David S. Channin, MD

By now, you have woken up and cleared your hangover. You vaguely remember that at the radiology department holiday party you made a loud, public, New Year’s resolution to digitize your mammography operation, introduce CAD, and improve efficiency. Someone also thought they heard you commit to reducing liability, while improving regulatory compliance and patient satisfaction, and making the mammographers happy. The boss wanted you to commit to making a profit (excuse me, I mean revenue in excess of expenses), but you were sober enough to realize you were not going to solve world hunger either. So now, it is Monday morning, you have got your capital line items in a row, you have dusted off the brochures and request for proposal responses, you have got that shiny reprint of the DMIST trial results, and you are ready to go. Is it that easy? Of course not!

Converting a mammography operation to filmless (and paperless!) operations is as complex as converting any other aspect of a radiology department. This is not an imaging project; this is an imaging informatics project. To be successful, you will have to solve complex health care information management problems that involve multiple information systems from multiple vendors. Wait! Sound familiar? Yes, that is exactly the mission of the Integrating the Healthcare Enterprise (IHE) initiative (www.ihe.net). IHE is a process to drive the use of standards in solving these very problems. IHE has already solved 15 problems (called “integration profiles”) in radiology and is working on more both within radiology as well as in other domains and across and between health care institutions.

The first thing to do is to make sure you are purchasing the IHE functionality that lets you take advantage of the solutions that are available today. To do that, you need to step back and look at the larger picture in your department. Identify the core functionality that each system must have and how that actor (as IHE calls them) needs to play in the sandbox with the other children.

You want to have efficient modality acquisition. IHE calls this the Scheduled Workflow integration profile. There is more to scheduled workflow than just powering up the x-ray generator. You need to have actors for registering patients, placing orders, filling orders, and managing and archiving the images. Note that I did not say RIS, PACS, HIS, or ADT. IHE speaks in terms of actor functionality. You need to look at the systems in your environment and decide how you are going to fill each of these roles. You might be able to buy one box from one vendor that does all of these things (I doubt it), or several boxes from one or more vendors (more likely) or you can have people do it with paper and film; but one way or another, you have to perform these functions.

Let us look, for a moment, at the full-field digital mammography (FFDM) modality, often a subject of angst. Your chief of mammography has selected the unit that gives the prettiest pictures. Your physicist has blessed the MTF, DQE, QA, QC, MQSA, U/L, and all of the rest. Your imaging informatician has blessed the informatics of the system. Has she?

It is not enough for the vendor to say that the modality supports the Digital Imaging and Communications in Medicine (DICOM) standard. DICOM is necessary but not sufficient. IHE scheduled workflow says the modality has to do DICOM storage. It also says (and you want) the modality to do DICOM modality worklist, DICOM storage commitment, and DICOM modality performed procedure step. Why? Efficiency, interoperability, and safety.

Note that nowhere does IHE or DICOM say that you have to push the images to a workstation. How will you know which workstation gets which images? Will the technologists remember to push all the images to each of the workstations? Automatic push to all workstations? Yikes, how will you correct mistakes? IHE says, among other things, that the modality must talk to image manager/image archive systems. As we will shortly see, the workstation can talk to them too.

MORE PURCHASING DECISIONS

Now, truth be told, there are a few other things you have to think about when purchasing a digital mammography acquisition modality. You want the modality vendor to do any proprietary post-processing of the images on the acquisition modality, even if you are buying workstations, and other systems, from that same vendor. Why? Because modalities make two kinds of images: For Processing and For Presentation. Acquiring the data from the detector and converting it into a For Processing image is usually a proprietary, closely guarded, and highly FDA-regulated activity. Converting a For Processing image into a For Presentation image (given many different names by many different vendors) is also a proprietary, closely guarded, and highly FDA-regulated activity. But the display of For Presentation is only a lightly FDA-regulated function and many vendors are able to display For Presentation images from multiple vendors. This is the logical place to make a break between the modality and the other systems.

You can then decide if you want to store the For processing images and the For Presentation images, and there are arguments both ways. I tend to lean toward the prostorage camp: storage is cheap, you can pull the For Processing images back to the same vendor’s modality and reprocess them (if necessary) when the vendor introduces a new algorithm or for research. The antistorage camp might say having the capacity to reprocess the data means that you must reprocess the data every time the algorithms change or face potential liability. OK, modality…done.

CAD AND ALL IT MEANS

Now, we will move down the chain to computer-assisted detection (CAD). What good is a shiny new FFDM without an equally shiny, new CAD? Your chief of mammography has picked out a couple of units that her colleagues use. Your physicist has evaluated the sensitivity and specificity claims of the vendor. The vendor asks that you push For Processing and/or For Presentation images to the system from the modality. They will then push the results to the radiologist on the workstation. Which CAD results to which workstations? Who is going to store the CAD results permanently, and how will they get there from the workstation? IHE says use IHE Postprocessing Workflow. How does it work? Make a worklist (postprocessing worklist), claim an item on the worklist, tell when you start (postprocessing performed procedure step), retrieve the image objects, do your thing, store your results (DICOM mammography CAD objects, right?!), commit your results, tell when you are done, tell what you did. Sound familiar? Just like IHE Scheduled Workflow.

THE WEAKEST LINK

Now we come to the workstation, the eye candy, and the weakest link; we need it to do so many things that it just does not know yet how to do. First things first: Does your PACS vendor or your digital mammography modality vendor make the best multimodality (do not forget breast CT, MR, US, PET, etc) diagnostic imaging workstation? Maybe, but maybe not. Do you buy all your modalities exclusively from your PACS vendor? I doubt you do, and IHE and DICOM make sure you do not have to. Why should you have to buy your diagnostic imaging workstations (for any area in your department) from the PACS vendor? Because, until now, efficient (relatively speaking) diagnostic imaging display means worklist-driven operation and those worklists are typically proprietary.

Did I hear, “What about the IHE Reporting Workflow integration profile?”

Precisely. The IHE reporting workflow integration profile specifies the transactions necessary so that any workstation can be driven by DICOM standard worklists. I love the smell of IHE in the morning.

Now, I have to be the first to admit that mammography display remains a challenge for many workstations, even those intended for mammography. I will also admit that a new IHE Mammography subcommittee of IHE Radiology has been formed to specify what needs to be done to ensure consistent display of mammographic images. You will be hearing a lot more about this activity in the near future.

Until then, you need to ask a few questions of your workstation vendor. Do they support:

  1. IHE Reporting Workflow;
  2. automated retrieval of historical studies based on the worklist selection;
  3. display of For Presentation images from any modality vendor;
  4. accurate size display of comparison images from different vendors with different detector sizes;
  5. correct, automatic orientation of all views from all vendors;
  6. DICOM value of interest (VOI) lookup tables (LUT);
  7. correct sigmoidal window/level manipulations;
  8. correct display of legally mandated annotations?
  9. Does the background get window/level with the breast (annoying and distracting)?
  10. How are multimodality hanging protocols implemented (did you know there is a DICOM standard for hanging protocols?!)?

Remember, we in radiology live in a market-driven engineering world. If you ask for functionality before making a purchase, the vendors will deliver it. With integration reliably distributing relevant information, vendors can focus on providing new features and functionalities to advance the art and science of medicine. What will the future bring? Stay tuned.

David S. Channin, MD, is associate professor of radiology, and chief, imaging informatics, Department of Radiology, Feinberg School of Medicine, Northwestern University, Chicago.