Five Levels of PACS Modularity:
Integrating 3D and Other Advanced Visualization Tools
 
Authors:
Kenneth C. Wang, MD, PhD, Johns Hopkins Hospital; Ross W. Filice, MD; Eliot Siegel, MD; James F. Philbin, PhD; Paul Nagy, PhD
 
Background:
The current array of PACS products and 3D visualization vendors presents a wide range of options for utilizing 3D tools in a clinical workflow, and the emergence of server-based rendering in 3D imaging applications expands the set of possible system architectures. At the same time, these technologies present opportunities for optimizing performance and improving the usefulness of advanced visualization methods. We present a classification scheme describing the integration of 3D tools into PACS systems, with the goals of providing: (1) a practical guide to utilizing 3D tools with a primary PACS, (2) a set of metrics for evaluating integration architectures, and (3) a vision of possible future architectures, to facilitate more widespread use of 3D technologies in routine practice.
 
Evaluation:
Until recently, the predominant mechanism for utilizing 3D visualization tools was based on a dedicated workstation, separate from the primary clinical viewing system, comprising a combined hardware and software solution designed to provide turnkey 3D capabilities. More recently, PACS vendors have increased native 3D capabilities within clinical viewers, and a degree of integration with third-party 3D vendors has also become possible. These options involve tradeoffs between performance, maintainability, and cost, and each of these factors should be considered with metrics such as: frame rate, speed of access to prior exams, client thinness, storage efficiency, modularity, initial cost, and longer-term cost. These may be applied to a series of five architectural models for incorporating 3D tools into a primary PACS, representing five progressively advanced levels of integration. These levels and metrics are delineated in Table 1 and discussed below.

Table 1

Level 1: Separate turn-key clients

Level 1 consists of a dedicated advanced imaging workstation, separate from the primary viewing application. The dedicated workstation is typically based on a thick local application, using locally-cached image data, which may be retrieved from the main imaging archive, or directly received from a modality. The dedicated workstation maintains its own local storage, and access of non-cached studies incurs the network penalty of full image data transfer. Once received, imaging performance is limited only by the local hardware, with no further network dependence. Lack of communication between the PACS system and the third party application interferes with comparison of image displays across multiple applications.

Level 2: Proprietary client side integration

In Level 2, the PACS vendor provides internal advanced visualization tools within its primary viewing application, and/or partners with a third party vendor to provide proprietary client-side integration. At this level of the classification scheme, both the primary PACS client and the advanced visualization tool are thick clients. Image data, while still transferred in full to the client machine, is typically shared by memory-to-memory transfer on the client machine, enabling the radiologist to relatively easily transfer a study from the PACS to the advanced visualization tool. The redundancy of dedicated hardware, local image storage, and duplicated network transfers of Level 1 are eliminated here, reducing costs and improving maintainability and performance. However, users are limited by the internal options or custom-integrated external tools provided by the PACS vendor.

Level 3: Thin 3D client separate from PACS, with context sharing and multiple archives

In Level 3, clients become thinner with the introduction of server-based image processing, such as the server-sided rendering, which is becoming increasingly important among 3D applications. This improves access to exams, as the full network transfer of image data required in Levels 1 and 2 is now traded for the transfer of server-computed renderings, although a new requirement for sustained low network latency is introduced in order to maintain interactivity of the display. Thinner clients also improve maintainability, as client hardware and software requirements are simplified. Standards-based context sharing on the client, such as via the HL7-CCOW standard, allows for potentially improved modularity when compared to Level 2. However, with the introduction of a server-based component to the advanced visualization tool, there is now a requirement for an archive that is capable of interfacing with this component outside of the primary PACS. In the absence of a general-purpose archive, this implies at least partial archive duplication, reducing storage efficiency and increasing costs.

Level 4: Archive integration, separate clients

Looking beyond Levels 1 through 3, in Level 4, the archive requirements of multiple server-based viewing applications are addressed through a single unified archive. This unified archive is used to serve the primary PACS, the server-based rendering engine of the 3D application, and any other server-based application components, eliminating the need for duplicate storage. This implies greater modularity, greater storage efficiency, and decreased complexity on the server, when compared to Level 3. Level 4 does require a standardized archive application programming interface (either open or proprietary). Clients remain largely separate in Level 4, each with its own dedicated user interface and special-purpose link to the server.

Level 5: Complete integration and modularity of archive and clients

Level 5 represents a vision of complete modularization of both client-side and server-side components of image viewing systems. An application programming interface is provided, both at the archive level on the server, as well as at the viewing level on the client workstation. Level 4 leaves client integration unaddressed, whereas multiple viewer applications in Level 5 share a common interface and communicate with the server through a standardized client API. This offers the potential for unified access to 3D and other advanced imaging tools alongside more routine display techniques.
 
Discussion:
There is a need for better interoperability between imaging systems, given the complex demands of radiologists, as well as the increasing number of tools available for advanced image viewing and analysis. The five integration architectures in this classification system represent a spectrum of increasing client thinness and system modularity, impacting the key metrics of performance, maintainability and cost. Advancing toward Level 5 leads to greater efficiencies and less dependence on customized linkages. The work of DICOM Working Group 23 and the National Cancer Institute’s eXtensible Imaging Project (XIP),[1] lay much of the needed groundwork. In this time of increasing PACS market penetration and commoditization, PACS and other imaging vendors have an opportunity to further develop and commercialize the key interfaces which will enable a future of modularized visualization applications. This would improve the usefulness of advanced visualization methods in routine clinical practice, would enable best-of-breed technology suites, and would potentially spark innovation in the marketplace for PACS and visualization technologies, much like the iPhone platform has driven recent innovation in mobile computing and telephony.
 
Conclusion:
We propose a five-level hierarchy for classifying and evaluating integration architectures, with a view toward a future of modularized visualization systems.
 
References:
[1] Prior FW, Erickson BJ, Tarbox L. “Open Source Software Projects of the caBIG In Vivo Imaging Workspace Software Special Interest Group.” J Digit Imaging. 2007;20:94-100.