When You Need the Image Now: How much storage is enough?

Image storage is a bit of a conundrum; if cost weren't a factor, healthcare organizations could store all of their digital medical images on the highest-performing, most reliable storage media on the market. But since cost is always a factor, medical imaging archives must incorporate ingenuity in their design, including lower cost media while still meeting end-user requirements for accessibility, reliability, availability and performance.

Image archiving is challenging; albeit not impossible. Some 20 percent of U.S. healthcare facilities have deployed PACS, according to industry estimates, and all are using a combination of storage hardware and storage software to automate management. But what constitutes a robust, reliable, scalable and affordable PACS storage plan? Strategizing can be arduous as technological improvements change options and healthcare organizations - large, medium and small - have different digital imaging, PACS and archiving needs.

But there is some common ground. PACS storage consists of short- and long-term caches that securely and reliably disseminate medical imaging data when needed to physicians and clinicians throughout the enterprise and remotely. The media chosen for each reflects the speed with which the image data needs to be accessed.

In the past, a healthcare facility may have had a couple years of long-term storage on slower-access, lower-costing media; putting the online archive on fast spinning disk. A tape unit may have backed up the long-term archive. Once the study was validated, the data would be moved from the online archive to the long term archive over to the disaster recovery. However, the declining cost of storage media, such as spinning disk, is changing the paradigm.

Every PACS team must evaluate the amount of data stored, retrieval times, costs and integration with various applications. State and federal laws will influence how long electronic patient images must be stored. Also, users should adopt a "buy as you need it" approach to storage media to mitigate the difficulty of data migration and avoid data obsolescence.

Storing smart

Trends in long-term storage are impacting PACS archiving. Exponentially increasing storage requirements due to the introduction of new imaging technology and other 'ology' image storage beyond radiology is one. The healthcare community is starting to look at spinning disk as a viable option for both short- and long-term needs as costs decline and capacities increase. Medical imaging data are getting replicated to off-site facilities for disaster recovery and business continuity requirements. Architectures are expanding to enterprise archives, rather than individual silos of storage managed separately by departments.

Policy-based data management, which permits medical imaging studies to be intelligently filed during its lifecycle, is another influential long-term storage trend. Phoenix-based Banner Health, which manages 21 facilities across seven states -Alaska, Arizona, California, Colorado, Nebraska, Nevada and Wyoming - uses this technology to cost effectively meet retrieval requirements and economically optimize storage costs.

Banner's greatest concentration of facilities and procedure volumes are in Arizona. In total, they perform more than 800,000 radiology and cardiology exams annually. Transitioning to digital imaging, Banner's Arizona facilities needed a storage infrastructure that could easily scale to 1.2 petabytes - its projected imaging volume by 2009.

Tim McFarlan, director of technology management, says the enterprise-wide project has high availability, multi-facility requirements. Retrieval of mission-critical data must be instantaneous as hundreds of clinicians access the archive daily. Scalability is crucial since huge amounts of electronic data will be produced annually. Security and data integrity must be maintained so that state retention mandates and federal government regulations, such as Health Insurance Portability and Accountability Act (HIPAA) requirements, are met.

McFarlan says that the storage must interface with multiple PACS. "We quickly realized that if we were to develop a common archive solution for Banner's Arizona-based facilities, we needed to find an archive that would support both the archival retrieval process of radiology PACS and also other DICOM-based image management systems used in other imaging-related departments, such as cardiology," says McFarlan.

Banner selected a multi-tiered storage hierarchy based on Hewlett Packard (HP) StorageWorksModular Smart Arrays integrated with HP ProLiant 380 servers.

The infrastructure is driven by Bycast's StorageGRID software that integrates with Banner's various image management systems. Information lifecycle management (ILM) policies allow medical images to be stored in the right place, on the right media, at the right time.

ILM - a buzz phrase in the storage industry - takes the approach that not every piece of patient information is needed at all times, so different systems can be used to prioritize the access and storage of data. ILM enables Banner to move older imaging studies to less expensive storage that still provides reliable access, but costs less to acquire and manage.

"The [technology] gives us flexibility so we can really tailor the storage requirements and optimize it for different modalities," says McFarlan. "Our storage strategy gives us the flexibility to isolate different storage management solutions for the different modalities, satisfy the clinicians, and at the same time in the IT world, manage different storage tiers."

To exist is to change

More healthcare organizations are adopting an "always online" approach to PACS storage and realizing the clinical, financial and IT-related benefits. Just ask Dirk Anderson, radiology manager of CT, ultrasound and PACS at St. Mary's Hospital Medical Center. He oversaw the implementation of Stentor's iSite PACS at the 380-bed facility in Grand Junction, Colo.

The installation included Stentor's iVault online long-term storage technology that employs an online data storage model to house medical images on an active server, thus permitting doctors immediate data accessibility throughout the enterprise. "Any study under 500 megabytes in size is guaranteed to be presented to the physician in less than 2.5 seconds," says Anderson.

"Online" storage is comprised mostly of spinning disks (redundant array of independent disks, RAID). The older HSM (hierarchical storage management) model manages data across a storage hierarchy that consists of high-performance disk, compressed disk, and tape libraries. Many implementations of online archive models use SAN (storage area network) or NAS (networked attached storage).

"Right now, St. Mary's has three year's worth of electronic patient images stored on spinning disk," says Anderson. "We back up for disaster recovery to an off-site facility, but locally we do not go to DVD or employ an HSM model where studies are dumped off the cache and go to a deeper archive. Everything is always contained on a hard drive." Future plans include bringing cardiology and pathology images online as well.

Paul Chang, MD, director, radiology informatics at the University of Pittsburgh Medical Center, says the advantages of an on-demand archive model are significant. UPMC's medical imaging department was one of the first to deploy an online PACS archive several years ago for its 19 hospitals that produce more than 1.4 million studies annually. "An on-demand archive ensures that all studies, even older prior images from several years ago, are immediately available without the penalty of data migration from slower media to disk. This is important for enterprise image delivery, where physicians want older studies immediately," says Chang.

Other advantages include high performance, simple archive design, decreasing costs and increasing capacities.

Keep in mind

"If the archive is not carefully designed, the required RAID disks could be very expensive. In order to work economically, the online model must be able to use relatively inexpensive RAID devices," Chang explains as a potential drawback of online storage. Redundancy also is key as medical imaging data are housed online for long-term periods. Hard drives are hard drives, and just like any PC or laptop computer, healthcare organizations must implement a permanent backup strategy.

At St. Mary's, medical imaging data resides in multiple locations. There are three copies of the images internally stored on RAID - immediate (weeks), short term (months) and long term (seven years). Anderson adds that a VPN between the medical center and a data center in San Jose, Calif., allows all imaging studies to be pushed off-site for disaster recovery purposes. "Stentor guarantees that they can get us a completely repopulated RAID model back onsite within 96 hours," Anderson says.

Paul Nagy, PhD, director, informatics research at the University of Maryland Medical System, says that while the total cost of ownership for spinning disk is lower than tape or magneto optical disk for PACS, users also must understand the cons and how to mitigate them.

"You need to have a plan to not have disks spinning for more than four years," explains Nagy. "Most hard drives [SATA or SCSI] are rated at 1.2 million hours between failures. If you don't have the right environmental conditions, you can get much less performance. Especially watch the temperature in the server rooms. Air conditioning is a must. Also, disks are not write once media and can be written by malicious worms. You should use snapshotting and write once file system."

Not all vendors provide spinning media solutions. Online archives count for a fraction of PACS deployments, but this is expected to change. Other vendors tout a more HSM-based approach where data are tiered. Medical imaging data gets stored on different media - online, nearline and offline - at different locations (or different media types) and are migrated based on defined business rules and policies. Nearline data, typically stored on tape or optical media, has slower retrieval times than online media. Offline storage refers to removable media that has been removed and stored on a shelf.

Contents of an archive

Sutter Health manages 27 Northern California hospitals dispersed throughout five regions: Greater Bay Area, Greater Sacramento Valley/Foothills, Central Valley, North Coast and Clearlake. With plans to go filmless in both radiology and cardiology, the healthcare system needed an online data repository where digital medical images could be securely stored and readily accessible for clinicians across the enterprise.

Sutter selected a storage technology that is gaining a foothold in the healthcare market: content addressed storage (CAS). CAS uses write once, read many (WORM) capability to ensure that once data are stored they cannot be overwritten. CAS is designed specifically for fixed content data - data that never get changed once they are written - such as medical imaging files. CAS enables management costs to remain flat as imaging volumes increase. It also ensures long-term retention and authenticity of medical imaging data.

EMC's Centera CAS will be the short-term and long-term online repository for Sutter Health's medical images. "Each of our five regions is building its own common image repository that each hospital within the region can access through PACS," says Mario Garcia, PACS - eICU program manager at Sutter. "Medical images will be stored in the local image management system (IMS) for a short-term period of time on RAID while the patient study or images will then be sent to the long-term EMC storage system at each region."

Sutter Health opted to forgo a centralized data repository for all 27 hospitals, which Garcia says would have been an expensive, massive undertaking that required each hospital to beef up its network. Rather, each region has its own centralized repository. Combined, the five regions have a total storage capacity of 200TB.

For example, Sutter's Sacramento region consists of five hospitals. Each hospital has an IMS where images are kept for 3 to 6 months. The IMS has a total storage capacity of 2TB. Each image study also gets replicated into Centera, which is located in the Sacramento Data Center. "The system has 115TB of raw data storage capacity, but we mirror each study, so the actual usable storage capacity is 55.2TB," explains Garcia. The data center is miles away from each hospital, but radiologists, cardiologists and other clinicians can still access image studies within two to three seconds.

To achieve system-wide access to PACS data without central long-term storage, Garcia explains that Sutter will incorporate its Patient Study Information within its existing Enterprise Master Patient Index (EMPI) product. The EMPI will allow Sutter to uniquely identify its patients across the system with appropriate pointers to patient PACS studies stored within various local PACS.

"The integration of PACS study information within the EMPI and other systems such as the EMR and physician portal will enable Sutter physicians access to the longitudinal view of their patient medical records," says Garcia.


Medical imaging trends, such as increasing study volumes and healthcare enterprise archive goals, in addition to technological improvements and declining storage costs, are redefining the essentials of a PACS storage plan. More medical imaging studies are being stored online, longer and storage management software is allowing users to match the value of patient data with corresponding price and performance of the storage media. Ultimately, all hospitals, whether large, medium or small, are looking for archive systems that will help them improve patient care and safety and optimize physician's time.