Diagnostic Imaging
November 2002

Viewpoint

Case grows stronger for software-only PACS strategy

Hospitals and practices find alternatives to expensive proprietary hardware and software systems

In the beginning, word processors were pieces of hardware. Office administrators bought an entire package from vendors like Wang and Digital. A "word processor" was the computer, the monitor, the storage device, and the keyboard, all optimized to work with the "built-in" software. Personal computers were only for tinkerers and experimenters. Today, of course, no one associates a word processor with hardware. It's just a piece of software you can purchase separately and run with a variety of different hardware choices.

Today's PACS are a lot like early word processors. Because the technology has been complex and difficult to manage, hospitals and imaging centers tend to purchase turnkey products from single vendors rather than attempting to integrate off-the-shelf hardware with PACS software applications. But the technology does exist for software-only PACS that, like word processors, will run on off-the-shelf hardware that could already be available in a hospital's existing server and network infrastructure. Institutions should no longer be required to purchase expensive proprietary hardware solutions. Unfortunately, in real life, most major PACS offerings are deeply embedded in legacy systems delivered through combined hardware/software installations. Software-only PACS will require a new breed of offering.

The fundamental building blocks of any PACS are image acquisition, cataloging, short-term storage, long-term archiving, data transmission, and image display. Some of these building blocks are already software-only, operating effectively on the newest off-the-shelf hardware.

The first requirement of any PACS is to acquire images from the various digital modalities and make them available to the institution's computer network. Usually, the images must be purged quickly from the modality to free up local storage and allow for the examination of more patients. The images are typically transferred immediately to a RAID (redundant array of inexpensive disks). This transfer uses software-only protocols (DICOM) that translate the data into a format the network and RAID can understand.

Cataloging, the process of remembering what is acquired and where it is stored, uses software-only, off-the-shelf database products such as Oracle, Sybase, or Microsoft SQL Server. If the PACS and radiology information system are separate products, as is common, this process first requires a synchronization step to interface the PACS to the RIS via HL7 or possibly an Intergrating the Healthcare Enterprise (IHE)-mediated direct database query. These are software-only solutions that can run on any standard hardware platform.

Short-term storage is necessary for the near-instantaneous retrieval of image data. Historically, storage-usually in the form of RAID-used proprietary hardware that customers had to purchase directly from PACS vendors. Many legacy PACS vendors still practice this antiquated approach, saying that it's necessary for hardware stability and adequate performance. What your PACS vendor may not want you to know is that you can now purchase your own RAID devices directly from hardware vendors in the form of directly attached or network attached storage (NAS) at very competitive prices (Table 1). When connected to your network, these devices provide cost-effective storage for any network application, including PACS.

Several vendors allow customers to purchase NAS RAID independently, an option critical to accommodating the inevitable growth of PACS storage requirements. A 1-TB RAID (enough to store about 40,000 examinations), for example, might cost hundreds of thousands of dollars from a PACS vendor, while the same RAID could be configured and purchased directly from a NAS vendor for $20,000 to $50,000.

Because the amount of RAID storage will directly affect PACS performance, the cost per terabyte is a good benchmark to use when comparison shopping for PACS. Even though RAID costs are falling dramatically, many proprietary PACS vendors maintain artificially high price tags.

A RAID is typically too expensive for long-term archiving of imaging data, so most PACS use some form of removable archiving media. This is cheaper but slower. When a user requests an examination that is no longer on the RAID, the PACS software retrieves it from the archive media, which can take several minutes (Table 2). To speed things up, most PACS software attempts to anticipate the next day's archive retrieval requests by monitoring the RIS's orders. The imaging data are transferred from the archive media to the faster RAID during the night, when the department is less busy. This semiautomated archive-retrieval process will become obsolete as prices fall and RAID becomes the dominant storage method for PACS. Removable media will be relegated to the storage of permanent records and possibly disaster recovery.

Communication and control of the archive devices, and possibly even the system RAID, are best performed by third-party off-the-shelf software packages known as HSM (hierarchical storage management). This area of PACS should not be controlled by your PACS vendor, who will not be able to keep up with ever-changing media types.

CONTROVERSIES

Although rapidly changing transmission technology is making software-only PACS more feasible, controversy has not been avoided. Conventional PACS use lossless compression strategies, which require high bandwidth. New products offer compression strategies that can compensate for slower wide area network transmission speeds. But controversy persists over how much an image can be compressed and still be of sufficient quality for interpretation, clinical review, and comparison as a prior.

The small body of literature that does exist suggests that wavelet compression at a 10:1 ratio has little effect on diagnostic acuity and is therefore seen by some, including the authors, to be acceptable for clinical review and comparison purposes in most instances. Others suggest that in many cases these compression levels are acceptable for primary interpretation as well. JPEG 2000 contains several standard forms of wavelet compression and is offered as a software-only solution from a variety of vendors, and it may soon be supported by Web browsers much as JPEG and TIFF compression formats are today.

As network prices fall and speeds continue to rise, these controversies may become moot. Most networks can already provide fast enough image retrieval to remote locations, and when JPEG 2000 is incorporated into DICOM, it could spell the end of these proprietary compression strategies from PACS vendors.

Finally, controversies surround image display for both software and hardware. Most Web-based display software is bolted onto a DICOM PACS as an afterthought, providing limited functionality. The newest Web solutions deliver all images through Web protocols over a browser but also incorporate sophisticated image manipulation tools. The advantages of the Web-based approach are simpler software upgrades, operating system independence, channel definition protocols, common user interfaces, and unending advances from improved browser technology. It won't be long before non-Web PACS, in any form, will be seen as an old and undesirable technology.

The American College of Radiology recommends a minimum monitor brightness for soft-copy interpretation of 50 foot lamberts (171 nits). Now that off-the-shelf flat-panel consumer displays are capable of pumping out over 200 foot lamberts (700 nits), it's not difficult for hospital personnel to purchase and deploy all the hardware necessary for primary interpretation and clinical review.

As consumer hardware technology has continued to advance rapidly, the vision of software-only PACS has taken shape. A hospital or imaging center will no longer need to buy hardware and software from a single vendor. Instead, hospitals will look for a solid software-only PACS solution from a mature software vendor offering associated services (such as managerial, operational planning, installation, training, support, and maintenance) to complement the institution's existing IT infrastructure. As many legacy vendors continue to sell high-priced, complex hardware/software solutions to a diminishing market, a new breed of PACS providers will emerge, delivering far more affordable software-only PACS with services that provide customers the ability to purchase competitively priced commodity hardware now and in the future.

Dr. Dreyer is vice chair of radiology, information systems, and Dr. Hirschorn is a clinical fellow in radiology informatics/MRI, both at Massachusetts General Hospital and Harvard Medical School in Boston. Mr. Schultz is a senior software developer for Partners Medical Imaging, assigned to Massachusetts General Hospital.