Wally Peppler, PhD (left), and Gary J. Wendt, MD, MBA, BSEE, worked closely with the University of Wisconsin Hospital and Clinics’ information technology manager to spearhead the PACS initiative in the radiology department.

It is not necessary to spend $5-$10 million to implement an enterprise-wide online PACS.

That is the message being spread by the Informatics team at the University of Wisconsin Hospital and Clinics at Madison, the state capital.

The pioneering PACS created by the UW Radiology Department utilizes brand-name off-the-shelf commercial hardware and proprietary software, and deploys an economical but nevertheless state-of-the-art imaging management system that seems destined to change the shape of PACS to come.

UW’s multisite PACS employs open systems architecture to support (1) network connectivity between its 463-bed main campus and six area outpatient clinics, (2) interconnectivity with information systems at other hospitals, and (3) the ability to run on hospital-supplied workstations.

Starting with an ultrasound mini-PACS installed in 1998, UW soon moved boldly forward to expand it to all modalities. Its initial request for proposals (RFP) brought responses with prices ranging from $1 million to $10 million for an archive and diagnostic workstations. But while data storage and computer hardware costs have dropped by a factor of 10 in the past 2 years, it seemed to UW that PACS vendors were still trying to recover legacy system costs and were not passing down any of these savings. Buying PCs from a PACS vendor for $80,000 each did not seem to be the way to go.

It therefore opted for a software-only solution. It selected a vendor in May 2000, and licensed commercial software that now links 28 modalities with 20 hospital-supplied diagnostic workstations and a dual archive.

The three-tier archive offers 118 terabytes (TB) of deep storage at each of two online sites, which are located several blocks apart on the Madison campus. (A TB is 1,000 gigabytes [GB] ; a multi-slice CT examination can generate 1 GB of information.) This online redundant approach is considered to be one of the most realistic disaster recovery options available to PACS users.

The initial deployment of UW’s software-only PACS solution cost less than $1 million, according to Gary J. Wendt, MD, MBA, BSEE, a neuroradiologist and assistant professor of radiology. Now in his second year as vice chair of informatics for UW’s department of radiology, Wendt has worked closely with medical physicist Wally Peppler, PhD, and the hospital’s information technology manager to spearhead the PACS initiative.

The goal of the project was to establish a “contemporaneous radiology environment that could deliver radiology information faster and cheaper than the current system,” Wendt says. “We wanted a networked system that was patient-oriented and would provide radiologists in the hospital and outpatient centers with universal access to examinations and integrated work flow in real time.”

UW, a major teaching center specializing in organ transplants and oncology, conducts more than 250,000 radiological examinations annually. These generate 12-14 TB of data.

UW currently captures images from seven ultrasound units, five nuclear medicine cameras, four angiography systems, four MRIs, three CTs (one of them high-speed, multi-slice), and four computed radiography (CR) units. Images are read by a staff of 35 faculty radiologists. UW also has a film digitizer, which is used primarily to scan into the archive films from outside facilities that will be compared to new studies.

Software-Only Savings

Gary J. Wendt, MD, MBA, BSEE, vice chair of informatics for the University of Wisconsin Department of Radiolgy, sits on one of the standard, Windows NT, plug-and-play dual monitor PCs that serve as workstations.

“We looked for a software-only solution that would enable us to extend image management throughout the enterprise,”? Wendt says. “But we wanted a mainstream system. We didn’t want to be developing custom software solutions, or creating one-of-a-kind or custom hardware. We wanted to use commercial, proven, off-the-shelf hardware. If you have a software license, you can upgrade PCs as you see fit and do not have to go back to a PACS vendor and negotiate trade-ins.

“We are using commercial hardware previously validated by our software vendor-or validated specifically at our request for our system. We wanted to use hardware that our software vendor is comfortable with and used to supporting. We are not trying to cut corners buying economy workstations,” he adds.

“If people go after a software-only solution with the idea that they can slap it on any piece of hardware they see fit, they are going to end up seeing a lot of support problems in the future simply because, as new versions of the software roll out, you are not going to be absolutely certain that they have been validated and tested on that hardware.

“The bottom line is that PACS vendors should not be selling PCs. There are PC vendors out there that know how to do it. We can get a much better deal by purchasing hardware ourselves, and by getting a maintenance contract integrated with our purchases. The dual Pentium III workstations that we are putting in now may turn out to be slow 2 years from now, but they will still be great office PCs for the support staff. We can roll them over and continue using them, and put newer, higher-end equipment in and roll our software license onto the new hardware.”

The UW diagnostic workstations are all standard Windows NT PCs that you can buy on the Internet. They are ready to plug-and-play right out of the box. The workstation array includes dual-monitor gray-scale and? color systems, two of them in a Systems Development Laboratory that connect to a second non-production PACS used for testing and development. In addition, there are three single-monitor quality assurance stations and a single-monitor case review station.

“Typically, people use high-resolution (2,000 x 2,500) monochrome monitors for reading CR and digital radiography (DR), but for cross-sectional modalities 21-inch color monitors with a resolution of 1,600 x 1,200, similar to a display you might use with a desktop business computer is very acceptable,” reports Peppler. “What we are doing is creating clinical workstations using standard components. We are going to encourage our people to minimize the number of monitors they use-four being the maximum and two or three being the more common number.”

Radiologists also can access an electronic medical record using standard office PCs. These are deployed near the diagnostic workstations.

Eternal Archives

UW’s archiving solution is both creative and low cost. It is not only redundant, facilitating disaster recovery, but images are expected to be available online. The archive and its servers were installed in November 2000. The first tier of the three-tier archive is 3.6 TB of high-speed RAID (redundant array of inexpensive drives). The RAID incorporates more than 100 hard drives and is used for short-term 6-9 month storage. This is the fastest and most expensive storage medium in the archive. It cost UW just under $160,000.

When RAID starts nearing capacity, UW moves to second tier-medium speed Network Attached Storage (NAS) boxes that are less expensive than a Storage Area Network (SAN) or fiber channel RAID. NAS is basically a computer with multiple hard drives, a power cord, and a network jack, but no keyboard or monitor. “It just makes a big chunk of hard drive space available on the network,” says Peppler.

NAS can be added to the archive at any time, and UW plans to buy NAS boxes only as they are needed. “We are using a just-in-time approach,” Wendt explains. “The longer we wait, the cheaper the prices will be.” A commercial NAS box currently costs about $5,000 for 328 GB, but prices are dropping rapidly. “We recently bought 640 GB of hard drive space for $3,000,” he notes.

The two deep archives, where older data are stored, use generic Advanced Intelligence Tape (AIT), which Wendt and Peppler refer to as the “media of last resort.” AIT is the slowest (1 to 2 minute response time) and least expensive storage medium in the three-tier system. Each deep archive initially consists of 237 slots for tapes and a robotic arm that takes the tape out of the slot and puts it in one of six drives. Each archive in its current configuration can be expanded more than fourfold-to 1,182 slots and 24 tape drives-with a capacity of 118 TB. A second module of 276 slots costing about $6,000 will be added to each archive after July 2001.

The present AIT archives cost $52,000 each, but a full 118-TB AIT archive is in the $70,000 range, notes Wendt. “We actually had a quote of $489,000 from a PACS vendor for a 14-TB archive that was not expandable,” he adds. “Two of that vendor’s 14-TB archives would have cost us more than we have spent to deploy our entire enterprise PACS.”

Although there is a limit on how long most images must legally be maintained, UW currently has no plans to ever erase or even move its deep archive tapes into offline storage. “Eventually they may go onto a shelf, but we cannot foresee when that is going to happen,” says Wendt.

In using off-the-shelf hardware to assemble their system, Wendt and Peppler will not have the safety net of a system-wide warranty or service provider that those who buy from a single source enjoy. The system instead will be supported internally by a PACS manager, two full-time in-house service technicians, two part-time service support people, and two full-time radiology staff support technicians.

Warranty support for the computer hardware is provided by the individual vendors, and PACS software support will be provided by in-house as well as 24 x 7 PACS vendor support.

Accession Numbers

A basic problem plaguing every PACS installation is that of multiple accession numbers. While a single number for multiple studies would be desirable, the UW team does not believe it is feasible. “One accession number per order per examination is not a viable option,” Wendt notes. “Having a separate number for every possible combination of body parts in a CT scan would result in hundreds of possible orders.”

He sees three options:

1. Make multiple copies of the whole examination, so if you have a CT head, neck, chest, abdomen, and pelvis, you could copy the whole examination for every order and store it multiple times.

2. Have the technologist manually intervene and break up each examination before it gets into the PACS. The technologist, for example, would select the images associated with the CT head, and ship those off to the PACS. Then the next images associated with the CT neck would be selected and sent off as another examination. And so on, creating different examinations and different data sets in the PACS environment. But the problem with that approach is the radiologist who wants to read, for example, only the head and neck. He has to open and close two examinations rather than just one, which reduces his productivity.

3. Have one data set and assign an accession number for each series of images. As such, the PACS presents a single data set as five different orders to the physician. This approach, which eliminates the need to make multiple copies of the data, is similar to the one that IHE (Integrating the Health Care Enterprise cooperative program cosponsored by the Radiological Society of North America and Healthcare Information and Management Systems Society) is pursuing. “Our needs to associate multiple accession numbers with one radiology examination are here today and we need to solve them,” Wendt says. “The IHE initiative is really looking at a global picture. But in order to look at everything from the origin of the images to the final end product on the PACS, you need to get the modality vendors, the PACS vendors, and the HIS/RIS vendors all on the same page. And that is going to take some time. We are going to try to make sure that our solution is something that will meld in with the full IHE initiative. However, I think we probably will have to develop a custom, interim solution.”

Other Considerations

Getting radiological interpretations to referring physicians faster involves more than just PACS.

“We are pushing ourselves to try to improve turnaround time,” Wendt says. UW has a targeted goal to produce written reports in less than a day. Presently, physicians can access reports by telephone immediately after they have been dictated. While many hospitals have turned to VRT (Voice Recognition Technology) to immediately convert the spoken word to text on a computer monitor, UW is not yet ready to jump on the bandwagon. “We want radiologists to be doing radiological work,” Wendt notes. UW, he indicates, is working on getting more efficient transcription and signature of reports using the current system.

“An alternative to VRT may be to adopt a correctionist approach,” Wendt suggests. “Reports dictated by radiologists would be accessed by transcriptionists who would edit or correct them on a computer monitor almost immediately, then put them in the queue for electronic signature by the radiologist.”

While the UW Radiology Department has used teleradiology in the past to read examinations done at remote hospitals, it is not currently doing so. It now uses it only to send images to radiologists at home for on-call review.

Wendt and Peppler agree that creating a contemporaneous radiology environment and a perfect PACS is likely to be a never-ending story. “I don’t think it is ever going to get simpler,” Wendt believes.

Richard B. Elsberry is a contributing writer for Decisions in Axis Imaging News.