Image Archiving Needs Explode

Boosts in the number of imaging procedures nationwide as well as increased volume per study have led to the need to manage more images than ever. Plus, more and more radiologists require the ability to view images remotely as do other specialists making use of images for diagnosis, surgery and other therapies. Combine that with an increased focus on security, on-demand availability and redundancy and you'll find that most healthcare organizations are re-evaluating their image archiving.


Accommodating the ongoing flow

Austin Radiological is an outpatient, physician-owned radiology practice that owns and operates 14 imaging centers in the Austin, Texas, area. The 62 radiologists also do professional readings at all of the area hospitals for a total of more than a million studies a year.

The practice uses EMC DiskXtender software to automatically migrate PACS images that are more than 30 days old from its EMC Symmetrix system to the EMC Centera. They are currently moving about 16 gigabytes of images per day onto Centera, keeping production capacity open for the ongoing flow of new PACS data and ensuring free storage space on the Symmetrix box.

Austin Radiological implemented this solution when they started running out of storage space on their primary storage area network (SAN), says CIO Todd Thomas. "We knew, at some point, we had to look at migrating images off our storage network to some other archive."

Deciding on this solution "came down to the type of hardware platform to deploy on the backend," says Thomas. "We already had an EMC SAN solution in place, so it was somewhat attractive to stay with the same vendor throughout all storage subsystems. When we evaluated different proposals, we found that the total cost of ownership was equal to, if not less than, other solutions we were looking at."

Thomas likes that the Centera is disk-spaced, self-healing, and has the ability to enforce compliance on retention of images. Also, he likes that, as the organization runs out of space on a Centera frame, they can simply add another frame and continue the migration. Because Centera is network-based, DiskXtender knows where the media is at any given time.


Starting off with redundancy

Medica Imaging Centers runs four independent diagnostic testing facilities in Atlanta and Birmingham, producing about 10,000 studies a year. When the organization got started in 2001, the first center opened in a filmless environment. They didn't want to start archiving images on site because they didn't know how many centers they'd eventually have, says Kay Baker, FAHRA, vice president of operations. Plus, "we couldn't predict how busy we'd become and how large of an archive to purchase if we kept it locally on site," she says.

The goal was ensuring redundancy. Baker was impressed with InSite One's business plan and model and felt that the then-burgeoning company had the necessary technology. "They were starting off with redundancy. At that time, a lot of people weren't offering dual data repositories."

As Medica Imaging grew, InSite One came in to install another server and take care of the technology hook ups. Studies up to a few months old stay in the local cache. The system is set up so that all studies go to the server during the day and the server sends to the archive at night. Baker uses a T1 line because the radiologists do a lot of remote reading off-site.


From direct attached storage to SAN

The Medical University of South Carolina is in the process of implementing the IBM Shark enterprise storage server, says Frank Clark, vice president of IT and CIO. The facility's 23 radiologists read about 10 petabytes of imaging studies a year and wanted to move imaging applications off of direct attached storage to the SAN.

The school already had made the decision to run all of its clinical systems on IBM platforms, so after looking at the top direct attached storage vendors, they decided that it made sense to go with IBM for this as well, Clark says. The SAN, as opposed to having information silos, is more economical, helps with HIPAA and JCAHO requirements, increases productivity, and meets broader security needs.

Clark says he felt that security was adequately protecting in his old legacy framework, but thinks that "the storage access network allows you to do even better as far as protecting security, integrity and confidentiality."

Clark's storage solution includes three tiers — 13 terabytes for high performance/high utilization; 35 terabytes for medium performance/economical disk; and 140 terabytes for tape and disk archiving. Image studies are kept on the high performance tier for nine months. The facility also plans to move all of its applications — such as medical records, and clinical, administrative and financial systems ó off of direct attached storage to the SAN. This allows for allocation of  storage space to whatever application most needs it.


Departments centralize storage

Timing and consulting advice led to the installation of enterprise storage at Springfield Memorial Hospital in Illinois. Both the radiology and cardiology departments were looking to upgrade their PACS. Meanwhile, "our current storage was filling up and couldn't be expanded any further," says Jesse Whitehead, IT manager. The storage for the facility's front-end workstations was five years old. "It was time to do a forklift upgrade."

After looking at several vendors, Springfield decided that Kodak would be the easiest to upgrade to and had the features they wanted. The two departments decided to combine their storage needs but each wanted their own front-end access to centralized storage.

This solution has worked out well. IT already had set up a centralized storage platform that handles storage for departmental systems. Each department can use the PACS they like best. However, because DICOM is not a perfect standard, they have to adjust the settings for each vendor product they choose to tie into Kodak's enterprise image management. "It's very configurable so the door is open to do anything we want," says Whitehead.

Springfield also used Kodak's enterprise information management software for storing PACS images. The system lets PACS Administrator Tracy Brown-Hagmann set up protocols as to how different images are going to move throughout the lifecycle.

Storage plans were developed based on retention policies and rules were based on where the image was acquired. "When a study comes in, it goes through the list of filters," says Brown-Hagmann. "Once it makes a match, that image stores to a particular place on enterprise storage. Several copies of every study are written on different media so redundancy is built in."


Redundancy & customization of images

Centralized storage was the best path for Fletcher Allen Health Care in Burlington, Vt., as well, says IT Director David J. Haber. He began working with Network Appliance (Net≠App) more than five years ago for network attached storage. Over the years, Fletcher Allen has grown from one storage device to nine.

Haber first went with a Compaq SAN and NetApp for network attached storage. "People were reluctant to believe NetApp could also be a SAN," he says. But after experiencing both, he decided all storage would go on Net≠App devices. "Because of [their] features, we can centralize our storage and also take the first steps toward good disaster recovery." The system is customizable so Haber could allow each radiologist to view images in his or her preferred manner.

Disk storage is much faster with better performance, rather than pulling images off of tape, he says. Haber set up a secondary cache for images for when the image servers fill up. Images go to the archive at least once a day — radiologists are supposed to sign off on their studies within 24 hours. Once they do, the study goes to the archive. "Every study we've ever done is on that archive," says Haber, "but backup is outside PACS." Until the radiologist signs off, he or she has a copy on the modality and on the server itself. "All of our image servers and database servers are built on clusters," he says. With two servers connected together, they share the data. If one goes down, the other takes over.