Medical University of South Carolina (MUSC) in Charleston, owing to its three-pronged mission of training clinicians, providing care, and conducting (currently) $180 million worth of biomedical research, is a busy institution. In 2005, its three hospitals together logged 30,000 inpatient admissions and more than 700,000 outpatient visits, so it comes as no surprise to learn that MUSC with its almost entirely digital fleet of imaging systems generates massive amounts of radiological data.

The man responsible for making sure those electronic images and related records are easily accessible and kept safe at all times is chief information officer Frank Clark, PhD. Clark holds the view that the management of radiological data is best undertaken if approached as a customer service rather than a mere daily exercise in the operation of information technology. An interesting philosophy, to be sure—and one of the reasons why Decisions in Axis Imaging News chose to spend some time with him discussing data-management approaches. In particular, we were curious to learn how his department is responding to the fast-growing volume of data that needs to move from point to point to point across the MUSC enterprise.

IMAGING ECONOMICS: How much radiological data does MUSC currently produce?

CLARK: In a year, about 14 terabytes, uncompressed.

Frank Clark, PhD

IMAGING ECONOMICS: What is the archiving strategy you’ve adopted?

CLARK: Using MODs [magnetic optical disks], we keep images online until they are 9 months old. After that, they’re transferred to a tape library.

IMAGING ECONOMICS: What about data integrity? What are you doing to ensure that the data you collect and store are uncorrupted?

CLARK: First, when we store data on the MODs, they are compressed, but only at a 2:1 ratio—we feel that degree of compression is enough to make our storage more efficient without increasing the risk of incurring data corruption and other problems. Also, our data-capturing devices are frequently checked for calibration, which helps ensure that we have good quality data as they go into storage.

IMAGING ECONOMICS: And how are you ensuring the safety of your data?

CLARK: We back up our data very frequently, and have redundancies built in, such as mirroring. Charleston, as you probably know, is subject to hurricane flooding, so we’ve taken the precaution of locating our data center on the third floor of the building—we don’t expect ever to see flood waters reach that high, not even in the worst of storms. Moreover, we’re currently looking at installing a redundant parallel archive system at a site 250 miles inland as an extra safeguard.

IMAGING ECONOMICS: We are told that image data collection at MUSC has been doubling in size each year since 2000.

CLARK: Yes, that’s correct.

IMAGING ECONOMICS: To what do you attribute that growth?

CLARK: The number one reason has been the improvements in imaging modality technology. Our imaging systems are producing better pictures that show more than ever and are thus more useful than ever, so doctors are ordering more studies—especially those involving multislice images. We estimate that, this year, the data collection will grow by 50%—about 7 additional terabytes—because we’re adding digital mammography and a PET scanner to our service mix. Data collection will grow again in 2007 by a significant amount as a result of the opening of a fourth hospital we right now have under construction.

IMAGING ECONOMICS: What sorts of challenges does this much growth present to you in the information services department?

CLARK: The biggest challenge is staying ahead of the demand curve. The obvious way that is done is by continually increasing our storage capacity. Fortunately, we have an archiving system that is very scalable. Additionally, we keep improving our bandwidth capabilities—we want to make sure we always have the ability to quickly move those ever-larger image files through the pipes, and that requires ample bandwidth. We’re also exploring the use of other types of storage media for our online data: we’d like to be able to extend the time an image is available online to a full 12 months. With regard to the images archived on tape, the challenge there is to reduce the time it takes for retrieval. To address it, we’re in the process of upgrading our tape infrastructure from LTO generation 1 to generation 3 technology. The most important step we’ve taken—and I’m talking here about all data, enterprisewide, not just radiology’s—has been to move away from a siloed approach for managing data.

IMAGING ECONOMICS: If you are moving away from the siloing of data, what then are you moving toward?

CLARK: Integrated storage. In mid 2005, we acquired storage access network [SAN] technology along with a state-of-the-art data management system. These are being implemented now, with completion expected by June. When fully operational, we’ll have dramatically greater efficiency of storage, plus economies of scale.


CLARK: In the siloed approach, you have direct-attach storage where each system has its own archive and its own separate access to the storage network. If, for instance, you went out and bought a materials management system and you bought the hardware on which to run it, that configuration would have a certain amount of storage that could be used only by that management system. But, over the course of time, you would discover that relatively little of the system’s storage was ever actually used, leaving you with a significant amount of surplus storage that could be used by no other system. By contrast, with a SAN model, you have a big, central depository of storage that can be partitioned and allocated for materials management, medical records, clinical documentation, lab, pharmacy, you name it. There’s no wasted storage capacity. That translates into an economy of scale. And a SAN approach is readily scalable, so you can add terabytes as needed to keep pace with growth.

IMAGING ECONOMICS: There are radiological data and there are data from other departments. What are the key differentiations inherent to managing each of those?

CLARK: Size is the most obvious difference. Radiology image files are many times larger than, say, the billing records coming out of the accounting office.

IMAGING ECONOMICS: Does that mean the primary consideration for any enterprise storage solution you develop is radiology?

CLARK: Yes. It’s the tail that wags the dog. If you plan to accommodate that, you can easily handle anything else that comes along.

IMAGING ECONOMICS: What are some of the ways you gird up in the IS department for radiology equipment acquisitions, implementations, and roll-outs?

CLARK: For starters, we try to get a handle on demand trends 6 to 12 months out so that we can appropriately factor that into our storage planning. Afterward, we remain in close communication with the radiologists and look at their actual volume and utilization trends. And we periodically undertake a review process of what’s currently being stored online—many hospitals and IS departments have the mistaken view that storage is cheap, so they go ahead and store everything on the fastest storage medium, which also happens to be the most expensive. That’s how we used to look at it too. Lately, however, we’ve begun to system by system look at what we currently have in storage, asking ourselves whether any of it can be stored instead on tape or otherwise moved off from the expensive online archives—tape being a comparatively less expensive medium for storage. We’re still in a very early stage with this, but already we’ve recovered and made available for reallocation many hundreds of gigabytes of online storage space.

IMAGING ECONOMICS: Do you periodically reassess your overall infrastructure strategy?

CLARK: Yes, quarterly. It was through just this process that we uncovered reasons to implement our SAN strategy.

IMAGING ECONOMICS: Who participates in these reviews?

CLARK: Key people in the affected departments. In radiology, for example, it’s the radiologists and the individuals responsible for the operation of PACS and RIS.

IMAGING ECONOMICS: How do you communicate among yourselves so the work can go forward in a smooth way?

CLARK: It’s not easy. Just as we used to have storage silos, we have what I call communications silos. We’re trying to break those down and get people talking more. One way is by encouraging the formation of groups of people from across the enterprise who have particular interests in common. We call them SIGs—special interest groups—that meet once a month.

IMAGING ECONOMICS: How do the fruits of the labors of the SIGs get acted on or implemented?

CLARK: I’ll give you the example of a SIG that is dealing with data storage issues. On that SIG are members of the medical center’s IT Governance Infrastructure Council. Consequently, whatever positive ideas come out of that SIG will end up being directly relayed to the infrastructure council and, from there, on up to the Information Management Council, which conducts decision-making about the strategic directions we want to take.

IMAGING ECONOMICS: Just to sum it all up, what do you see as the most important aspect of radiological data management?

CLARK: The strategic planning. But, to plan well, you need a true understanding of the situation you’re in. That’s why good communication between data users and enterprise decision-makers is so vital to staying ahead of the demand curve. The last thing you want is to be taken by surprise on any level where the management of radiological data is concerned.

Rich Smith is a contributing writer for Decisions in Axis Imaging News.