The development of digital mammography began in the mid 1980s, but accelerated tremendously in 1993, when the National Cancer Institute (NCI) funded the National (later International) Digital Mammography Development Group. The underlying rationale for digital mammography is that by overcoming the technical limitations of screen-film mammography, improved performance in detection and radiological diagnosis of breast cancer can be achieved. These limitations are the restricted image contrast and latitude of film, film granularity, and inefficiencies in the use of radiation in forming the image. At the time of writing, four systems have received FDA approval for clinical use; one of these systems is no longer produced.

Currently, the accuracy of digital mammography (DM) is being evaluated versus screen-film mammography (SFM) in the application of screening in the Digital Mammography Imaging Screening Trial (DMIST), funded by NCI and administered by the ACR Imaging Network (ACRIN). In this trial 49,500 women at 35 facilities received both DM and SFM on the same day, and the images were read independently by different radiologists. Although the results of that study will not be available until early 2005, many facilities worldwide are now using DM both for workup and for screening.


The two critical components defining the overall performance of DM are the x-ray detector and the display workstation. In order to provide comparable or improved performance over SFM, the detector must meet several requirements including:

  1. Efficiently absorb incident x-rays; ie, the detector must have high quantum efficiency
  2. Produce a signal large enough to be reliably measured
  3. Provide adequate spatial resolution to depict fine details
  4. Allow subtle changes in tissue composition to be detected. In technical terms, a high signal difference to noise ratio (SDNR) is required.
  5. Have adequate dynamic range (latitude) to allow imaging all regions of the breast.
  6. Include as much of the breast tissue as possible in the image.

Detectors are frequently described by a quantity called detective quantum efficiency (DQE), which varies between 0% and 100% and combines the first four criteria listed above. It characterizes the efficiency of using the radiation to convey information (in the form of SDNR) to the image. The DQE for a screen-film system is on the order of 45%. Digital systems promise to increase DQE considerably to 75% or higher.

Since their introduction in the first digital mammography systems, detectors have been refined to improve their performance and reliability. The first systems introduced used cesium iodide (CsI) phosphors to absorb the x-rays and produce light, and either photodiodes or charge coupled devices (CCDs) to convert the light to an electronic signal that could be digitized. In addition, the photostimulable phosphor plate used in general radiography was adapted to provide higher spatial resolution for mammography. Currently, there are four major system types that are being used clinically for digital mammography and these are described briefly here. The image field coverage and detector element (del) size vary among the systems.

Type A. The detector consists of a CsI phosphor, coupled directly to a large matrix array of photodiodes fabricated on a large-area plate using amorphous silicon technology. Readout is achieved by activating control lines running down each column of the matrix. These are connected to thin film transistor (TFT) switches located in the corner of each detector element adjacent to the photodiodes. The light from the phosphor causes charge to be deposited in the photodiodes. Activation of the switches transfers the charge to readout lines running along each row of the matrix and then to a digitizer. The del size is 100 µm. The detector is robust and provides fast readout, facilitating procedures requiring serial exposures. Future developments include reduction of noise of the detector electronics, which will increase DQE.

Contrast digital mammography. Mask cranial caudal digital mammogram showing a density and some microcalcifications.

Type B. In this system, a CsI phosphor is coupled through fiberoptic plates to CCDs, which convert the light to an electronic signal and provide a readout. The detector is slot-shaped, and scans across the breast in synchrony with a fan beam of x-rays to acquire the image data under low scatter conditions. The standard del size is 50 µm. In a high-resolution mode, a pixel size of 25 µm is provided over a limited field of view. An acquisition system with 14-bit precision, reduced noise, and an automatic exposure control system is expected to be introduced in the near future.

Contrast digital mammography. Subtraction image where the uptake associated with angiogenesis is clearly seen. This was an invasive ductal carcinoma.

Type C. The system uses photostimulable or storage phosphor plates composed of BaFCl or BaFBr. The nominal del size is 50 µm. The initial signal consists of excited electrons stored in “traps” within the crystal molecular structure. The number of filled traps is proportional to the x-ray exposure received by the plate. The plate in its light-tight cassette is removed to a separate reader unit, where it is scanned by a red laser. The energy from the absorbed light releases electrons from the traps. This causes blue light to be generated by the phosphor. The emitted light is collected and measured with a photomultiplier tube and the resulting electronic signal is digitized. Collection of the light from both the top and bottom of the plate in the reader improves sensitivity of the detector and reduces noise. Analog image signals are compressed logarithmically before image digitization. Because this system uses an existing conventional mammography unit and a reader, which can be common to several systems, conversion to digital is less expensive than with the other systems. It is necessary, however, to physically transfer the plate to the reader, so there is a greater delay between exposure and image availability than with the other systems.

Type D. This system is referred to as a direct converter as a phosphor is not used; the energy of x-ray absorption is converted directly into electronic charge liberated in the amorphous selenium absorber. The system is based on a flat plate detector with an active-matrix readout as in Type A, and a 70 µm detector element. An electric field imposed across the selenium layer collects the charge before it has had an opportunity to spread laterally and blur the image. One of the potential advantages of this approach is that the detector can be made thick for high quantum efficiency while maintaining high spatial resolution.

There are also two new digital mammography systems, developed in Sweden, that are of interest. Both of these utilize slot scanning for image acquisition. Unlike the systems already described, which accumulate the light or electronic charge produced by many x-rays to form a signal, the detectors in these actually count the number of individual x-rays interacting with the detector and store a matrix of these numbers as the image signal. This may offer some advantages in signal-to-noise ratio in the images.


On all of the systems except Type C, the image that is digitized is first corrected for spatial nonuniformities in the sensitivity of the detector. This is referred to as a shading or flat field correction. In the Type C system, correction is made for nonuniformities of the laser reader only. The image at this point is referred to as the “raw image.”

Further processing is frequently performed to improve contrast and sharpness of structures in the breast. This generally includes peripheral equalization, which corrects for the large change in signal due to the rapid change in thickness at the edge of the breast. It preserves the fluctuations in image signal due to variations in internal structures while suppressing the change due to thickness. This results in a reduced range of signal and allows the image of all regions of the breast to be more easily displayed in one setting of window and level. Other image processing may include edge enhancement or multi-frequency processing to sharpen the image.


The user friendliness of workstations varies among the manufacturers with considerable potential for improvements. Additionally, in general, workstations were not designed to display images from other manufacturers’ equipment and often employ proprietary image processing software. This limitation is now being addressed by third-party manufacturers, who are developing multimodality workstations that will allow viewing of digital images for mammography, ultrasound, and magnetic resonance imaging from different vendors. Monitors continue to improve with higher luminosity and pixel number.


Martin J. Yaffe, PhD

Screen-film mammography has been shown in randomized controlled trials to contribute to a decrease in breast cancer mortality. Published data have so far not demonstrated a significant difference in cancer detection between screen-film and digital mammography. The results of the DMIST will influence future decisions on using the digital technology for screening. Digital mammography can potentially improve productivity for technologists by eliminating the need to handle cassettes. This time-saving ability is particularly significant for radiologists during procedures such as a preoperative wire localization. It also means the woman’s breast is in compression for less time during this procedure. Similarly, time is saved during galactography. However, this does not apply to the CR system.

As images viewed on monitors can be adjusted to provide the necessary brightness and contrast, the need to repeat images for inadequate exposures is reduced. The need to apply interactive image display operations such as “window,” “level,” and “zoom” to view the image at its full acquired resolution can increase the time required to report digital examinations. Comparison to previous examinations can also be more difficult.

Roberta A. Jong, MD

Images from digital mammography can be stored on PACS and made available quickly for radiologists and clinicians throughout the department. The large size of the images makes a significant impact on storage requirements. Images can be sent to outside facilities on CD or DVD, although these images are generally of lesser quality when viewed on monitors (eg, personal computer screens) that have lower resolution than that of the digital mammography workstation.


Possibly the greatest impact from digital mammography will come from the array of applications that it enables or facilitates. These include: computer-aided detection and diagnosis, contrast-enhanced imaging, tomosynthesis, telemammography, and dual energy imaging.

Computer-assisted detection. Double reading of mammograms by two radiologists has been shown to increase cancer detection by about 10%. Computer programs can act as a second reader and are now used to flag possible lesions as areas of concern. In the past, this required films to be digitized, adding time and cost as well as noise associated with digitization. Digital mammography is better suited for this application. The computer searches for suspicious features such as mass density, irregularity of margins, spiculations, and microcalcifications in the digital mammogram. These features can be calculated in mathematical terms from the digital mammogram. By combining these features with optimal weighting using a neural network algorithm, the sensitivity and specificity can be maximized by accepting suspicious findings and rejecting those that are likely to be false-positive findings. Marks are then appended to the displayed mammogram to indicate areas suspicious for the presence of masses, architectural distortion, and microcalcifications. Based on these suggestions, the radiologist then considers whether these marks should change the initial interpretation of the mammogram. This could help avoid perception errors of the radiologist. Studies have shown that CAD can detect findings on prior mammograms that can be attributed to the cancer diagnosed at a later date. This could mean that cancers could potentially be diagnosed earlier. In the future, with adequate training, CAD may be able to provide an estimate of the probability of malignancy for a particular finding.

Improving conspicuity. Lesions are often missed because they are not adequately conspicuous, ie, they have inadequate contrast or they are masked by the complexity of overlying and underlying dense tissue structures. Similarly, superposition of such structures in the image can cause false-positive results to increase. Several new applications of digital mammography may help improve conspicuity of lesions.

Tomosynthesis. One way to reduce the effect of superposed tissue structures is to image different planes within the breast separately. This can be done by a method called tomosynthesis. In tomosynthesis, a number of individual exposures, made at different angles about the breast over a range of approximately ±11 degrees, are recorded and read out by the digital detector. By filtering, shifting, and adding the data appropriately in the computer, images of individual sections of the breast can be reconstructed. When viewed as a series (eg, in a movie loop), overlapping structures are separated and the presence of disease or confirmation of lack of abnormalities is determined.

Contrast-enhanced digital mammography. Another way of improving conspicuity is to image the effect of functional changes associated with cancer angiogenesis. Cancers produce growth factors that stimulate the formation of the blood vessels necessary for growth. Because the vessels formed in malignant angiogenesis are of poor quality, they leak and the pooling of intravascular contrast media can be imaged with digital mammography using a subtraction technique similar to that used in digital subtraction angiography. The mask and a subtracted image for such a procedure are shown in the figure on page 66. A similar result can be achieved using dual energy images rather than temporal subtraction. Imaging is performed at kVps above and below the K edge of iodine. In a manner very similar to that of breast MRI, the contrast medium would highlight the presence of tumor angiogenesis to make the cancers perceptible even against a background of dense fibroglandular tissue.

Telemammography. Telemedicine is now commonplace. The ability to send digital mammographic images to another site for reporting improves access to breast care for women in remote areas where no radiologist is available. In a mobile application, a digital system on a bus can be taken to the remote area for screening and even diagnostic views might be performed before the bus leaves the area. Reporting could be done more quickly and the time to diagnosis would be improved. This would lessen the period of anxiety of having no answer for women.

Currently, the capital cost of a digital mammography system is about four times that of a screen-film unit. To be justified clinically, digital mammography must offer either improved detection or diagnostic accuracy or increased efficiency or both. At present, digital mammography provides increased throughput for the technologist, but requires longer interpretation time. Improvements in workstations will ameliorate this situation. In addition, the results of DMIST and of some of the evaluations of applications such as those discussed here will help determine whether digital mammography will gain widespread acceptance.


  • Jong RA, Yaffe MJ, Skarpathiotakis M, et al. Contrast digital mammography: initial clinical experience. Radiology. 2003; 228:842-50.
  • Lewin JM, D’Orsi CJ, Hendrick RE, et al. Clinical comparison of full-field digital mammography and screen-film mammography for detection of breast cancer. AJR Am J Roentgenol. 2002;179:671-677.
  • Lewin JM, Isaacs PK, Vance V, Larke FJ. Dual-energy contrast-enhanced digital subtraction mammography: feasibility. Radiology. 2003;229:261-268.
  • Niklason LT, Christian BT, Niklason LE, et al. Digital tomosynthesis in breast imaging. Radiology. 1997;205:399-406.
  • Skarpathiotakis M, Yaffe MJ, Bloomquist AK, Rico D, Muller S, Rick A. Development of contrast digital mammography. Med Phys. 2002;29:2419-2426.
  • Vyborny CJ, Giger ML, Nishikawa RM. Computer-aided detection and diagnosis of breast cancer. Radiol Clin North Am. 2000;38:725-40.
  • Wu T, Stewart A, Stanton M, et al. Tomographic mammography using a limited number of low-dose cone-beam projection images. Med Phys. 2003;30:365-80.

Martin J. Yaffe, Phd, is senior scientist, Imaging/Bioengineering Research, Sunnybrook & Women’s College Health Sciences Centre, and professor, Departments of Medical Imaging and Medical Biophysics, University of Toronto.

Roberta A. Jong, MD, is director of breast imaging at Sunnybrook & Women’s College Health Sciences Centre, University of Toronto.