Long-Term Image Archiving

Images are getting larger in volume and quantity. As state and national laws mandate that healthcare organizations store these gigantic files for multiple years, storage system vendors are souping up their technologies that more efficiently manage dozens upon dozens of terabytes of data. And most hospitals today set up their PACS archive so that new studies are stored online for near-instant access and historical studies are moved to a reliable, cost efficient long-term data repository - keeping in mind in the long run that accessibility and retreivability are key.

Transitioning from a film room archive to an electronic storage system has its advantages, as well as challenges. The benefits include faster retrieval of historical studies, comparison of relevant priors, better management of images produced by new digital modalities, increased productivity and streamlined workflow.

On the other hand, storage technology is advancing quickly and prices continue to decline. What's "in" today may not necessarily be the technology choice of tomorrow. So how do you find a high-performing archive that is scalable, reliable and secure? And don't forget to factor in obsolescence protection and data migration, disaster recovery, business continuity and overall cost of ownership.

The good news is there are many options hospitals have when it comes to archiving medical images in a PACS environment. While a larger facility may opt for a system that stores new medical images on spinning disc technology and older studies in an onsite and offsite data repository, specialty hospitals and imaging centers may select to outsource their images to an offsite storage service provider (SSP).

To select the right storage system, decision-makers need to consider: "The size of each element of data that will be stored; the speed in which it needs to be retrieved; the cost per gigabyte for that storage; and its integration with different applications," says Tomas Hough, principal of True North & Consulting Associates. "Knowing these four points, you can then start to weigh out what is the best application for you given what your needs are."

One approach to storing medical images uses a hierarchical storage management (HSM) system where regularly accessed files are stored on high-speed hard drive arrays and then gradually migrated to some form of removable media for long-term storage. The declining cost of fiber-channel and ATA-based drive arrays, such as NAS (network attached storage) and SAN (storage area network) have started to change the paradigm.

Estimating that the diagnostic imaging department at the Medical College of Wisconsin would produce roughly 4 terabytes (TB) of data for every 100,000 studies, the facility installed a super advanced intelligent (AIT) tape library for their long-term archive. For online storage, the facility uses spinning disc on NAS. "We never ever pull from [the tape library]," says Paul Nagy, assistant professor of radiology and director of radiology informatics laboratory. "The only thing we use the tape for is a legal copy and or disaster recovery. We use an ATA drive on the NAS and we get away with that by saying that this is our convenient copy that is being used for interpretation and clinical use."

Long-term archives facilitate data migration, something Nagy insists all healthcare organizations must include in their storage plan. "Be aware of migration and exit strategies so that when you go with another vendor, you know how much is it going to cost to get the data back," instructs Nagy. "Don't expect that this is going to be your last vendor. Expect that this will be one vendor amongst many. It won't be your last archive either, so plan for the migration ahead of time because usually it's the most painful part."


HOW WILL MY ARCHIVE GROW?

When it comes to storage costs, there are three factors to consider, says Barbara Dumery, senior product manager of eMed Technologies: hardware for the server, software and the cost of storage. "The cost of storage depends on the type of storage it is and how much is needed," explains Dumery. Capacity is more often an approximation at the time of installation since image volume in most facilities is continuously increasing. As a result, the total cost of ownership for a hospital's long-term archive typically includes both initial and reoccurring costs.

"In the beginning, we did not anticipate annually that we would grow 25 percent," states Scott Dent, IT PACS Administrator at Bronson Healthcare in Michigan. But that was before the installation of 4- and 16-slice CT scanners and fixed MRI unit that required the facility to enlarge its storage capacity.

Adding more did not spark any major technical problems because of the scalability of Bronson's storage system. Bronson's diagnostic imaging department uses eMed's PACS and Network Appliance Inc.'s NAS F820 filter for short-term storage and NetApp NearStore R200 for long-term archiving. As a fail over, they currently write to MOD (magneto-optical disc), but it is being phased out with the implementation of another NearStore R200 in an offsite location.

To boost storage capacity, Dent uses a Web interface to add drives dynamically without the user even knowing it. "It's a system that is designed to handle multiple trays and you slide in the tray and put 16 drives on that tray," explains Dent. "You can activate one or all of them on request."

Medical images are simultaneously stored in both the short and long-term archives. The studies remain online for up to eight months. But are always retrievable and permanently stored in the deep archive. "The archive pulls in the study and writes it onto short-term storage on the NAS," says Dent. "We replicate our database to another archive server. That archive server handles the archiving of images long-term with a storage management tool (Amass for Unix by Advanced Digital Information Corp.). That writes the image out to one NearStore, and then that system sends it to another NearStore. The devices automatically replicate all by themselves. If we put an image on one system, it will automatically copy it off on the other system, making our long-term archive 100 percent redundant."


DEFINING THE TERMS OF YOUR ARCHIVE

The IS department at St. Anthony's Medical Center in St. Louis, Mo., went into PACS knowing that they did not want to manage disparate storage systems. "Since we were already using IBM's SAN [based on the IBM TotalStorage Enterprise Storage Server (ESS)] in our IS data center, we asked the vendor if they would substitute the device that they were going to sell us for another [ESS]," says Rick Faszold, manager of IS services.

The device stores St. Anthony's diagnostic images for up to 18 months before it is sent into the deep archive. IBM's Tivoli Storage Manager (TSM) keeps tracks of the SAN's capacity and moves it onto tape when it reaches its high water mark. Faszold says the facility opted to use tape for its deep archive due to the size of the medical facility. "Our primary concern about moving to DVD was the number of DVDs we would be creating on an annual basis," explains Faszold. "We decided to use high-capacity tape drives. We have a six drive tape library and right now each tape will take 200 gigabytes. We will eventually upgrade to the new ones that hold 400 to 500 gigabytes."

Faszold adds that using tape requires a PACS that can pre-fetch historical studies since the access time of the media is slower than optical technology. "When you are looking at a tape library system, you slow your retrieval time into multiple minutes," says Faszold. "What you positively have to do is make sure that your PACS [recognizes] the patient orders that come across so that it will pre-fetch any images related to that patient. If a patient comes in after his/her image has been migrated off to the deep archive, you have to find a way to make sure that as soon as the radiologist is ready to read that exam again, the [historical study] has already been dug up from the deep archive and is immediately sent to the right workstation."


THE ASP ROUTE

When looking at what would be the best long-term storage alternative for Children's Healthcare of Atlanta, the hospital selected the route of an application service provider (ASP). Through their PACS vendor, GE Healthcare, the hospital utilizes Quest Diagnostic Inc. for offsite long-term storage of their pediatric images. Children's is charged a fixed fee per procedure. In 2003, Children's performed 200,000 diagnostic imaging studies, which includes MR, CT, ultrasound, nuclear medicine, computed radiography and fluoroscopy.

"As a pediatric hospital, we are required to store medical record information for 5 years beyond the age of majority," explains David Moyer, IS/PACS administrator for diagnostic services. "With newborns, we have to store their images at least 23 to 26 years. That would pose quite a bit of storage issues over time, especially as the technology changes over time."

To ensure the network is both reliable and secure when transmitting images and pre-fetching images from the offsite archive, Children's installed a point-to-point DS3 connection. "Most of our physicians in-house like to see [the studies] that are most recent," says Moyer. "Our radiologists are going to want to see a lot of previous images that have been archived [long-term], so we have set up our system so that it will pre-fetch any images on the long-term archive back to the short-term archive."

Children's factored into its PACS agreement the ability to migrate the images to some form of in-house storage if they ever desired. "Any time that you are using an ASP solution, it's with the sight in mind that it may not be a long-term commitment," opines Allana Cummings, IS director of information systems and technology. "You must make sure that there are built-in options to migrate data back in-house."


STORAGE LEVELS

The Florida division of Adventist Healthcare Group, consisting of seven hospitals and three imaging centers, produces 1.2 million imaging studies annually (roughly 32 TB of data) after implementing Agfa Healthcare's IMPAX PACS. The image management system and digital modalities in turn increased the storage requirements of the hospitals, which initially consisted of RAID for short-term and a tape library for long-term storage. "Our requirements actually grew," says Fred Saberian, manager of IT Operations. "Once we had physicians and clinicians on board with the system, their usage exponentially grew as well." As it becomes easier to store, the propensity to store more grows.

Knowing that they would eventually go live with cardiology on PACS, Florida decided to change its storage system rather than increase its capacity. Florida implemented StorageTek's BladeStore BC84 blade system and SL8500 modular library system. In addition to redundancy and speed, Saberian says the three-tier approach to storage with the SL8500 on the end allows for one single point of management for all of their tape archiving requirements.

Greenville (S.C.) Hospital System (GHS) will soon implement a similar three-tier system. Currently the facility uses StorageTek's 9176 full-Fibre RAID disc Subsystem for patient images that are up to a year old. "That way when the patient is active and people are really looking at those studies, you are getting sub-second response time and really fast retrieval rates," notes Tip Jones, technical director at Greenville. "Then there is an interim period when the patient is discharged and physicians may still be looking at the studies, but not nearly as much. What we are looking at doing is using a combination of high speed disc and BladeStore as an intermediate disc. We may keep one year on high speed, two years on intermediate and then go to deep archive."

BladeStore technology uses low cost ATA/IDE disc drives for data storage rather than higher cost SCSI or fiber-channel disk drives traditionally used in enterprise disc systems. "You want to be able to get the most storage with the quickest response time for the least amount of money," opines Jones. StorageTek's B-series is designed as a low-cost alternative for high-capacity secondary disk or fixed content storage and application systems.

The industry is experiencing a trend where storage equipment vendors are partnering with PACS providers, and both hospitals agree that this has helped facilitate their projects. "It is very important that your PACS vendor partner cooperate with your storage vendors so that you get the best information," says Brad Moody, senior IS manager at Greenville. "The [systems] need to integrate to give PACS an image quickly, timely and reliably."


FIXED CONTENT

Beth Israel Deaconess Medical Center (BIDMC of CareGroup Healthcare) purchased their storage system through their PACS provider. Previously using MOD for long-term storage, Beth Israel Deaconess installed EMC Corp.'s Clariion for short-term and Centera for long-term storage. The systems utilize a relatively new "fixed-content" approach to data archiving called content addressed storage (CAS). When a patient's image is stored, a unique identifier is generated. The identifier is used to retrieve the images without the user having to know where they are stored. In addition to the two systems, everything that is stored on the Centera is replicated onto DVD in a jukebox in another building.Beth Israel Deaconess has 65 different modalities attached to the PACS and about 70 radiologist workstations dispersed throughout the campus. According to Ron Mitchell, director of IS at the , the quick retrieval rate of images from the Centara compensates for the cost of the hospital spinning disc technology (even though the overall cost of the technology has declined significantly over the years). "Spinning disc technology used to be far too expensive to consider as a long-term archive but the cost continued to come down to the point where the system was very attractive for us," notes Mitchell.

With the Centera in place, the healthcare organization is now undergoing a massive data migration project to transfer 10TB of data off the MOD archive. The year- long project will be managed by their PACS vendor since they run Beth Israel Deaconess' data center. "It is going to be a slow process because we want to do it at night or other times when the system is not being fully utilized," explains Mitchell.