Applied Radiology

Dr. DeJarnette is a Johns Hopkins University trained physicist who has been in the medical imaging industry since 1980. He is the author of numerous technical papers in the fields of high-energy physics, heavy ion physics, and PACS. Dr. DeJarnette is currently president of DeJarnette Research Systems, Inc., in Towson, MD.

T he earliest PACS systems were installed in the mid-to-late 1980s. These prototype systems were limited by computer technology, cost, rudimentary workflow concepts, and limited modality interface technology. In most instances they were little more than large-scale video acquisition and storage systems.

AT&T/Philips was the primary vendor for the first generation systems. Only a few dozen were ever installed, and few if any are in active use today. The U.S. military was heavily involved in the evaluation of the AT&T/Philips product offering (Commview) at Georgetown University Hospital in Washington DC.

The U.S. military learned from the Commview experience. In 1991, a request for proposals (RFP) was issued to the medical imaging industry to build a second generation PACS for deployment in Army and Airforce medical facilities worldwide. This RFP came to be known as the MDIS (Medical Diagnostic Information System). In September of 1992, the contract was awarded to a joint venture between Loral Aerospace and Siemens Gammasonics. MDIS was to be developed with a strong reliance on communication standards, particularly the ACR-NEMA V2, and later, digital imaging and communications in medicine (DICOM).

A large number of the resultant second generation PACS systems from various manufacturers have been installed. These systems were characterized by the utilization of ACR-NEMA V2 and early DICOM implementations for the acquisition of images from modalities; limited integration of hospital information systems (HIS) and radiology information systems (RIS); limited image viewing paradigms; limited network printing capabilities; and largely proprietary communications in all aspects of the system except for image acquisition. Companies such as Loral/Siemens (now GEMS), E-Systems (EMED division, now Access Radiology), Kodak (Vortech), Agfa, Cemax-Icon (now Kodak), Olicon (now ALI), 3M (now Kodak) and Siemens Medical Systems (Erlangen, Germany) offered the second generation systems; most are still in use today.

The industry is now at the beginning of the next generation--the third generation--of PACS solutions. Third generation solutions are characterized by a heavy reliance on interface and communication standards throughout the system, like the DICOM standard and the Health Level 7 (HL-7) standard.

DICOM is the direct descendant of the ACR-NEMA V2 standard. This standard was developed by industry under the auspices of a joint venture between the American College of
Radiology (ACR) and the National Equipment Manufacturers Association (NEMA). DICOM is in fact the ACR-NEMA V3 standard. The name was changed in 1993 to DICOM in order to disassociate it from any one standards body and to gain widespread international support.

HL-7 is purely medical informatics meant to standardize the transfer of information between the HIS products of different manufacturers. In order to optimize the workflow efficiency of an institution making use of PACS, the PACS must be integrated with the hospital information system and the radiology information system.

Some near-third generation solutions, outgrowths of second generation systems, are currently in the marketplace. These systems are limited by their second-generation design concepts and architectures, which reduce performance. The U.S. Military and the Veterans Administration, two early adopters of PACS technology, are again helping to drive PACS concepts and design. The U.S. Military has recently awarded a new PACS contract vehicle, known as DIN-PACS, to replace the earlier MDIS contract. This contract is for the deployment of third generation PACS throughout the military; it calls for heavy usage of communication and application standards such as DICOM and HL-7.

What have we learned?

There have been numerous lessons learned from the deployment of first and second generation systems. These lessons account for all components in a PACS.

Architectural concept --All first and nearly all second generation PACS systems have been characterized as "central storage" systems, as opposed to "distributed storage" systems. Central storage systems have the advantages of reduced hardware costs and easier maintenance. However, such systems have the disadvantages of poor scalability and single point failure. Performance of central storage systems is also relatively poorer, as the architecture requires that all acquired images move to a central storage device before being distributed to display stations. This deficiency can be made up for by utilizing higher bandwidths and higher cost communications technology.

The cost advantage of a central storage system is being eroded. As the cost of storage components comes down, there is less reason to use a central storage system. This was a significant advantage in the past, but no longer. Such systems, designed for a single radiology department, are less well suited to the distributed environment radiology will require in the future. Moving images to a central site for storage increases the likelihood of catastrophic disk failure and increases communication costs.

The MDIS systems deployed in the early-to-mid 1990s are perfect examples of central storage architecture. These systems made use of image acquisition gateways to acquire images from modalities. These images had to be forwarded on to the central storage facility (WSU--working storage unit), where they would be made known to the system database and would be stored on a redundant array of independent disks (RAID), backed up by an optical jukebox. This method provides higher performance with less chance of catastrophic disk failure. The database co-resided with the image storage on a single computer. This central storage facility would serve as a central server for the PACS workstations, all connected in a point-to-point fashion over proprietary high-speed fiber optic connections.

The image retrieval performance of this type of system was far and away the best in its day and remains one of the highest performing systems available. However, this architecture did have a number of weaknesses. The co-location of the image database and image storage made the central archive a single point of failure. The tight integration of the database and the storage system made upgrading to new storage technology difficult, expensive, and of high impact from a clinical perspective. Upgrading these early systems required that the entire PACS facility be brought down for periods of up to a few weeks. It is really this issue that, arguably, most distinguishes central storage architectures from distributed storages. Distributed storage architectures completely de-couple the image storage components from the image database (management) components; image database components are potentially redundant and self-replicating. The independent storage components are suitably distributable to minimize image movement within the PACS.

Long-term archival --Distributed storage systems can minimize communications bandwidth requirements, reduce the likelihood of catastrophic failure, have a minimal impact on clinical operations and, when properly designed, can make upgrade to new storage technology less expensive.

As anyone who has lived through a PACS deployment will tell you, you can be certain that your archive will grow over time. Storage technology is a rapidly advancing science. The storage system you purchase today will be functional but technically obsolete within a few years. Therefore, your PACS system must allow for easy migration to future storage technologies, or you will find yourself re-purchasing your PACS sooner than you thought. This is the single greatest advantage of physically separating your image management (database) and image storage components.

Even today, when a vendor attempts to sell you a DICOM archive, this single device is designed to include short-term storage, long-term storage, and the database to manage these images. In a true distributed storage architecture, these components are de-coupled, creating individual components known as "short-term store," "image manager (the database)," and "long-term archive." The long-term archive is the storage component whose upgrade must be the most carefully planned. It is this upgrade (or technology migration of this component) which was not considered early on by the designers of PACS.

The difficulty in migrating to new storage technology, especially long-term storage technology, arises out of interfacing any new storage technology to the existing PACS. The electronic and software interfaces of today's technology and tomorrow's technology may be very different. In a tightly coupled PACS archive component, integration of a new long term storage device (at this "hardware level") generally will require a significant amount of software rewriting, re-integration, and perhaps even hardware modifications. These tasks can make such an upgrade a very expensive proposition for the vendor, which means that the customer will most certainly incur some significant costs.

In a distributed storage architecture, the long-term archive stands as a separate entity. Its interface, is a "system level" interface, not a "hardware level" interface. These system level interfaces are known as "hierarchical storage management" (HSM) systems (they have also been referred to as application storage management [ASM] systems). HSMs are designed to make the long-term archive look like a network disk, from a software engineering perspective. Typically, these systems are designed to work with a hardware configuration consisting of a small, high speed disk subsystem, a database, and the long-term storage hardware. The HSM manages the migration of stored objects (files) between the high speed disk subsystem and long-term storage. The HSM also presents a simple "file handle" interface (a type of file naming system) to the rest of the world. A distributed storage architecture PACS makes use of this file handle interface. A DICOM long-term archive is built by layering DICOM storage and DICOM query/move services on top of the HSM.

The following section serves as a simple DICOM dictionary, fundamental to any thorough discussion of PACS and their components.

SOP: service operator pair --A SOP is a pair of software entities: one provides the ability to make requests for a particular service, the other provides the service. In DICOM, one entity is known as an SCU (service class user) and the other as an SCP (service class provider).

Storage class SCU --This is an entity which makes a request of another entity (storage class SCP) that an image be stored. This component is the image sender; the storage class SCP is the receiver.

Query class SCU --This is an entity which makes a query request of a query class SCP. The query in question is for a list of images with certain attributes. Typically, a query class SCU would be an imaging workstation and the query class SCP would be an image database that would know where the images are stored.

Move class SCU --This is an entity which makes a request of a move class SCP to move a set of images to another location. The move class SCP is generally also a storage class SCU entity, such that it receives a request to move a set of images to another location and does so by acting as a storage class SCU. The images are moved to a storage class SCP, which could be the same entity which made the initial query. If a storage class SCU makes a request to move images to some entity other than itself, it is referred to as a "third party move."

Query/retrieve provider --This entity acts as a query class SCP and a move class SCP. Such an entity has the ability to respond to queries for lists of images with specific attributes and then to move those images to another location, based on a request from a move class SCU.

The distinction between a PACS archive and a long-term archive may seem a trivial and semantic distinction, though this is not actually the case. It is true that one could build a PACS archive (which includes the long-term storage, short-term storage, and the image management database) which makes use of HSM technology. The result would be the two databases--a PACS image management database and the HSM database--co-residing. From an engineering perspective, this is problematic in terms of performance, machine loading, and other technicalities; it is also more costly. Inevitably, the vendor engineers choose to have one database which serves both the image management and storage migration (HSM) functions. It is in making this choice that we (vendors and customers) are backed into a corner, making migration to future storage technology expensive and difficult.

Simple migration to future storage hardware requires that the current archive (storage system) not have a great deal of application knowledge resident. Combining the two databases into one mixes an application knowledge database (image management database) with what could be an application knowledge-free database (HSM database). There would now be no clean high level interface to new storage technology. This interface is embedded in the single database, which increases the engineering work to be done if one migrates to new storage technology.

The real difference between what has been termed a PACS archive and a long-term archive is the level of application knowledge which is resident on these devices. A PACS archive has knowledge of the storage application it is performing, whereas a long-term archive as defined in this article has no application-specific knowledge. A long-term archive stores neutral objects. These objects may be images, but the archival system is not aware of this; it is the image manager (database) that has this knowledge. This division of application knowledge makes migration to future long-term storage technology easier from an engineering standpoint and less obtrusive from the clinical standpoint. This type of long-term archive is known as a neutral object application habitat (NOAH).

HIS/RIS interface

It is generally recognized that a PACS without an RIS makes little sense. One of the major benefits of a PACS, workflow improvement, cannot be recognized without the integration of the PACS with the radiology information system. Enterprise-wide benefits are attained with the integration of the PACS/RIS and an HIS.

In first generation PACS, such integration was rare; integration was more common with second generation systems but was still highly limited. Today, it is not uncommon to see an RIS and a PACS integrated in a unidirectional fashion, where information flows from the RIS to the PACS but not back. Bi-directional PACS/RIS integration is far less common.

Generally, an RIS will pass ADT (admission, discharge, transfer) and order information to the PACS. It is also desirable that reports be passed from the RIS to the PACS and that study status and preliminary report information be sent back to the RIS. Today, second generation PACS do this by making use of multiple databases. On a basic level there would be a PACS image database and a RIS database sharing (duplicating) information. In the majority of second generation systems, however, three databases are required to achieve this integration: an RIS database, a PACS image database, and a middleware engine database, such as a PACS broker--an entity which acts as an interface between the HIS/RIS systems and a PACS. It enables the transfer of patient information, radiology department orders, and ADT information to the PACS. All of these databases maintain a common subset of data.

Multiple databases in a PACS/RIS system are not cause for concern. However, the two or three databases required for PACS/RIS integration in second generation systems is highly undesirable, as the maintenance involved in the sharing of a common subset of data is problematic. Such databases can become unsynchronized, and they invariably do. This increases the maintenance burden and can lead to clinical confusion. Therefore, it is ideal to have the PACS image management database and the RIS database be one in the same. It is important not to view PACS and RIS as independent systems or products; PACS is a value-added addition to a RIS and a RIS is a value-added addition to a PACS.

Some late second generation PACS/
teleradiology systems have made use of what has come to be known as "Web technology." These early Web systems have proven useful in many cases. What is perceived as Web technology has come to be associated with the Internet, Web browsers, and low cost applications. However, this is not a useful definition. In actuality, it is a software technology which allows for the widespread distribution of information and standardizes the interaction between information systems. Web technology does not imply usage of the Internet, and does not, in and of itself, imply lower cost. Web technology can be used over the Internet, over intranets, or over other communication infrastructures which support TCP/IP.

The Internet, however, has proven an excellent infrastructure for teleradiology, where cost is a driving concern. There has been some usage of pure Web technology in this application. Low cost, general purpose Web browsers have been used to view radiological images. These browsers, however, are suboptimal for this task. To improve the viewing capabilities of these browsers, "add ons" (applets) have been developed to remove some of the limitations. For example, the ability to window/level cross sectional images requires an add on. In general, teleradiology viewing applications based on Web technology are poorer in terms of functionality and performance than dedicated teleradiology viewing products.

It is generally thought that the usage of Web technology has brought the cost of at home teleradiology systems down. In the case of equal functionality and performance, the development cost of a radiological image viewing enabled Web browser is the same as the development cost of a dedicated non-Web based teleradiology viewer. The widespread availability and usage of the Internet, TCP/IP and other standards, as well as an increasing market is what drives costs lower in this application, not the usage of Web based software technology.

The remote HIS/RIS data entry and review application and the external distribution of radiology reports are true web applications, which are gaining acceptance. The amount of information moved in these applications is far less than is moved in a teleradiology application. The information moved, patient data, reports, and scheduling information is better suited to the development of specialized application applets.

Both the internet and pure Web technology will find increasing usage in the next generation of PACS.

Modality interface

First generation modality interfaces were achieved primarily through video acquisition techniques, whereas second generation interfaces made use of ACR-NEMA V2, proprietary digital, and early DICOM storage interfaces. Such interfaces allowed for the exportation of image information into the PACS. The "double entry" problem, that of entering one set of patient information into the RIS and an incorrect variant of that same information into the modality, was not significantly addressed by first or second generation systems. Some second generation modality interface gateways ("protocol converters") made use of DICOM modality worklist interfaces to rationalize the RIS and modality patient information. However, such interfaces were the exceptions to the rule.

Modality image acquisition, as a technical problem in PACS, has largely been solved by the widespread implementation of the DICOM standard and the availability of image "protocol converters," used to interface pre-DICOM modalities. The modality interface problem facing third generation PACS is that of truly solving the "double entry" problem and improving workflow. Although most, if not all, modalities manufactured today support the DICOM standard, they support only the DICOM storage SOP portion of the standard. A smaller number of modalities support DICOM worklist management SOP or DICOM detached study management SOP, which are required to solve the "double entry" problem; even fewer modalities manufactured today support the DICOM modality performed procedure step SOP, which is used to track the status of image acquisition workflow.

The next generation modality interfaces ("protocol converters") will have to address these informatics problems. Such devices will have to provide for the acquisition of images, the rationalization of the patient information with the RIS patient information, and the reporting of modality workflow status. These interfaces will provide a complete DICOM interface to the modality. In the case of some modalities, these interfaces will be supplied as a part of the modality itself.

The best examples of such modality interfaces can be found in the world of computed radiography. These interfaces present a modality worklist to the operator, associate records from the worklist with the digitized images, provide image quality control, transmit the digitized images to multiple destinations, print images to both local and network printers, and report modality workflow status.

Display station

The most visible PACS component is the display station. In first and early second generation systems, many vendors and customers alike thought that the most important component in the PACS was the display station, though we now realize that there is not really any single most important component.

First generation PACS displays grew out of experience with multimodality display technologies. These systems were largely suitable for viewing cross sectional modalities. With the advent of second generation PACS, more thought was given to the viewing of radiographic images. The optimal display and viewing paradigms differ dramatically for cross sectional images and radiographic images.

Early in the deployment of PACS, the debate raged as to how many monitors were needed to make a useful display station. That debate has largely been settled, with most vendors and customers feeling that for diagnostic purposes, a dual head, high brightness, 2 K * 2.5 K display station is suitable. Four-head systems, common in first and second generation PACS, are no longer thought to be necessary (and certainly not cost effective) by the majority of PACS customers. Some researchers argue that even lower cost display stations are all that is needed for diagnostic purposes, though this is yet to be proven. ¹

Today, PACS display stations are broken into three types, based on functionality: 1) primary diagnostic displays; 2) clinical review displays; and 3) low cost teleradiology displays. Viewing paradigms differ among these three types of displays. All first and most second generation display stations were generally designed with viewing paradigms not optimized for primary diagnostic reading. Such systems served the roles of clinical review and teleradiology display much better than that of primary viewing. These systems are generally characterized by large areas of the display monitor being taken up by distracting logos, icons, menus and the like. In fact, researchers have done eye motion studies on such display stations and determined that a radiologist spent up to 25% of their reading time not looking at the images, but rather searching for menus, controls, and interacting with these controls. ^2,3

Results of the above mentioned research pointed the way to a new generation of display stations. ⁴ The new display stations are characterized by a "dark and stark" appearance with few visible controls and distracting non-image information on the display monitor. Previously, viewing controls consisted of drop down menus and icons, each one needed to indicate that a specific operation was to be performed. In such systems it was not uncommon to have to work through three or more menus or selections to perform some simple function. The new generation of display stations has put the most common controls on a three-button mouse, making the mouse buttons context sensitive to the state of the display. This new control paradigm, termed "heads-up imaging," allows the viewer to spend more time concentrating on the images and less time looking for the controls.

Conclusion

The next generation of picture archiving and communication systems will differ dramatically from what we have seen in their first and second generation. The new systems will make use of more sophisticated modality interfaces, which address not just the issues of acquiring images, but the consistent entry of patient information and the improvement of departmental workflow. In the future, PACS will be much better integrated with RIS and HIS. Additionally, the new systems will be the first with product offerings that include a full-featured RIS at a reasonable price.

In any sophisticated technical product or technology, it is not until the arrival of 3rd or 4th generation products that the early promise of that product or technology is realized. From here on, PACS will bring significantly improved radiology department workflow through improved display station viewing paradigms and the tighter integration of HIS/RIS and PACS. The new generation will be characterized by higher performance at a lower cost, lower cost maintenance, and lower cost technology upgrade and migration. Finally, PACS will see a widespread adoption, both internally in radiology and in the wider medical enterprise, by hospitals and the external medical community. AR

References

1. Stewart BK, Langer SG, Taira RK: DICOM image integration into an electronic medical record using thin viewing clients. Medical Imaging 1998. PACS Design and Evaluation. Proceedings of SPIE 3339: 322-328, 1998.

2. Krupinski, EA, Lund PJ: Comparison of film vs monitor viewing of CR films using eye position recording, pp 269-279. Proceedings from SCAR, 1996.

3. Horii, SC, Grevera G, Feingold E, et al: Prototype controls for a plain radiography workstation. Medical Imaging 1998: PACS Design and Evaluation. Proceedings of SPIE 3339: 87-91, 1998.

4. Stockham CD, DeJarnette WT, Levy VH, et al: Implementation and experiences with a minimal GUI diagnostic display station, Proceedings of SPIE, Medical Imaging, 1999.