

Age (years)

Number of exams

Number of cancersa detected by mammography

Gain by adding the second modalityb

All

Film-screen

Digital

Both

Film-screen only

Digital only

Lewin [33]

>40

6,736

42

33 (78.6%)

27 (64.3%)

18 (42.9%)

15 (+55.6%)

9 (+27.3%)

OSLO I [34]

50–69

3,683

31

28 (90.3%)

23 (74.2%)

20 (64.5%)

8 (+34.8%)

3 (+10.7%)

DMIST [35]

All

42,555c

237

174 (73.4%)

185 (78.1%)

122 (51.5%)

52 (+28.1%)

63 (+36.2%)

≥50

28,200

183

142 (77.6%)

137 (74.9%)

96 (52.5%)

46 (+33.6%)

41 (+28.9%)

<50

14,355

54

32 (59.3%)

48 (88.9%)

26 (48.1%)

6 (+12.5%)

22 (+68.8%)

aAll breast malignancies including invasive and in-situ breast cancers

bPercentage values are relative to the number of cancers found by the other modality

cExcluding 205 women who underwent only one type of mammography exam

Digital mammography with the possibility to locally optimize image contrast has, however, a clear advantage in younger patients with dense breasts, as was impressively demonstrated by the Digital Mammographic Screening Trial (DMIST) [38]. Interestingly, the rapid decline in sensitivity as typically seen with film-screen mammography in denser breasts [40] was not observed with digital mammography in the DMIST trial, where the sensitivity of digital mammography in the subgroup of women with dense breasts was identical to the sensitivity in the entire group [38]. This advantage of digital mammography in women with dense breasts will be especially valuable in patients with a genetic predisposition for breast cancer, in whom intensified early detection measures including mammography may have to start as early as 25 to 30 years of age [41, 42]. However, it is uncertain whether the DMIST results can be translated into the European situation, where screening mammography exams are usually double-read and recall rates are much lower. Per Skaane in the Oslo II study at a recall rate of 3.7% for digital and 3.0% for film-screen mammography (compared to around 10% for women <50 years of age in the DMIST trial) found a much smaller, statistically not significant advantage for digital mammography in women below the age of 50 (Table 2). To be able to take full advantage of digital mammography in women with dense breasts, it may therefore be necessary to aggressively recall even subtle findings, so-called “minimal signs” as defined in the Dutch screening program [43]. In European population-based screening programs, however, there is a tendency to initially ignore these minimal signs in order to keep the recall rate at an acceptable low level [43].

Microcalcifications

For a long time, the question whether digital mammography with a spatial resolution lower than film-screen mammography systems can adequately visualize small microcalcifications has been at the center of an intense debate. In theory, digitization with a limited spatial resolution may impair visualization of small details in two ways. Objects smaller than the pixel size of a digital detector will be shown larger and with lower contrast. In addition, the shape information of small objects may be lost, since objects slightly larger than the pixel size will be depicted by a few square pixels [4, 44]. However, both experimental studies [45, 46] as well as clinical trials [47, 48, 49] have shown this to be irrelevant both for detection and characterization of microcalcifications. This is due to the fact that in overview (unmagnified) mammographic images, only microcalcifications larger than approximately 130 μm can be detected [2, 50]. With these small microcalcifications just at the detection threshold, also with film-screen mammography no real shape information is discernible due to screen unsharpness, scatter radiation and geometric blur associated with the larger focus. Although on average there may be no differences in the depiction of microcalcifications between film-screen and digital mammography, both systems may have advantages and disadvantages in certain patient populations. Integrated digital systems with a high DQE imaged at sufficient dose may be superior to film-screen mammography in depicting microcalcifications in dense parenchyma due to higher contrast. This is not true for CR systems, which at clinically acceptable dose levels have a relatively high noise level in areas of dense parenchyma, limiting the visualization of subtle microcalcifications. While in general the lower spatial resolution of digital mammography will not play a role in clinical practice, digital mammography may be at a slight disadvantage in older patients with small and transparent breasts, in whom film-screen mammography may depict details smaller than the usual visibility threshold of around 130 μm. Although not analyzed separately, there may have been a slight advantage for film-screen mammography in the DMIST trial in patients ≥50 years of age and with transparent breasts [51, 52], which could support this hypothesis.

Some authors have suggested that with digital mammography fewer magnification views may be necessary [53]. Based on theoretical considerations and our own clinical experience with digital mammography for more than 7 years now, this theory cannot be supported. Electronic magnification (zooming) of digital mammograms contains less rather than more additional information compared to using a strong magnifying glass with high spatial resolution film-screen mammography. With digital mammography systems, due to the limited spatial resolution, small microcalcifications will be depicted by just a few individual pixels. Electronic magnification can be done in two ways, pixel replication or interpolation. When pixel replication is used, small microcalcifications will always be shown with a ragged border due to the blown-up square pixels. Both with bilinear and bicubic interpolation, the two most common forms, small microcalcifications will always appear round on the electronically magnified images. In both cases, no relevant additional information is provided by the electronic magnification other than that the microcalcifications are easier to see. Both with conventional film-screen and digital mammography, additional small-focus spot views with geometric (e.g., ×1.8) magnification are necessary for more detailed analysis of microcalcifications. Due to the higher spatial resolution related to the magnification as well as the reduced geometric unsharpness offered by the smaller focus, true geometric magnification views will show additional smaller microcalcifications not seen on the overview mammogram and the shape of the individual microcalcifications will be more clearly depicted (Fig. 4). Both with film-screen and digital mammography, it may sometimes be difficult to reliably distinguish subtle amorphous microcalcifications at the detection threshold from image noise. Again, electronic magnification will not help in this situation, since both noise as well as microcalcifications will be shown enlarged, and true geometric magnification views will be necessary to resolve this question.
Fig. 4

Impact of true geometric magnification on the visibility of microcalcifications. A 52-year-old patient with a suspicious cluster of microcalcifications on mammography (diagnosis: high-grade DCIS). Electronic magnification of the overview digital mammogram using pixel replication (a) and bicubic interpolation (b) compared to the geometric ×1.8 spot magnification view (c). The shape of the individual microcalcifications is much better defined on the geometric magnification view, in addition several additional smaller microcalcifications are seen. The images shown correspond to an area of approximately 1 × 1 cm

Impact on workflow

Introduction of digital imaging into mammography has significant workflow implications. Most of the advantages of digital mammography are related to getting rid of film. With integrated digital systems, the lack of film cassette handling allows for a higher patient throughput [54] and lets the technologist concentrate more on the patient. Especially interventional procedures such as preoperative wire localizations are much faster with integrated digital systems without the need for films to be developed between each step of the procedure. Digital images can automatically be transferred, stored and retrieved without the need for human interaction. There are no lost or misplaced films and digital images can be viewed by several different people at the same time. Film library space and personnel are freed up, and the higher investment costs for digital mammography are at least in part compensated for by these savings [55]. When considering the impact of digital mammography on the reading of mammographic studies, the picture is less clear-cut. There is no doubt that images acquired digitally should best be read as soft-copy on a monitor. Only in this way can the main advantages of digital mammography such as tailored image processing and contrast optimization be harvested. However, depending on detector area and pixel size, digital mammograms may have an image matrix of up to 4,800 × 6,000 pixels with a file size of more than 50 MByte. These images are too large to be displayed at full 1:1 resolution on a high-resolution 5-megapixel monitor, the current standard for mammography review workstations. Both hard- and software of mammography review stations has improved significantly over the last few years. Dedicated mammography review stations now allow to switch almost instantaneously between different image layouts including a sequential magnified quadrant zoom, with softcopy reading speed approaching or even surpassing that of batch film reading [56, 57]. Viewing the entire image piece by piece in full resolution may be tedious, but necessary to ensure that no microcalcifications are overlooked. There has been some discussion recently about the use of computer-aided diagnosis (CAD) techniques, which have a near perfect sensitivity for microcalcifications of more than 98%, as a preprocessing tool, only showing those areas of the image in full resolution to the radiologist, where the CAD system has detected possible microcalcifications [58]. This concept has the potential for significantly speeding up softcopy reading of digital mammograms.

Different opinions exist on how to handle prior mammographic films when reading digital mammogram soft-copy. Digitization of prior films is expensive, and image characteristics of digitized films are different from primary digital mammograms, making direct comparison difficult. Keeping a film viewer next to the computer workstation is not only cumbersome, but light from the view box may interfere with the image display on the monitor. Some groups have therefore decided with softcopy reading not to offer prior mammographic films during the primary reading session, but only in the consensus conference in case abnormal findings exist on the current exam requiring comparison with older exams [36].

Another limitation of softcopy reading is related to the lower maximum contrast of monitors compared to viewing film in front of a high luminance alternator. Most vendors try to compensate for this disadvantage of monitor reading by specialized non-linear image processing, compressing the dynamic range of the mammographic image so that the entire breast from the chest wall to the skin can be viewed simultaneously at maximum contrast [59, 60]. Key to this technique is a so-called thickness compensation or density equalization, which increases the brightness of the dark peripheral breast areas closer to the skin to match the brightness of the central parts of the breast (Fig. 5). In digital mammography, a variety of other image processing algorithms are available, e.g., to enhance the conspicuity of certain relevant findings such as masses or microcalcifications. Image processing may vary substantially among different digital mammography vendors, even endangering interoperability between mammography review workstations. Not surprisingly, there is also no agreement on the optimal image processing among radiologists [61], and care has to be taken not to prolong reading times by switching too much between different image processing and window settings.
Fig. 5

Peripheral density correction in digital mammography. A 57-year-old patient with low-grade DCIS. The relatively superficial lesion with internal microcalcifications is only partially seen on the raw unprocessed image (a), which is similar to conventional film-screen mammography. On the processed image (b) the skin and subcutaneous tissue are shown with the same contrast and brightness as the central areas of the breast, and the lesion is shown in its entirety

Computer-aided diagnosis

Computer-aided diagnosis (CAD) can be defined as a diagnosis made by a radiologists taking into account the computer output as a second opinion, similar to the use of a spell-checker program. The concept of computer-aided diagnosis in mammography has now been around for more than 40 years [1]. Much of the earlier research as well as the first FDA-approved clinical system introduced in 1998 were based on digitized film-screen mammograms. However, integration of CAD into the workflow is much easier with digital mammography, where the CAD output can be shown directly on the mammography review workstation [62]. Although mammography CAD systems have received wide-spread adoption in the US, where there is additional reimbursement for the use of CAD, the clinical value of CAD is still being debated [63]. Several clinical studies have demonstrated that by using CAD more and smaller cancers can be detected, usually at the expense of a slightly higher recall rate [64, 65, 66]. The usefulness of CAD will vary with the mammography experience of the reader, and thorough training in the use of the CAD system will improve results [63]. One major problem of CAD is the still relatively high number of false-positive computer marks, on average between one and two per case [67], which means that in a screening situation often less than 1 in 100 computer prompts actually represents cancer (Table 4). It is important to realize that the sensitivity of CAD systems is different for masses and microcalcifications. While microcalcifications can be reliably detected by the computer with a sensitivity of more than 98%, the sensitivity of CAD systems for mammographic masses is significantly lower [68]. Even with high sensitivity/low specificity settings, current CAD systems miss around 10% of masses, which can be detected by a human observer [67]. Due to the very high sensitivity for microcalcifications, CAD systems may in the future be used to improve the reading workflow of digital mammograms. Radiologists would no longer need to look at the entire mammographic image at full resolution, something that depending on the pixel matrix of the image may take a long time, but would only need to look in full resolution at areas with possible microcalcifications detected by the CAD system [58]. Other future applications of CAD include prescreening, where a radiologist would no longer need to look at all at a mammogram with no computer-detected abnormalities, and double reading with CAD, where one of the human readers in the double reading process would be replaced by the computer.
Table 4

Positive predictive value of CAD marks in a screening setting

Average number of CAD marks per normal case

Positive-predictive value (ppv)a

Number of positive CAD marks/abnormal readings to detect one cancer

5

0.001

1,000

1b

0.005

200

0.1

0.05

20

Radiologistc

0.1–0.5

2–10

CAD = computer-aided diagnosis

aAssuming a cancer detection rate of 5 per 1,000 screening exams

bPerformance of current commercial CAD systems

cBased on a range of radiologists’ recall rates between 1% and 5%

Conclusion

Digital mammography has established itself as a true alternative to film-screen mammography offering significant workflow improvements, and there is no doubt that in the long run digital mammography will replace film-screen mammography. Although in general the diagnostic accuracy of digital mammography is similar to that of film-screen mammography, digital mammography may have specific advantages in younger women due to the possibility to selectively enhance image contrast in areas of dense parenchyma. Digital mammography enables an array of advanced applications such as contrast-enhanced mammography, tomosynthesis and computer-aided diagnosis, although the value of these new techniques in clinical practice has yet to be shown. Future efforts should aim to further optimize image acquisition parameters in digital mammography resulting in the lowest possible radiation exposure to the breast as well as to improve and standardize image processing techniques.

References

  1. 1.
    Winsberg F, Elkin M, Macy J, Bordaz V, Weymouth W (1967) Detection of radiographic abnormalities in mammograms by means of optical scanning and computer analysis. Radiology 89:211–215Google Scholar
  2. 2.
    Cowen AR, Parkin GJS, Hawkridge P (1997) Direct digital mammography image acquisition. Eur Radiol 7:918–930PubMedCrossRefGoogle Scholar
  3. 3.
    Kheddache S, Thilander-Klang A, Lanhede B, Mansson LG, Bjurstam N, Ackerholm P, Björneld L (1999) Storage phosphor and film-screen mammography: performance with different mammographic techniques. Eur Radiol 9:591–597PubMedCrossRefGoogle Scholar
  4. 4.
    Bick U (2000) Full-field digital mammography. Fortschr Röntgenstr 172:957–964CrossRefGoogle Scholar
  5. 5.
    Fischer U, Hermann KP, Baum F (2006) Digital mammography: current state and future aspects. Eur Radiol 16:38–44PubMedCrossRefGoogle Scholar
  6. 6.
    Pisano ED, Yaffe MJ (2005) Digital mammography. Radiology 234:353–362PubMedCrossRefGoogle Scholar
  7. 7.
    Center for Devices and Radiological Health of the U.S. Food and Drug Administration (2006) Mammography. Information for mammography facility personnel, inspectors, and consumers about the implementation of the Mammography Quality Standards Act of 1992 (MQSA). Last accessed on 25.9.2006. Available at: http://www.fda.gov/CDRH/mammography/
  8. 8.
    Maidment AD (2003) Digital mammography. Semin Roentgenol 38:216–230PubMedCrossRefGoogle Scholar
  9. 9.
    Mahesh M (2004) AAPM/RSNA physics tutorial for residents: digital mammography: an overview. Radiographics 24:1747–1760PubMedCrossRefGoogle Scholar
  10. 10.
    Huda W, Sajewicz AM, Ogden KM, Dance DR (2003) Experimental investigation of the dose and image quality characteristics of a digital mammography imaging system. Med Phys 30:442–448PubMedCrossRefGoogle Scholar
  11. 11.
    Berns EA, Hendrick RE, Cutter GR (2003) Optimization of technique factors for a silicon diode array full-field digital mammography system and comparison to screen-film mammography with matched average glandular dose. Med Phys 30:334–340PubMedCrossRefGoogle Scholar
  12. 12.
    Noel A, Thibault F (2004) Digital detectors for mammography: the technical challenges. Eur Radiol 14:1990–1998PubMedCrossRefGoogle Scholar
  13. 13.
    Bloomquist AK, Yaffe MJ, Pisano ED, Hendrick RE, Mawdsley GE, Bright S, Shen SZ, Mahesh M, Nickoloff EL, Fleischman RC, Williams MB, Maidment AD, Beideck DJ, Och J, Seibert JA (2006) Quality control for digital mammography in the ACRIN DMIST trial: part I. Med Phys 33:719–736PubMedCrossRefGoogle Scholar
  14. 14.
    Perry N, Broeders M, de Wolf C, Törnberg S, Holland R, von Karsa L, Puhaar E (2006) European guidelines for quality assurance in breast cancer screening and diagnosis, 4th edn. Luxembourg: Office for Official Publications of the European CommunitiesGoogle Scholar
  15. 15.
    Young KC, Cook JJH, Oduko JM (2006) Automated and human determination of threshold contrast for digital mammography systems. In: Astley SM, Brday M, Rose C, Zwiggelaar R (eds) Digital mammography. Proceedings of the IWDM 2006, LNCS 4046. Berlin Heidelberg: Springer, pp 266–272Google Scholar
  16. 16.
    Gennaro G, di Maggio C (2006) Dose comparison between screen/film and full-field digital mammography. Eur Radiol 16:2559–2566PubMedCrossRefGoogle Scholar
  17. 17.
    Gosch D, Jendrass S, Scholz M, Kahn T (2006) Radiation exposure in full-field digital mammography with a selenium flat-panel detector. Fortschr Röntgenstr 178:693–697CrossRefGoogle Scholar
  18. 18.
    Hermann KP, Obenauer S, Marten K, Kehbel S, Fischer U, Grabbe E (2002) Average glandular dose with amorphous silicon full-field digital mammography-clinical results. Fortschr Röntgenstr 174:696–699CrossRefGoogle Scholar
  19. 19.
    Marshall NW (2006) A comparison between objective and subjective image quality measurements for a full field digital mammography system. Phys Med Biol 51:2441–2463PubMedCrossRefGoogle Scholar
  20. 20.
    Bosmans H, Carton AK, Rogge F, Zanca F, Jacobs J, Van Ongeval C, Nijs K, Van Steen A, Marchal G (2005) Image quality measurements and metrics in full field digital mammography: an overview. Radiation Protection Dosimetry 117:120–130PubMedCrossRefGoogle Scholar
  21. 21.
    Bick U, Diekmann F, Grebe S, Marth F, Juran R, Friedrich M, Hamm B (2001) Contrast-detail resolution of dull-field digital mammography in comparison to conventional film-screen mammography. In: Yaffe M (ed) IWDM 2000. Proceedings of the 5th International Workshop on Digital Mammography. Madison: Medical Physics Publishing, pp 627–632Google Scholar
  22. 22.
    Grosjean B, Muller S (2006) Impact of textured background on scoring of simulated CDMAM phantom. In: Astley SM, Brady M, Rose C, Zwiggelaar R (eds) Digital mammography. Proceedings of the IWDM 2006, LNCS 4046. Berlin Heidelberg: Springer, pp 460–467Google Scholar
  23. 23.
    Burgess AE, Jacobson FL, Judy PF (2001) Human observer detection experiments with mammograms and power-law noise. Med Phys 28:419–437PubMedCrossRefGoogle Scholar
  24. 24.
    Kotre CJ (1998) The effect of background structure on the detection of low contrast objects in mammography. Br J Radiol 71:1162–1167PubMedGoogle Scholar
  25. 25.
    Diekmann F, Diekmann S, Bick U, Hamm B (2002) Reduced-dose digital mammography of skin calcifications. AJR 178:473–474PubMedGoogle Scholar
  26. 26.
    Riedl CC, Jaromi S, Floery D, Pfarl G, Fuchsjaeger MH, Helbich TH (2005) Potential of dose reduction after marker placement with full-field digital mammography. Invest Radiol 40:343–348PubMedCrossRefGoogle Scholar
  27. 27.
    Diekmann F, Diekmann S, Jeunehomme F, Muller S, Hamm B, Bick U (2005) Digital mammography using iodine-based contrast media: initial clinical experience with dynamic contrast medium enhancement. Invest Radiol 40:397–404PubMedCrossRefGoogle Scholar
  28. 28.
    Dobbins JT, 3rd, Godfrey DJ (2003) Digital X-ray tomosynthesis: current state of the art and clinical potential. Phys Med Biol 48:R65–R106PubMedCrossRefGoogle Scholar
  29. 29.
    Bloomquist AK, Yaffe MJ, Mawdsley GE, Hunter DM, Beideck DJ (2006) Lag and ghosting in a clinical flat-panel selenium digital mammography system. Med Phys 33:2998–3005PubMedCrossRefGoogle Scholar
  30. 30.
    Berns EA, Hendrick RE, Cutter GR (2002) Performance comparison of full-field digital mammography to screen-film mammography in clinical practice. Med Phys 29:830–834PubMedCrossRefGoogle Scholar
  31. 31.
    Couwenhoven M, William Sehnert W, Wang X, Dupin M, Wandtke J, Don S, Kraus R, Paul N, Halin N, Sarno N (2005) Observer study of a noise suppression algorithm for computed radiography images. Proc SPIE 5749:318–327CrossRefGoogle Scholar
  32. 32.
    Lewin JM, Hendrick RE, D'Orsi CJ, Isaacs PK, Moss LJ, Karellas A, Sisney GA, Kuni CC, Kutter GR (2001) Comparison of full-field digital mammography with screen-film mammography for cancer detection: results of 4,945 paired examinations. Radiology 218:873—880PubMedGoogle Scholar
  33. 33.
    Lewin JM, D’Orsi CJ, Hendrick RE, Moss LJ, Isaacs PK, Karellas A, Cutter GR (2002) Clinical comparison of full-field digital mammography and screen-film mammography for detection of breast cancer. AJR 179:671–677PubMedGoogle Scholar
  34. 34.
    Skaane P, Young K, Skjennald A (2003) Population-based mammography screening: comparison of screen-film and full-field digital mammography with soft-copy reading-Oslo I study. Radiology 229:877–884PubMedCrossRefGoogle Scholar
  35. 35.
    Skaane P, Skjennald A, Young K, Egge E, Jebsen I, Sager EM, Scheel B, Sovik E, Ertzaas AK, Hofvind S, Abdelnoor M (2005) Follow-up and final results of the Oslo I Study comparing screen-film mammography and full-field digital mammography with soft-copy reading. Acta Radiol 46:679–689PubMedCrossRefGoogle Scholar
  36. 36.
    Skaane P, Skjennald A (2004) Screen-film mammography versus full-field digital mammography with soft-copy reading: randomized trial in a population-based screening program-the Oslo II Study. Radiology 232:197–204PubMedCrossRefGoogle Scholar
  37. 37.
    Pisano ED, Gatsonis CA, Yaffe MJ, Hendrick RE, Tosteson AN, Fryback DG, Bassett LW, Baum JK, Conant EF, Jong RA, Rebner M, D’Orsi CJ (2005) American College of Radiology Imaging Network digital mammographic imaging screening trial: objectives and methodology. Radiology 236:404–412PubMedCrossRefGoogle Scholar
  38. 38.
    Pisano ED, Gatsonis C, Hendrick E, Yaffe M, Baum JK, Acharyya S, Conant EF, Fajardo LL, Bassett L, D’Orsi C, Jong R, Rebner M (2005) Diagnostic performance of digital versus film mammography for breast-cancer screening. N Engl J Med 353:1773–1783PubMedCrossRefGoogle Scholar
  39. 39.
    Venta LA, Hendrick RE, Adler YT, DeLeon P, Mengoni PM, Scharl AM, Comstock CE, Hansen L, Kay N, Coveler A, Cutter G (2001) Rates and causes of disagreement in interpretation of full-field digital mammography and film-screen mammography in a diagnostic setting. AJR 176:1241–1248PubMedGoogle Scholar
  40. 40.
    Carney PA, Miglioretti DL, Yankaskas BC, Kerlikowske K, Rosenberg R, Rutter CM, Geller BM, Abraham LA, Taplin SH, Dignan M, Cutter G, Ballard-Barbash R (2003) Individual and combined effects of age, breast density, and hormone replacement therapy use on the accuracy of screening mammography. Ann Intern Med 138:168–175PubMedGoogle Scholar
  41. 41.
    Burke W, Daly M, Garber J, Botkin J, Kahn MJ, Lynch P, McTiernan A, Offit K, Perlman J, Petersen G, Thomson E, Varricchio C (1997) Recommendations for follow-up care of individuals with an inherited predisposition to cancer. II. BRCA1 and BRCA2. Cancer Genetics Studies Consortium. JAMA 277:997–1003PubMedCrossRefGoogle Scholar
  42. 42.
    Lehman CD, Blume JD, Weatherall P, Thickman D, Hylton N, Warner E, Pisano E, Schnitt SJ, Gatsonis C, Schnall M, DeAngelis GA, Stomper P, Rosen EL, O’Loughlin M, Harms S, Bluemke DA (2005) Screening women at high risk for breast cancer with mammography and magnetic resonance imaging. Cancer 103:1898–1905PubMedCrossRefGoogle Scholar
  43. 43.
    Maes RM, Dronkers DJ, Hendriks JH, Thijssen MA, Nab HW (1997) Do non-specific minimal signs in a biennial mammographic breast cancer screening programme need further diagnostic assessment? Br J Radiol 70:34–38PubMedGoogle Scholar
  44. 44.
    Ruschin M, Hemdal B, Andersson I, Borjesson S, Hakansson M, Bath M, Grahn A, Tingberg A (2005) Threshold pixel size for shape determination of microcalcifications in digital mammography: a pilot study. Radiation Protection Dosimetry 114:415–423PubMedCrossRefGoogle Scholar
  45. 45.
    Ideguchi T, Higashida Y, Kawaji Y, Sasaki M, Zaizen M, Shibayama R, Nakamura Y, Koyanagi K, Ikeda H, Ohki M, Toyofuku F, Muranaka T (2004) New CR system with pixel size of 50 microm for digital mammography: physical imaging properties and detection of subtle microcalcifications. Radiat Med 22:218–224PubMedGoogle Scholar
  46. 46.
    Chan HP, Helvie MA, Petrick N, Sahiner B, Adler DD, Paramagul C, Roubidoux MA, Blane CE, Joynt LK, Wilson TE, Hadjiiski LM, Goodsitt MM (2001) Digital mammography: observer performance study of the effects of pixel size on the characterization of malignant and benign microcalcifications. Acad Radiol 8:454–466PubMedCrossRefGoogle Scholar
  47. 47.
    Kim HH, Pisano ED, Cole EB, Jiroutek MR, Muller KE, Zheng Y, Kuzmiak CM, Koomen MA (2006) Comparison of calcification specificity in digital mammography using soft-copy display versus screen-film mammography. AJR 187:47–50PubMedCrossRefGoogle Scholar
  48. 48.
    Fischer U, Baum F, Obenauer S, Luftner-Nagel S, von Heyden D, Vosshenrich R, Grabbe E (2002) Comparative study in patients with microcalcifications: full-field digital mammography vs screen-film mammography. Eur Radiol 12:2679–2683PubMedGoogle Scholar
  49. 49.
    Diekmann S, Bick U, von Heyden H, Diekmann F (2003) Visualization of microcalcifications on mammographies obtained by digital full-field mammography in comparison to conventional film-screen mammography. Fortschr Röntgenstr 175:775–779CrossRefGoogle Scholar
  50. 50.
    Karssemeijer N, Frieling JTM, Hendriks JHCL (1993) Spatial resolution in digital mammography. Invest Radiol 28:413–419PubMedCrossRefGoogle Scholar
  51. 51.
    Crystal P, Strano S (2006) Digital and film mammography. N Engl J Med 354:765–767PubMedCrossRefGoogle Scholar
  52. 52.
    Dershaw DD (2005) Film or digital mammographic screening? N Engl J Med 353:1846–1847PubMedCrossRefGoogle Scholar
  53. 53.
    Fischer U, Baum F, Obenauer S, Funke M, Hermann KP, Grabbe E (2002) Digital full field mammography: comparison between radiographic direct magnification and digital monitor zooming Radiologe 42:261–264PubMedCrossRefGoogle Scholar
  54. 54.
    Berns EA, Hendrick RE, Solari M, Barke L, Reddy D, Wolfman J, Segal L, DeLeon P, Benjamin S, Willis L (2006) Digital and screen-film mammography: comparison of image acquisition and interpretation times. AJR 187:38–41PubMedCrossRefGoogle Scholar
  55. 55.
    Ciatto S, Brancato B, Baglioni R, Turci M (2006) A methodology to evaluate differential costs of full field digital as compared to conventional screen film mammography in a clinical setting. Eur J Radiol 57:69–75PubMedCrossRefGoogle Scholar
  56. 56.
    Pisano ED, Cole EB, Kistner EO, Muller KE, Hemminger BM, Brown ML, Johnston RE, Kuzmiak CM, Braeuning MP, Freimanis RI, Soo MS, Baker JA, Walsh R (2002) Interpretation of digital mammograms: comparison of speed and accuracy of soft-copy versus printed-film display. Radiology 223:483–488PubMedCrossRefGoogle Scholar
  57. 57.
    Roelofs AA, van Woudenberg S, Otten JD, Hendriks JH, Bodicker A, Evertsz CJ, Karssemeijer N (2006) Effect of soft-copy display supported by CAD on mammography screening performance. Eur Radiol 16:45–52PubMedCrossRefGoogle Scholar
  58. 58.
    Malich A, Fischer DR, Bottcher J (2006) CAD for mammography: the technique, results, current role and further developments. Eur Radiol 16:1449–1460PubMedCrossRefGoogle Scholar
  59. 59.
    Bick U, Giger ML, Schmidt RA, Nishikawa RM, Doi K (1996) Density correction of peripheral breast tissue on digital mammograms. Radiographics 16:1403–1411PubMedGoogle Scholar
  60. 60.
    Byng JW, Critten JP, Yaffe MJ (1997) Thickness-equalization processing for mammographic images. Radiology 203:564–568PubMedGoogle Scholar
  61. 61.
    Pisano ED, Cole EB, Major S, Zong S, Hemminger BM, Muller KE, Johnston RE, Walsh R, Conant E, Fajardo LL, Feig SA, Nishikawa RM, Yaffe MJ, Williams MB, Aylward SR (2000) Radiologists’ preferences for digital mammographic display. Radiology 216:820–830PubMedGoogle Scholar
  62. 62.
    Baum F, Fischer U, Obenauer S, Grabbe E (2002) Computer-aided detection in direct digital full-field mammography: initial results. Eur Radiol 12:3015–3017PubMedGoogle Scholar
  63. 63.
    Astley SM (2004) Computer-aided detection for screening mammography. Acad Radiol 11:1139–1143PubMedCrossRefGoogle Scholar
  64. 64.
    Freer TW, Ulissey MJ (2001) Screening mammography with computer-aided detection: prospective study of 12,860 patients in a community breast center. Radiology 220:781–786PubMedCrossRefGoogle Scholar
  65. 65.
    Birdwell RL, Bandodkar P, Ikeda DM (2005) Computer-aided detection with screening mammography in a university hospital setting. Radiology 236:451–457PubMedCrossRefGoogle Scholar
  66. 66.
    Cupples TE, Cunningham JE, Reynolds JC (2005) Impact of computer-aided detection in a regional screening mammography program. AJR 185:944–950PubMedCrossRefGoogle Scholar
  67. 67.
    Roehrig J (2005) The manufacturer’s perspective. Br J Radiol 78:S41–S45PubMedCrossRefGoogle Scholar
  68. 68.
    Warren Burhenne LJ, Wood SA, D’Orsi CJ, Feig SA, Kopans DB, O’Shaughnessy KF, Sickles EA, Tabar L, Vyborny CJ, Castellino RA (2000) Potential contribution of computer-aided detection to the sensitivity of screening mammography. Radiology 215:554–562PubMedGoogle Scholar

Copyright information

© Springer-Verlag 2007

Authors and Affiliations

  1. 1.Department of RadiologyCharité-Universitätsmedizin BerlinBerlinGermany

Personalised recommendations