Applied Radiology

Dr. Nagy is the Director of the Radiology Informatics Laboratory at the Medical College of Wisconsin, Milwaukee, WI.

What does it take today to have a successful enterprise-wide picture archiving and communication system (PACS) encompassing the entire institution? While many radiology departments are enjoying the fruits of a filmless workplace, few institutions have been able to shut down their film processor operations to the rest of the hospital, thereby achieving the true cost effectiveness of PACS. All PACS vendors claim that they have the right stuff to wean the hospital from film, but do they really? And do you and your institution have the right stuff to implement the change?

Finding the right recipe for PACS can be a daunting task. The job is one part technology assessment, one part change management, one part vendor management, and at least three parts understanding your own environment. At the Medical College of Wisconsin, we have distilled the essence of a successful PACS implementation down to four critical components: (1) staffing, (2) infrastructure, (3) integration, and (4) provision of a powerful common user interface across the enterprise.

It takes a team

PACS is an information system, not an imaging modality. Hospitals that treat PACS as a modality will suffer from inadequate support, training, and from a lack of understanding of how implementing PACS will affect their workflow. A PACS creates major changes in how images are managed. Vendors cannot restructure your organization to accommodate these changes; the restructuring has to be done by a competent internal team. This team must understand the technical, business, and user aspects of the changes brought about by PACS. Many of the often-overlooked project requirements are business related, such as how to sell PACS to the enterprise and how to get upper management to buy in to the PACS vision.

A real technical challenge for the PACS team is in understanding the concept of PACS as a mission-critical system. Most other major information systems within the hospital have a proxy service available in the event of downtime. For example, a physician can always call the lab to get patient results if the information system is down. But in the case of PACS, there is no way to describe how an image looks over the phone to a surgeon. To instill a sense of security for the PACS users at the University of Pittsburgh, a "drop what you are doing," 90-second on-site response time was implemented for all critical areas such as the emergency room, intensive care unit, and radiology department. Dr. Paul Chang, ¹ director of informatics, stated at the PACS 2002 conference in San Antonio, "Our response team members wear white coats and act as PACS consultants. They are more than just help-desk support. They are part of the solution and need to understand how and where PACS is being used."

To hire a PACS team you might need to think outside the box. Do not confine your manpower search merely within the medical environment; advertise in places where you can tap into qualified information technology (IT) professionals. Acquiring talented IT personnel and teaching them DICOM and PACS can be easier than trying to find a qualified PACS expert. Look for individuals with good project management, communication, and systems administration competencies.

A powerful user interface for the enterprise

Supporting proprietary workstations throughout the hospital is a challenge for any PACS team. Not only does geography slow down the response time, but maintaining workstations and providing periodic application and system upgrades is also a real resource drain for the team. Using a Web-based, or thin-client, distribution mechanism alleviates many of the problems. A web-client software interface can be distributed automatically to any workstation within the hospital, as long as the receiving computer meets the technical requirements.

The Web is a great distribution mechanism, but the user interface has to be just as powerful, intuitive, and responsive as the clinical and diagnostic workstations. Also, many physicians need advanced navigation and visualization tools, not just the radiologists. A trend among PACS vendors is to provide a common user interface for radiologists and referring physicians. This approach simplifies training and support, and keeps referring physicians from feeling like second-class citizens.

In implementing an enterprise PACS, one challenge is to predict how many workstations will be needed. A good exercise is to count the number of light boxes in the facility. Dr. Steve Horii (personal communication, 1/2002) performed this experiment while he was at the New York University Medical Center. Before he gave up counting, he exceeded 1,000 light boxes in just the first three floors of the 18-story building. To believe that a few dozen dedicated workstations will replace all those light boxes is just wishful thinking.

Integration

PACS is not an island unto itself. The buzz at the 2001 RSNA meeting was about PACS and the integration with the radiology information system (RIS). This reflects the realization that PACS is not about moving images from point A to point B, but about improving the productivity and efficiency of the department and the institution. Combining a PACS and a RIS can be accomplished through a proprietary interface, but will be more flexible for future changes if it is done with an open architecture and industry standards. The open standard initiative leading this integration is the Integrating the Healthcare Enterprise (IHE) movement (www.rsna.org/ihe). PACS/RIS integration should be part of any request for PACS proposal. The IHE focuses on the integration of RIS and PACS and reinforces core industry standards such as HL7 and DICOM. "Radiology is leading the charge in workflow integration for the entire healthcare with IHE" stated Jay Gaeta (personal communication, 3/2002), a technical and planning committee member of the IHE.

Manually synchronizing databases between the hospital information system (HIS), RIS, and PACS is an error-prone and time-consuming task. IHE outlines how to reconcile patient information corrections among the three databases automatically. Skip Kennedy ² presented a paper at SPIE this year that collected issue statistics for the year 2000. He determined that >80% of the radiology workflow problems encountered at the University of California­Davis were solvable under the framework of the IHE.

Another trend is HIS/PACS integration. Physicians who just want the report and a few key images shouldn't necessarily need to learn PACS, but should be able to get the images as a transparent layer through the HIS. Additionally, HIS/PACS integration saves a lot of user-account duplication and eliminates the need to train the entire hospital to use the PACS software. Having the entire patient record at one's fingertips has been the goal for many years; it is now eminently possible.

Infrastructure

What makes PACS possible is that it is no longer an engineering novelty. PACS relies on the computer networking, storage, processing, and display industries. These major industries have caught up with the technology requirements for PACS, which is now becoming a coat-tail industry. Radiology can now enjoy the cost savings of off-the-shelf commercial hardware.

This is important because performance is important for everyone, not just the radiologist. Two important new trends are to put everything on an online redundant array of inexpensive disks (RAID), and to upgrade the backbone of the hospital network to gigabit Ethernet. Physicians are not known for their patience, and forcing them to stare at an hourglass while the system fetches the images from a deep archive will not win many converts.

PACS is such a storage-demanding application that it was actually driving the storage industry between 1988 and 1993. ³ There are now several advantages to putting all images online and keeping them there. First, this option has become very cost effective, with storage capacity doubling at no differential cost every 14 months. This growth rate even exceeds Moore's law (Gordon Moore, co-founder of Intel, stated in 1965 that the number of transistors on a microchip will double every 18 months, effectively doubling the processing power of computers. The ability to move that data on the network and store that data on a hard drive has also matched and even exceeded Moore's prediction for the past 35 years.) As an example, large hospitals generate approximately 10 terabytes of uncompressed data per year. That amount of data storage would have cost $10 million for hard drives alone 8 years ago. Today the cost is under $0.5 million.

Second, online storage offers instant access to the entire archive. Third, system administration and architectural complexity is reduced by getting rid of complicated prefetching, or purging and fetching on the RAID back and forth to the deep archive, which slows the RAID's primary function of getting images quickly to the desktop. Lastly, migration from one physical media format to another is avoided.

Conclusion

PACS is going mainstream as a cost-effective film alternative. This is due to the highly accelerated pace of the computer industry as well as the fact that the PACS industry has had 14 years to develop standards and stable solutions that can now solve the problems of image management for radiology and throughout the enterprise. The challenge is really upon the IT and radio-logy leadership of the hospital to understand and incorporate this technology into their environment. AR