For the University of California at Los Angeles Medical School’s Department of Radiology, migrating digitally stored images from an old picture archiving and communications system (PACS) to a new one has turned out to be a frustrating journey into the land of the unforeseen. A process once slated to take months, the migration is going to take years.

Osman Ratib, MD, PhD, FAHA, who heads the UCLA migration project, says there is a light at the end of the tunnel, “but I haven’t seen it yet.”

He estimates it could be another year and a half before UCLA will be able to transfer its archives and shut down its old PACS completely. Only then will the hospital fully enjoy the benefits of a new enterprise-wide PACS that will link departments, facilities, referring doctors, and patients in an image and report repository that will be a big part of the school’s EMR (electronic medical record). 

This is not to say that UCLA is sitting with its new PACS idle waiting for the old data to arrive. Part of what is making the data migration so difficult is that the new PACS is being deployed at the same time that the old one is being edged out.

Ratib says it might have been easier to build a new archive from scratch. But that was out of the question, he adds. UCLA doctors were used to having prior studies on their workstations, and to have turned the clock back to fetching film priors simply was not an option. The electronic priors had to be on the new PACS.

Ratib is a fascinating man. Egyptian by birth, he grew up in Switzerland. He still holds a visiting professorship at the University Hospital of Geneva, and he has retained his Swiss citizenship.

Ratib remembers the old days at UCLA when radiology professors and their graduate students were boldly constructing one of the first PACS in the world, engineering the system from the ground up.

“There were no PACS vendors then,” he says. “We were just a bunch of PhD students, and we developed a multiple-generation PACS over 10 years.”

Ratib took his first medical degree in cardiology. That was in 1979. By 1986, he was at UCLA working on a doctorate in medical imaging in the Department of Radiological Sciences.

“I got into PACS from cardiology imaging, and that led to the idea that I needed more background,” he says. “UCLA was the only place to offer a PhD in digital imaging.”

Ratib helped design UCLA’s first PACS and received his PhD in 1989. Then he returned to Switzerland to practice medicine. He helped build a Swiss PACS and wrote some PACS software that was widely used. In 1996 he became a board-certified radiologist.

In the summer of 1998, he returned to UCLA to take up his current position as professor and vice chair of information systems in the Department of Radiological Sciences. He holds a joint appointment in UCLA’s Department of Molecular Medicine, Pharmacology, and Nuclear Medicine. He says he spends about half his time on the information technology side and the other half seeing patients.

POINT A

The legacy PACS at UCLA when Ratib returned in 1998 was not a single PACS; it was a collection of intradepartmental PACSs in different settings. “But after 10 years of developing your own stuff, the original authors of the system are long gone and nobody knows how to maintain things,” Ratib says.

The medical school’s collection of old interconnected PACSs was operational but frustrating to use. For one thing, images could be prefetched only to predetermined workstations set up to handle particular interpretations, Ratib says. Radiologists had to run across streets on campus in order to read. In the 10 years UCLA had been building its system, biomedical physics in many vendors had come to the market with PACS arrays far more utilitarian and agile than UCLA’s old homegrown system. Ratib puts it bluntly, “Our old PACS was dying.”

Ratib says he was hired in part to “star t shifting gears” toward a new PACS. At the same time, UCLA was building a new hospital on campus that is due to open in 2005. Administrators wanted the new PACS up and running before the hospital opened. “You don’t go to a new hospital with a completely new IT (information technology) infrastructure, because it will fail. People will get lost,” Ratib says. “You want to move into the new hospital with something familiar, with the work flow worked out. You don’t want to go to the new environment where things don’t work or people aren’t happy.”

Even though he had been hired to change the PACS, Ratib says Point A for him was not the PACS but the hospital’s RIS (radiology information system). “If you want a PACS that works, you need to make sure that your information system whatever drives the requisition, the scheduling, the technologist, the report, and the billingis robust and solid before you start putting images in there. Images floating around without an information system are useless. They are not linked.”

The problem with installing the RIS prior to PACS was that it took a long timein Ratib’s estimate, 2 or 3 years. After a vendor selection process, the data from the old RIS had to be sent to the new RIS. The new RIS then had to be interfaced with the old legacy PACS systems until the new PACS arrived. “But you didn’t want to tie the new RIS too strongly to the old PACS because the new RIS had to be ready for the new PACS when it came in,” Ratib says.

“For that reason, we decided that the new RIS had to be web-based. Web technology allows you to be independent. The application doesn’t reside on the workstation so there’s no conflict with other applications on your workstation. It’s what we call thin client; the program is not on your computer, it’s on the web server.”

Before the PACS project had even gone out for bid, the migration of data involving the RIS had begun. This was the first step in UCLA’s PACS migration strategy.

POINT B

By spring 2002, UCLA had finished its PACS vendor selection process. But it was several months before the migration of imaging data could begin in earnest. For this to happen, other choices had to be made. How much data would be migrated? How would it be moved while keeping the old PACS operational? Where would it be archived?

On the question of where to store data, Ratib says he lobbied hard to use an ASP vendor (application service provider) to handle the archive. This resulted in an outsourcing contract where the ASP vendor essentially took control of the archive, warehoused it, and made it accessible for a fee. The reason for using the ASP model was threefold, Ratib says. First, UCLA would be liberated from the headache of archiving. “We had been through so many difficulties running our own archive through different generations that we thought that the real value of an ASP was we won’t have to worry about it anymore.”

Second, the ASP vendor would have to take on the responsibility of upgrading technology over time. “Our goal was that we didn’t want to hear about anything regarding technology.” The ASP vendor would have the incentive to upgrade to keep its costs low.

Third, and most important, UCLA wanted to avoid what had plagued it in the past having its archive tied to its crumbling PACS deployments. It wanted the imaging archive separated from the new PACS for the same reasons of flexibility that it had chosen a web-based RIS. The goal was to have the ASP vendor agree to store images for the long-term. Ratib typifies this as 15 years. If during the 15 years UCLA wanted to change its PACS, it wanted to be able to do so without having to migrate to a new archive all over again. The ASP vendor would maintain the archive. If there was a PACS change, the new system could then be easily interfaced with the existing archive. “We wanted the flexibility to contract different vendors for different components of the PACS,” Ratib says (see story “Deal Points”).

On the question of how much to archive for the new PACS, UCLA drew the line at the point in timeabout 1998, according to Ratibwhen all the data on its existing archive had become DICOM (Digital Imaging and Communications in Medicine) compliant. To have migrated images that did not comply with DICOM standards would have been pointless because, without conversion, they could not have been effectively used on the new PACS, Ratib says. The non-DICOM images, if needed, would have to be scanned from film and “go into DICOM as a secondary capture,” he adds. While this would require a librarian, it would be less expensive than attempting to convert all the old studies. Even with conversion, the non-DICOM images would not have been very manipulable on the new PACS, Ratib adds.

UCLA actually began migrating images from its old archive in February 2003. The PACS vendor had the responsibility for the migration, says Ratib: “It was a one paragraph clause in the contract.” The vendor assured the school the task could be completed in 3 months. Ratib says he knew better. “No matter what you do, migration is going to be your worst nightmare and your worst headache,” he says. He was right. It has been more than a year, and the migration is far from over. “If you ask me today, when we are going to be done, I can’t answer that question,” Ratib says.

HURDLE AFTER HURDLE

For all its difficulties, UCLA is making big strides with the migration. In March it was finally able to stop sending new radiological studies to the old PACS. Those studies are accessed with the new PACS now.

“The ASP is running. We are archiving everything on the off-site. But for a year we have been archiving on the two systems (the old PACS and the new), just in case. We have been drawing out some more extensions or implementing the work flow with the new RIS and PACS. It took us 1 year to really start deploying our new PACS. We’re not completely done yet, but we’re on the verge of stopping our old one now that the new one is working. However, our legacy images [those generated between 1998 and 2002 and still residing on the old PACS] have not migrated from the old to the new yet, at least not completely, and we’re still dealing with a lot of problems,” Ratib says.

A book could be written about the problems of the migration, but the major ones can be summed up:

  • The Timeline. While a 3-month migration was always judged unrealistic, Ratib says it has been impossible to establish a realistic timeline. “It’s important. How long are you going to have to maintain your old PACS? It’s costing us money. If you go back to the CEO and say the old PACS will have to be maintained for another 2 years, be prepared to be yelled at. We had stopped the old contracts. It’s not a question of if the old PACS will break down; it’s a question of when. You better have a timeline that you trust to begin with. Unfortunately, in our situation we’ve had to revise the timeline several times.”
  • Missing Studies. UCLA’s migration strategy was to use backup tapes, which had been kept on all studies, to do the migration while the primary studies remained on the old PACS for activeinterpretation. The tapes were fed over to the new archive through a data acquisition computer where software prepared them to be sent on to the data warehouse at the ASP vendor. They were sent in reverse chronological order, with the newer and therefore the most in demand studies migrating first. But, says Ratib, the backup tapes were not always pristine. Studies would bemissing or incomplete or in such poor condition that they could not be used. When that happened, the images for the new PACS users would have to be prefetched off the primary archives on the old PACS (in reality several PACSs that had been linked). “We had to do a lot of things to quality control the images,” Ratib says.
  • Mismatches. “The PACS doesn’t know how to send images if it doesn’t have an accession number, or requisition, procedure code, and so forth that are generated by the RIS. Otherwise, the images will stay orphaned,” Ratib says. With the old PACS, there was no control to make sure that images matched with the patient order on the RIS. “They were supposed to match, but they didn’t always.” When the old RIS data was migrated to the new RIS, which was to drive the new PACS, mismatches occurred frequently between the RIS orders and the migrated images on the new PACS.

Initially, Ratib and his colleagues tried to solve the problem by assigning a two-person team to rectify mismatches on a keyboard manually. The workers were overwhelmed quickly. “We had to go back and reengineer,” Ratib says.

An effort was made to encode rules in the transfer software that would profile the migrated data such as patient name, date, time, and modality, so that the RIS and the PACS files would be matched. “The profiling was the most difficult. You can’t think of everything and you discover everyday new problems,” Ratib says. “We put rules in there, and we found out that the rules were overdoing it.” He offers the example of missing accession or order numbers that would force an attempt to match the study with the patient’s name, date of the study, or modality. “OK, 90% of them go through, but 10% don’t go through. Why? The patient may have had two studies the same day. What if the time stamp is not correct on the PACS?” A study would be matched on the RIS only to find that a second study was already there. “So we had to go back and rewrite the rulesit’s very complicated,” Ratib says. “The problem is one of standardization. People are still accumulating data that may or may not migrate easily to the next level.”

  • Prefetching. A final headache was prefetching of prior images when a patient was scheduled for a new radiological examination. On the old PACS, the radiologists had access to the priors. They needed (and demanded) the same availability of priors on the new PACS. But often the priors had not been migrated yet. What to do?

After a dealing with a little resistance, Ratib says the vendor team and UCLA’s group of technologists worked together to create a “prefetch engine” that would scan incoming patient schedules and automatically fetch the needed priors off the active old PACS archive. The backup tapes were migrating the normal data run to the new archive, but the prefetch engine was programmed to archive needed priors at a higher priority that would supercede the routine archiving. “We used our old PACS for an ad hoc query when the patient was scheduled,” Ratib says.

But doing the ad hoc queries raised a new issue. If the images had already been archived ad hoc with a priority override, what would stop them from being archived doubly when they rolled off the normal tape run? “We had to make sure the program didn’t do that,” Ratib says, “but the difficulty is to determine if it is the same study sent twice or two different studies of the same patient done on the same day. We want to avoid storing duplicates without losing any studies by mistake.”

Ironically, a step UCLA had taken to move images between its old PACSs turned out to be super-serendipitous when it came to fetching images off the old PACS (see story “Open Sesame: The Web Fetch Portal”). UCLA’s IT people had created a “query and copy” portal that allowed images to be located at one source in the old PACS array and copied to another. The portal had actually been designed to make sure that images and other data could be moved only in compliance with patient-privacy regulations, but it was a familiar tool and it could be used to copy priority images from the old PACS to the new archive. “It saved us,” Ratib says.

Even so, Ratib still must pay for what he calls “an army of people” (actually four or five staffers) to manually intercede in the data migration in those instances where the computers and software are not up to the task. But this is slow and expensive, so the migration is continuing to be problematic. A lot of the difficulty continues to come from mismatches between the old RIS data and the old PACS data that have been migrated without the benefit of standardized codes. “It’s all about standardization,” Ratib says, “but I can’t give you a formula for handling it. There’s no magic bullet.”

THE PAYOFF

There is, of course, the bright side. As more and more images collect on the new PACS archive, the great advantages of the new PACS are becoming apparent. The new PACS connects five inpatient/outpatient buildings and the hospital on campus as well as two off-campus hospitals operated by UCLA. The archive can be accessed with clearance via the Internet. UCLA radiologists no longer have to scurry between buildings to read at workstations that were the only sites where some images could be placed on monitors. When it is completely deployed, Ratib sees the new PACS as becoming “generic for everybody,” whether they are in cardiology, nuclear medicine, or radiology. Instead of a collection of PACSs, UCLA will have only one.

Ratib says whether the new PACS will pay for itself through increased efficiency, saved film, and the like is “a very difficult question to answer.” But he says that work flow has been streamlined and patient care improved. He also thinks that faster report turnaround time will mean more income for the hospital. “Turnaround time for cash flow is one of our biggest initiatives,” he says.

Ratib admits some radiologists still complain that the old PACS had all the prior images, whereas the new PACS does not. “Some users see a step backI tell them, Sorry, we’re working on it.'” .

ADDITIONAL READING

Ratib O, Liu BJ, Kho HT, Tao W, Wang C, McCoy JM. Multigeneration data migration from legacy systems. Medical Imaging 2003PACS and Integrated Medical Information Systems: Design and Evaluation. SPIE Proceedings. 2003;5033:285-288.

Deal Points

One of the elements of installing a new PACS system is risk. For UCLA, part of the risk concerned hiring an ASP (application service provider) to handle the PACS archive. UCLA’s goal of separating the PACS vendor from the archive storage was based on the idea that the archive might operate for 15 years while the PACS would have a shorter lifetime. UCLA wanted to keep its archive intact with the freedom to change its PACS.

According to Osman Ratib, MD, PhD, FAHA, who is directing the hospital’s archive conversion to the ASP model, choosing to outsource the archive did involve trust that the ASP vendor would not only handle the storage but that it would be in business over the long term. That is one reason UCLA chose as its ASP vendor a subsidiary of its PACS vendor, but that created problems that had to be hammered out in lengthy negotiations, Ratib adds.

The PACS vendor wanted to bundle the ASP service into a single contract, just what UCLA sought to avoid. Moreover, UCLA wanted the ASP vendor to agree to what Ratib calls “a nonlinear cost model.” Basically, UCLA wanted a decrease in storage fees based on increasing volume. As volume went up, fees would go down. Similarly, it wanted a price decrease over time based on the idea that storage would continue to become cheaper for the ASP vendor as new technology developed. “Why should we pay the same price to store a study 3 years from now, when I know that the storage cost in 3 years will be a fraction of what it is now?” Ratib asks.

He says lawyers on both sides were nervous about the nonlinear cost model because it was essentially uncharted water for ASP archiving. “The negotiations took forever; it was tedious,” he recalls. “When it came to the legal people, it got crazy. They had all these fears about going too far in commitments and pricing, and what if 5 years down the road, it turned out they wrote a bad contract.”

In the big picture this was just one of the steps that UCLA had to take to lay the groundwork for its archive conversion and the installation of its new PACS.

OPEN SESAME: THE WEB FETCH PORTAL

By opening the application and filling in the query fields, medical personnel with the proper clearance at UCLA can sit at a computer and gather images, reports, and other patient-related data from a number of PACS deployments spread through several locations across the campus. The developers of the program call it an open source portal, and have given it the name WebFetch II. 1

WebFetch II is a software query and copy tool that can access data in a DICOM format from multiple sources and bring it to the user who needs it. WebFetch II, as its name implies, uses web technology and can fetch radiological imaging data from any of several existing archives at UCLA.

WebFetch II was originally designed to provide access to legacy archives in a way that ensured compliance with the Health Insurance Portability and Accountability Act, the federal patient privacy act. It was also meant to be a tool that allowed the user quick, easy access to images and other information that was distributed on several DICOM-compliant systems.

Now, UCLA is migrating this legacy imaging data to a new archive where it will be used with a new PACS that will span the whole enterprise, ending the need for disparate archives at different sites.

While WebFetch II was designed to manage image collection on the legacy archives, it has been an essential tool to quickly migrate urgently needed studies to the new archive that serves the new PACS.

According to Osman Ratib, MD, PhD, FAHA, the radiology information systems specialist who is heading UCLA’s migration of legacy images, WebFetch II allows a radiologist or a staff person to call images off the old system and place them on the new archive. It is used primarily to prefetch prior images for patients scheduled for new examinations.

While it was never designed for this purpose, it has, says Ratib, taken some of the edge off the complex task of transferring legacy data to a new archive.

George Wiley is a contributing writer for Decisions in Axis Imaging News.

References:

  1. Tao W, Ratib O, Kho H, et al. Open source portal to distributed image repositories. Medical Imaging 2004PACS and Imaging Informatics. SPIE Proceedings. 2004;5371.