Some imaging providers may question whether they can afford digital mammography. Others will justify the expense associated with the transition to digital mammography using the productivity that they expect to gain. Initially, however, implementing digital mammography will be a time-consuming process because of the major change in workflow that it represents for radiologists and technologists.

Unfortunately, its developers initially saw digital mammography as another imaging modality, not as a system. For this reason, successful integration of digital mammography into a radiology practice calls for more than an acquisition of capital equipment; it is an imaging informatics project.

It is important for a breast center not to settle for a digital mammography system that seems good enough for today; good enough never is. The best that can be done today cannot be allowed to set the practice standard for tomorrow. Achieving a breast-care delivery system that sustains quality while improving efficiency, productivity, and revenues is a journey, not a destination. The pursuit of excellence must be relentless and continuous if the breast center is to achieve optimal results from its implementation of digital mammography.

Workflow Challenges

Table 1. Technical
Attributes for Display

Exposure time
Filter material
Anode target material
Compression force
Body-part thickness
Primary angle of positioner
Relative x-ray exposure
Entrance dose
Organ dose
Date of last calibration

Systems integration is as important as modality integration and detector specifications. Likewise, workflow design and management are as critical as choosing the digital mammography system. For service line planning, four aspects of workflow should be considered: images and other digital data, patients, technologists, and radiologists.

Some challenges in digital infrastructure remain. For general workflow, robust worklists for physician review workstations are needed, along with communication between review workstations that allows physicians to manage the daily workload among themselves and in cooperation with each other. Continuing improvements in automated prefetching capabilities and hanging protocols, especially where more than one type of digital mammography system is in use (for example, needing to compare mosaic with nonmosaic studies), are needed to put mammography at productivity levels comparable to those of other digital imaging technologies. Operating in both the analog and digital worlds continues to present challenges to workflow and productivity. When staff continue to hang prior film-screen mammograms, with or without laser-printed current digital mammograms, in addition to querying and retrieving current and prior digital mammograms on the workstation, productivity potential is reduced.

Performance specifications should include the detector size (coverage and dose) and the resolution, with sampling pitch matched to detector resolution.
—Laurie L. Fajardo, MD, MBA

Challenges for technologist workflow include the management of worklists and the ability to view workflow from within the digital mammography room. There is a need to prefetch and retrieve from the PACS to the modality or acquisition workstation, as well as a need to preview prior studies before acquiring new images. It is also necessary to preview the previous year’s annotations, annotations on a recent screen (for callbacks), and annotations made in real time during the diagnostic workup for a symptomatic patient.

Radiologist workflow is subject to challenges that affect the efficiency of interpretation. The annotation of abnormalities is critical for year-to-year image comparisons. Electronic tools for this purpose are not always intuitive, and automatic storage of annotations is sometimes inefficient because it requires too many clicks. The process can even be less efficient than the old wax-pencil annotation method, since it can require clicking on an annotation or region-of-interest tool, making the annotation, saving it, and transferring a secondary capture with the annotation to PACS.

For maximal interpretation efficiency, the radiologist needs a diagnostic display workstation, similar to a general PACS review workstation that supports other modalities, including ultrasound, MRI, PET, and nuclear medicine. Systems should support automated processing software for other modalities, such as MRI, including third-party software. Communication between radiologist workstations should be possible, including the ability to create separate screening and diagnostic sessions, as well as to view and monitor workflow. The ability to store unique data sets, such as case-review, teaching, and audit sets, is needed.

Because digital mammography is an imaging informatics (not simply imaging) project, vendors of digital mammography systems should view systems integration as a product offering that is fully as important as technical and modality products. It is necessary to determine the center’s digital infrastructure before selecting and purchasing technologies. Efficient acquisition of digital mammography capabilities should be based on the workflow integration profile of the center. This profile covers registering patients, placing orders, filling orders, and managing and archiving images.

The actor functionality within the Integrating the Healthcare Enterprise (IHE) project is achieved through the information-technology functions performed by the RIS, PACS, HIS, and admission-discharge-transfer system.

There are vendor variances in image data and attributes (where processing is applied). It is important to have access to prior studies for comparison. In addition, computer-aided detection (CAD) is becoming the standard of care, and all regulatory expectations must be met.

Informatics, Performance Specifications

The Digital Imaging and Communications in Medicine (DICOM) standard has both required and optional components, so it is important to determine what a vendor means when claiming to support DICOM. In general, such standards as DICOM and Health Level 7 allow providers to use information more effectively, helping them to optimize care. The IHE Mammography Image Profile specifies:

  • how DICOM mammography images and evidence objects are created, exchanged, and used;
  • how acquisition modalities transfer digital mammography images;
  • how CAD systems act as evidence creators; and
  • how image displays should retrieve and make use of CAD results.

Informatics specifications for digital mammography should cover DICOM storage and modality worklists; the DICOM modality performed procedure step, which permits postprocessing with CAD and storage of DICOM mammography CAD objects; and communication with the image manager and the image archive system. In addition, it should be possible to perform proprietary image postprocessing on the acquisition modality and to store both images for processing and images for presentation. This allows postprocessing or reprocessing of prior studies when new algorithms become available, which would be impossible if only the for-presentation images were stored.

Performance specifications should include the detector size (coverage and dose) and the resolution, with sampling pitch matched to detector resolution. An uptime percentage should be specified, as should detector life; since technology typically changes before detectors break, a 3-year warranty for detectors is reasonable. Throughput should be specified as well—in Europe, it is usually 12 to 15 patients per hour. Specifications should cover the dose level at which the system is capable of producing images of diagnostic quality and support for upright stereotactic biopsy.

Workstations and Image Display

Because there is no universal breast-center model, specification of digital mammography systems must be approached individually and with great care, if the center’s needs are to be met.
—JeongMi Park, MD

Workstations should support IHE reporting of workflow integration profiles for postprocessing, CAD, and worklists across several workstations. Automated retrieval of prior images should be based on worklist selection, and display of for-presentation images should be available from any modality vendor. Displayed images from different vendors, with different detector sizes, should still be of accurate size. Mandated annotations should be displayed correctly. Correct and automatic orientation should apply to all views from all vendors, as should standard hanging protocols. Header information should include DICOM value of interest lookup tables. Sigmoidal window/level manipulations are needed so that the parenchymal aspects of images will not change when these manipulations are applied to breast tissue.

Image display should support calibration as described in the DICOM Grayscale Softcopy Display Function and horizontal image justification when the aspect ratio of the viewport does not match the aspect ratio of the image. True-size display should be available, so that a measurement taken of an object within the image, taken at the surface of the display, is approximately the actual physical size of the object. Even if images have different pixel sizes because they were obtained using detectors with different pixel matrices, it should be possible to display them at the same size. If it is not possible to display all encoded pixels simultaneously at maximal resolution, it should be possible to view actual pixels using a pan/quadrant function.

In addition, image display should support the display of certain technical factors (Table 1) that are encoded in the image integration profile; these are required to detect and resolve quality problems. It should be possible to turn various display functions on and off at the user’s discretion, and image display should support the variety of hanging protocols specified by IHE.

Print Composers, Print Servers, and Storage

Print composers should support true-size printing, since this ensures accurate measurements on printed film. Images should be justified so that the chest wall is printed as close to the edge of the film as the print server is capable of printing it. All annotations defined for image displays, along with a ruler or distance scale, should be burned into the pixel data. A pixel transmission of 10 bits or more to the print server should be specified.

Print servers should support true-size printing and should create a border of 2 mm or less between the chest wall and the edge of the film. Printing should use a maximum optical density of 3.5 or more, with receipt of pixel data at a bit depth of 10 bits or more from the print composer.

Image-storage requirements vary greatly according to the field of view, type of compression, and system. For example, a center performing 10,000 examinations per year (30% of those with a large field of view) on a 50-µm-per-pixel detector system and using 4:1 lossless compression would need 2.8 terabytes of storage for its 7-year archive. Using a 70-µm-per-pixel detector system without changing other characteristics would reduce the storage need to 1.3 terabytes, and a 100-µm-per-pixel detector system would require 0.7 terabytes ( Table 2 and Tables 3–5). Naturally, storage costs vary with archive size (Table 6 below).

Because there is no universal breast-center model, specification of digital mammography systems must be approached individually and with great care, if the center’s needs are to be met. Benchmarks may not exist because no two breast centers operate in identical environments. It will be necessary to innovate, and to make innovation part of the center’s culture; this may be challenging initially, but it is what makes the best breast centers great.

Laurie L. Fajardo, MD, MBA, is professor and chair, and JeongMi Park, MD, is professor in the Department of Radiology at the University of Iowa, Iowa City. This article has been adapted from “Digital Mammography: How to Make It Work in Your Practice,” presented at the 26th Annual Breast Imaging Conference, held on September 28 in Las Vegas. For more information, contact .