Let’s face it — PACS have changed everything about medical practice for the better. Picture archiving and communications systems save time and space, eliminate film and darkrooms, provide wider and faster availability of images, speed diagnoses and shorten hospital stays. What’s not to like? From departmentally focused systems to enterprise-wide behemoths, PACS are everywhere. And initially, the technology advanced so rapidly that the first revolution — the switch from proprietary to PC-based systems — is already behind us. Up next: Ushering PACS from glorified storage bin to clinical force majeure.

Just how this might play out was the main topic at the annual U.K. Radiological Congress in London last May. Their most important announcement revealed that the drivers of advances in radiology would be: Increased computing capability; the combination of anatomical and biological imaging techniques; and one-stop image acquisition.

It’s a forecast that’s tough to dispute, especially since innovation in these areas is already well underway. But like most predictions, it assumes the road will take a logical course. Like that would ever happen.

Let’s review. PACS progress was put on hold by the distraction of Y2K compliance, then by HIPAA and the wavering U.S. economy. Efforts to address serious performance deficiencies in speed, language conversion, and storage capacity were already lagging when the September terrorist attacks hiked the bar even higher on security and disaster-recovery standards. Other issues consequently slid further down the to-do list, trapping PACS R&D in a seemingly interminable game of catch-up.

And so there you have it, the state of multimodality imaging in a nutshell: glamorous drivers of advances cruising with post-apocalyptic slackers. Let’s take a closer look.

Growing pains

While technology vendors hustled to meet new data security mandates, the IT research firm Gartner, Inc. (Stamford, Conn.) published a wake-up report that concluded: “Approximately two billion radiographs are taken around the world each year, including chest X-rays, mammograms and CAT scans. It takes about 8MB to digitize a chest X-ray, and it has been estimated that to store all the world’s X-rays would require 17 petabytes each year.”

In lay terms, that’s nearly 17.5 trillion bytes. Fortunately, the report also examined a powerful new way to leverage PACS power: storage area networks (SANs). A SAN is a high-speed, dedicated network that links storage devices and servers. It creates a pool of storage that users can access directly, but which functions independently of the LAN. Transmissions are up to six times faster than with Ethernet. SANs also increase bandwidth availability and can be updated with software, rather than more hardware, making them a logical enhancement to HIS/RIS integration. SAN-based storage was used by 16.3 percent of the multi-user storage market in 2000; Gartner says 70.7 percent will use SANs by 2005.

Compaq Computer Corp. (Houston) is the largest supplier of SANs, one of which found its way into North Arundel Hospital in Glen Burnie, Md. The facility’s existing NT and UNIX storage and communications systems are now consolidated through a Compaq StorageWorks fiber channel SAN. It provides massive storage capacity, speeds up operations by 12 percent and cuts back-up time by 90 percent. North Arundel’s intensive care unit has since been named in a Solucient Leadership Institute (Evanston, Ill.) benchmark study as one of the top 100 I.C.U.s in the country.

North Arundel’s success validates Gartner’s conclusion that “PACS … are beginning to gain momentum beyond radiology departments.” In fact, it’s a momentum that’s barreling straight toward the dream PACS: the one that does it all.

Many hats, 1 central nervous system

“We got our start in the radiology domain,” Robert Cooke explains about the new generation of megaPACS from Agfa Corp. (Greenville, S.C.). Cooke, V.P. and general manager of global IMPAX systems, says, “Our server technology is capable of managing data from a variety of different image-generating departments within the healthcare enterprise, and deploying specific applications in each one of those departments for primary interpretation.

“IMPAX now has the capability of handling cardiology data, opthalmology, and essentially any modalities that produce DICOM objects. We’re capable of interfacing to non-DICOM [analog] modalities via technology known as Paxport. The system is highly configurable to articulate a variety of departmental workflows. Because the systems are so tightly integrated, a user in radiology can access cardiology data if it’s required, and vice versa.”

The systems also utilize speech recognition technology, notes Cooke, “so the reader can immediately dictate, edit, and sign a result on the same workstation.”

IMPAX runs on a Sun UNIX or Windows 2000 server, with Windows 2000 workstations. Agfa has 600 IMPAX installations at locations worldwide, including Brigham and Women’s Hospital and Massachusetts General Hospital, both in Boston. The larger installations average 20 to 30 primary interpretation workstations and have Web-based applications for the convenience of referring physicians.

Likewise, GE Medical Systems IT (GEMSIT of Milwaukee, Wis.) sells a combination RIS/PACS system called PathSpeed. It runs on Sun Solaris and Windows NT or 2000. “It’s very broad support for multimodality,” says Vishal Wanchoo, V.P. and general manager of radiology systems at GEMSIT. CT, MR, CR, DR, nuclear medicine, ultrasound, PET — whatever you’ve got, says Wanchoo, PathSpeed will support it. “We’ve got sites that are doing 150 connections. You can connect every modality that’s produced today, by any vendor. Typically in a department we’ll have about 30 to 50 modalities connected into a single PACS.” GEMSIT has some 300 systems installed worldwide that average 120,000 procedures per year.

Siemens Medical Solutions (Iselin, N.J.) offers a similar PACS called MagicStore that runs on NT or UNIX. It can be configured with a SAN and integrated with Siemens’ Sienet HIS system. The workstations run on Internet Explorer and support voice recognition technology.

Tomorrow’s headaches today

Numerous vendors now offer these one-PACS-runs-all systems, the magic of which resides in devices called gateways. They translate DICOM into language that can be understood by PCs, which is good. Where things get tricky is the “one-runs-all” part.

“There are many versions of DICOM. They don’t necessarily talk to each other very well in the real world,” says Piotr Slomka, Ph.D., F.C.C.P.M. Slomka is a medical physicist in the nuclear medicine department at the Robarts Research Institute Health Science Centre in London, Ontario, and assistant professor of diagnostic radiology and nuclear medicine in the University of Western Ontario medical biophysics department. He also heads a start-up company called Forward Imaging that develops fusion software.

“The DICOM systems are not necessarily plug-and-play solutions,” he says, “especially when it comes to things like PET and nuclear medicine, functional MRI and so forth.” In Slomka’s facility, for instance, the radiation therapy station doesn’t recognize PET DICOM. “You need to have some special customized solutions for that.”

Robert Taylor, Ph.D., executive V.P. of TeraRecon Inc. (San Mateo, Calif.), agrees that the all-purpose PACS is still “down the road,” for various other reasons. “There’s a big obstacle to that, which is getting the applications onto that PACS, and getting enough power into the PACS to do that processing.” Besides that, “fully real-time, interactive 3D registration is something not available today, and it’s really what people need.”

Sophisticated new multimodality applications such as image fusion, intensity modulated radiation therapy (IMRT), and image-guided surgery have only recently begun to be integrated into existing systems, with uneven results. The upshot is the flourishing of departmental or single modality PACS or miniPACS as they were once known. GEMSIT’s Wanchoo notes, “Some departments will deploy niche PACS — nuclear medicine PACS, or ultrasound PACS, or a PACS to support an angio department. Workflow inefficiencies based on such deployment are pretty tremendous. There are lots of things that you have to do in terms of integrating these departmental miniPACS into an enterprise PACS.”

Some do succeed, only to run into another problem: system support. More and more, CT and MR are being combined for IMRT planning, PET and CT for neurology oncology, SPECT and CT for breast cancer, and CT, ultrasound and funduscopic images for ophthalmic modeling. As Billy Crystal put it, “You look mah-velous!” But the question is, who fixes these grand synergies when they break?

“Typically, if a hospital purchased a system from Vendor A, and then they get a Vendor B, it’s not a given that this is going to work,” says Slomka. “And it’s not a given that Vendor A or Vendor B are going to support it, because Vendor A will say, ‘Oh, this is a Vendor B problem.’” Such disagreements have frustrated Slomka enough times that Forward Imaging is now in high gear, developing a high-performance, Java-based turnkey solution to interface multi-vendor devices and facilitate advanced operations such as 3D imaging.

Better, slowly

By and large, multimodality image blending is wide-open territory for developers. Special-purpose solutions are gradually surfacing, including non-vendor-specific software. 3D-Doctor, for example, is an add-on tool for registering and combining MRI, CT, and/or microscopy images from Able Software Corp. of Lexington, Mass.

PET and nuclear medicine, however, remain too highly specialized for off-the-shelf help. And although functional/anatomical image fusion has been in development for a while, until recently it was truly successful only for neuro. The reason, simply put, is that brains don’t commonly jiggle.

“For the rest of the body, you’ve got all these problems of organs moving around and patients in different positions, and fusion by software only in the rest of the body is very labor intensive and doesn’t always work,” says David Townsend, Ph.D., a physicist in the department of radiology at the University of Pittsburgh.

Townsend developed one of the first combination PET/CT scanners in partnership with CTI Inc. (Knoxville, Tenn.) and Siemens. “We started making images with cancer patients in 1998,” he says. “This triggered a lot of interest from the medical community, because PET has been notoriously difficult to read properly. CT gives you some anatomical framework to the functional image.” The first production version, the Biograph, was installed at the University of Pittsburgh last August (the second is at Memorial Sloan-Kettering Cancer Center in New York).

Townsend and his colleagues are able to put the new Biograph images onto the university’s enterprise PACS, although not elegantly. The PACS utilizes iSite streaming technology marketed by Stentor Inc. (South San Francisco, Calif.), a strong radiology management product also developed at the university.

“It’s all DICOM-compatible,” Townsend says. “But this PAC system — and I think most PAC systems — don’t read PET DICOM. So we actually convert it to CT or MR DICOM, and then the images can go into the PAC system. The problem is there’s nothing to look at them with, except up here in the PET facility. But the whole point about this is the images are intrinsically aligned, because they’re acquired on the same scanner. Then you don’t need a fusion algorithm. All you need is a way of looking at them.

“There’s a lot of work going on now, especially on the software side, both for driving the machines and viewing the images,” Townsend says.

The fusion software being developed by Slomka at Forward Imaging, for example, is targeted for thorax imaging. It achieves interdepartmental readability via a novel concept. Unlike other approaches that rely on the use of fiducial markers or invasive methods of landmarking for co-registration, Slomka’s utilizes PET and CT attenuation mapping. A gateway manufactured by Nuclear Diagnostics (Stockholm, Sweden) takes care of DICOM conversion.

“If you use the Web viewer and Java viewer,” Slomka says, “you’d be able to see these images in oncology without worrying about what data format they’re in.”

TeraRecon took a different tack with the same problem, stirring real-time 3D into the brew. Its VolumePro 1000 volume-rendering hardware performs CT/PET fusion and powers the company’s new AquariusNet 3D PACS.

“One of the things about fusion is it requires the kind of workstation level of power that’s usually not available where you need it,” says TeraRecon’s Taylor. 3D workstations in radiology departments, he says, “haven’t received a lot of attention because it’s too complicated and time consuming to get up, walk over to the PACS, sit down, put up the 3D review module, and get the benefit of those tools. So we’re blurring the line between a PACS station and a powerful workstation by making a one-size-fits-all product that delivers your 2D PACS review and your 3D, all seamlessly integrated into one server, serving multiple clients.”

Are we there yet?
Combining modalities has quickly become the wildcard in medical imaging. Yes, we have the technology. But the evolution of fusion far outpaces the sophistication of ancillary services necessary to support it.

A truly multifunctional PACS requires quite a lot more from a system than mere storage and display. Additional connections are needed for treatment planning devices in addition to radiosurgery, 3D and surgical navigational workstations. Also necessary is additional software to run those devices and stations, and for co-registering and combining multimodality images. In image-guided neurosurgery for cancer, for example, images may be obtained with CT, MRI, PET, fluoroscopy and ultrasound for just one case. They must be acquired, archived, depicted, modeled, calibrated and registered, and then applied in radiotherapy preplanning and/or surgical guidance. Still thinking about conjoining HIS, RIS and PACS? Sometimes, just imagining how total enterprise integration will actually be achieved induces brain freeze. For now, let’s just consider the OR of tomorrow.

Lasting investment?
So, OK, they’re versatile. They’re scalable. They’ve got a ways to go, maybe, but already they perform miracles. There’s even hope for adequate storage, speed and conversion capability. Now the only question left is this: Can multimodality PACS give good value?

“There’s a couple of things customers will typically do in growing a department,” says GEMSIT’s Wanchoo. “Either they’ll add more connections if they’re purchasing new devices, or change out the connections if they upgrade from one type of scanner to another scanner.” Storage also is a concern, he says, that is “typically built into a growth program. We factor that into the proposal for the system. That’s something the customers kind of have in their own minds: a five-year road map.”

Planning ahead is wise. The only problem with plans, however, is surprises.

Greg Karnaze, M.D., president of Austin Radiological Association (ARA in Austin, Texas), is installing an enterprise PACS. He says the teleradiology system he’s replacing “is about six years old and it’s really obsolete. We can’t get parts for it anymore, so we’re doing that first.”

Karnaze hopes his new one will last longer. ARA is comprised of nine hospitals, 14 outpatient centers and a tenant, the three-hospital Seton Healthcare System in Austin, which leases teleradiology services from ARA. All will be managed with ARA’s new Synapse PACS from Fujifilm Medical Systems U.S.A. Inc. (Stamford, Conn.).

“There are 52 radiologists in our practice. We do about 375,000 outpatient exams per year,” says Karnaze. “We have MRI, CT, ultrasound, nuclear medicine, PET, plain film — pretty much all modalities.”

Some 100 devices are being linked to ARA’s PACS. “It’s a brokerless PAC system, and so there is integration between the Fujifilm PACS and IDXrad v. 10.18 RIS [from IDX Systems Corp., Burlington, Vt.]. We have an OC-3 network among our 14 outpatient imaging centers. It’s an optical network. It’s 155 megabits per second, which is 100 times as fast as a T-1 line. From our archive we have an OC-48, which is 2.5 gigabits per second, which is hugely fast. We’re using EMC [Hopkinton, Mass.] equipment for storage. We’re starting off with something like 20 terabytes of storage and we can go up from there. We’re co-locating a disaster-recovery system, which we own. So we’re duplicating much of our storage and also having some servers and software loaded on them so that if, let’s say, a tornado or terrorist airplane should go into our computer center, within a matter of minutes we can be back up again without any loss of data or functionality.”

ARA’s new system is an enviable investment, microplanned to the last pixel. “State of the art” is likely the ideal term for it. But as terms go, it’s also one of the most ephemeral.

Closing the windows
Among many other things, PACS technology is a constantly evolving art. Just as the first generation of proprietary filmless systems was routed by the PC revolution, now PC-based installations face the latest challenger: Microsoft’s superfast, memory-hogging Windows XP Professional, the replacement for the Windows NT workstation and Windows 2000 Professional. Now in beta testing is the Windows NET Server, the XP replacement for Windows NT and 2000 servers. At least one vendor, Fujifilm, plans to convert its PACS installations to XP.

Though Windows NT is famously crash-prone, consider this before tossing it: XP won’t run on many older PCs; Microsoft won’t even allow upgrades from Windows 95. XP is not DOS-based, and requires scads of memory, a gigantic hard drive and a way-fast processor to be of minimum practical use. Replace NT with XP, and you may be replacing all the PCs in your network, too.

“The new operating system may not be able to work with old legacy software; it’s not very backwards compatible,” warns Gwendolyn Bubb, a Microsoft consultant based in Brooklyn, N.Y. “If you have to depend on adding old software which may not be able to work on the new OS, then you’re out of luck.”

XP also includes a can of worms called self-help, a programmed opening through which Microsoft can enter and manipulate your system at will. If you’re slow to register your new operating system after installing it, it shuts down and you’ll have to call Microsoft to restart it. Once installed, you can’t reconfigure your PC or swap out machines without first obtaining special codes from Microsoft — in short, your worst nightmare when a drive crashes in the OR.

The consumer version of XP, which could easily find its way onto radiologists’ home computers, has no encrypting capability for security (a HIPAA no-no), nor does it allow the user to return downloaded files from whence they came. (So much for signing off on studies in fuzzy slippers.) “The activation for the home user is a pain, if you don’t know what you’re doing,” Bubb adds.

The good news is, XP hardly ever freezes. But neither is it stacking up to be a panacea for healthcare. However, it soon may be your only option if Microsoft stops supporting its earlier products, the foundation of so many existing PACS installations — the ones whose owners were assured, as they signed those seven-figure checks not long ago, that they were eternal.