[Standards] considerations using digital mammography

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon

So you've purchased a digital mammography system, and now what? Deploying such a system not only opens up a new world of opportunities, but also a lot of potential challenges.

Part of these challenges have to do with interfacing the equipment properly, which does not merely mean connecting "the plug" but also, and even more importantly, fitting the system into the workflow of your department. Of course, you can take the easy way out, and connect it straight into a printer and continue to generate films. However, I would argue that this is the same as buying a digital photo camera, and still dropping it off to your local drugstore for the development of your all-to-familiar paper snapshots. To really use the power of digital mammography, you should deploy a digital viewing system, archiving the images electronically, and use some of the computer aided detection (CAD) and automated reporting tools that can be nicely integrated with these systems.


Let's first see what the DICOM standard has specified for the interface of these digital modalities. The DICOM standard actually specifies two different interfaces. First of all, it allows sending images out in an unprocessed manner, also known as a "For Processing" image, which can be considered raw data. The image is not optimized for viewing, which is obvious from its appearance if it is displayed. The reason for this particular format is so that the image can be sent to a modality workstation for subsequent processing, and also to a CAD system, which prefers to operate on the raw data. One has to realize though that this option is only useful when the image data are sent to a workstation of the same vendor that knows how to process the information. Here is where the workflow impact becomes obvious: If the "For Processing" image is the only option (as is the case with a few systems on the market), it always has to be routed to this particular processing station of the same manufacturer. If not, one can send the images to a viewing station from a different manufacturer.

Assume we have viewable digital images, i.e. the "For Presentation" format, which allows an image to be viewed as-is, i.e. without additional image processing, which allows it to be displayed on a different viewing station than the one that might be provided by the digital mammography manufacturer. There are several concerns, first of all, from a hardware perspective. One needs to make sure that the resolution, both spatial and contrast, is sufficient, as well as the brightness. Needless to say that a 1 MPixel monitor simply would not meet the requirements for displaying these images which can be up to 40 MByte. Second, one needs to make sure that the images are "hanged" appropriately. What this means is that the proper views are positioned for the radiologist to look at. As one radiologist explained to me, the best reference for a radiologist of the particulars of a woman's breast is the other one, so that peculiarities, if any, can be compared. That is why they always insist on the "mirror view", i.e. displaying these images side-by-side, as if mirrored. The good news is that the instructions for the software to "hang" these, or "hanging protocols" are relatively simple to implement, and identical for most radiologists.

The DICOM image header is interpreted to find the appropriate information for the display, which consists of the viewing code, laterality, patient orientation, and body part (breast). You can imagine what happens if a technologist by accident enters any one of these parameters incorrectly during the acquisition; the information is incorrectly filled into the header and the images are not displayed properly, maybe upside down, flipped, or at the incorrect position on the screen. Remember, a lead marker on an image to indicate positioning is simply not enough, user input is required.


The third concern with the proper display is the image quality and consistency. It is critical that the display support the DICOM Greyscale Standard Display Function (GSDF). This means that they can be calibrated according to the proper transformation curve as defined by the DICOM standard, which defines exactly how to map the digital driving levels into luminance values, i.e. appearance on the screen. In addition, I have seen at least one vendor recently that went beyond the regular GSDF corrections and also implemented an extra uniformity correction for its flat-panel displays, especially for mammography. Calibrating the monitor is not enough; this only guarantees that the performance is correct as of the moment of calibration. What about the calibration to be done one, three, six months from now? It makes sense to have a monitor that is automatically calibrated and/or of which the display performance is managed remotely.

Unfortunately, there will still be a mixed environment for a while, whereby images are generated digitally today but the previous study is still available on film, either from the file room of the institution, or brought in by a patient from another clinic. In addition, there could be paper records, such as previous reports, that need to be made available for review on a workstation. Therefore, without a film digitizer, a digital system does not make sense, unless you want a physician to look both at the monitor and lightbox at the same time, which is a real challenge and I can imagine might compromise the way a physician can do proper comparison. Note again that proper orientation information in the DICOM header is a requirement for the images to come up properly, which also applies to digitized films. This requires both implementation support at the device and properly trained staff to enter this information correctly. If this information is entered by the file room staff, they might need additional training to recognize the particular views and orientation of the films.

Now that we have a set of images with a consistent greyscale image on a monitor, what's next? The logical step is reporting, with or without the use of CAD. The CAD software typically generates a series of detections (if present), based on a certain analysis, using a set of images. Not only should the detections be archived, but the analysis as well, because a follow-up exam could generate a detection correlating the previous and new analysis. The appropriate mechanism to document and archive the results of the CAD process is a DICOM Mammography Structured Report (SR). The reason for using DICOM is that is nicely fits within the existing system architecture, i.e. it can be archived on the main PACS archive. One might not necessarily expect that general purpose viewing stations would be able to interpret this type of information, however, it will have to be interpreted by specialty viewers that are able to interpret the information stored in the DICOM SR and, for example, generate the appropriate overlays on the images to indicate the detected pathology.

Lastly, how does the reporting fit in with all of this? There are specialized mammography reporting systems, which can be very effectively used because of the high degree of standardization and guidelines for these reports. These systems also allow for standardization of the used templates. A special DICOM SR in the form of a breast imaging report for mammography has been standardized recently, something you might want to check into with your vendor for support.


As with any relatively new standard, it always takes some time for the industry to support it. There is a time period of proprietary and semi-proprietary solutions, and often a chicken-and-egg situation: mammography vendors wait for the workstation vendors, that wait for the reporting vendors that wait for the CAD vendors to all support these new standards. Only by user pressure and general awareness, the movement to a "plug-and-play" environment whereby a user can select best-of-breed components based on an open standard can be accelerated.



Herman Oosterwijk is president of OTech Inc. (www.otechimg.com), a healthcare training and consulting firm, specializing in training and e-learning about these digital technologies. He just published a new book called PACS Fundamentals, which is available for sale on the OTech web site.