PACS - Next Generation
Review: Parikh, A. & Mehta, N. (2015) PACS: next generation. Proc. SPIE 9418, Medical Imaging 2015: PACS and Imaging Informatics: Next Generation and Innovations. 9148(94180G), 1-8. doi: 10.1117/12.2081987 (Note: I was later able to download this paper directly from the authors' web site: http://www.netdicom.net/Uploads/Documents/94180G.pdf).
The authors argue that the future of Picture Archiving and Communication Systems (PACS) is web-based because the expense of building and supporting traditional PACS may be financially unrealistic in many environments, while a web-based system could be more compelling due to potential for reduced costs and increased flexibility.
The authors' position is that web technologies have recently arrived at a point at which all the functionality of a traditional PACS can be replaced with web-based cloud infrastructure and "zero footprint viewers" that could run without installation, from a web page within a browser, securely without need of a dedicated VPN. They state that existing PACS technology is limited in that it cannot readily be used to push, for example, "send alerts to patient care providers" whereas a web-based platform could, using existing technology.
The authors note that existing PACS deployments are aging, and they assert that now is the time for institutions ready to replace their PACS to consider the possibilities of a new web-based architecture.
The article defines a traditional PACS as consisting broadly of three component capabilities:
- Image management and display, meaning the client functionality of viewing DICOM images
- Communications, such as query-retrieve functionality
- Administration, including filing of images, image storage, and access security.
Traditional PACS Limitations
The article outlines particular limitations of traditional PACS specific to each of these categories.
- Image display
- The authors note that PACS tend to lack streaming support for a lot of media, so files may have to be downloaded completely before any viewing is possible. Relatedly, many vendors require workstations specific to their modalities, adding to procurement and maintenance cost.
- PACS traditionally being constrained to a LAN means that radiologists may have a hard time sharing images with outside providers; many hurdles exist to providing access to PACS data while maintaining security.
- PACS systems may not have a fully- or accurately-implemented DICOM API, meaning that some queries could fail under certain circumstances; that is, one PACS' results could be inconsistent with another for the same query.
- The DICOM standard is domain-specific, so a lot of specialized knowledge is needed to develop and support PACS systems, leading to cost increase.
- Some traditional PACS do not use relational databases, so data organization can be challenging.
- Traditional PACS reliance on hardware architecture also leads to many costs; hardware maintenance and management is expensive, and acquisition can be out of reach for institutions in developing countries.
Future of PACS - Web-Based
A web-based PACS solution, according to the authors, is particularly well-suited to limiting access to medical images to authorized users, and to using RDBMS to handle image organization.
The HTML5 standard, only recently officially recommended, is crucial for displaying DICOM image data without lossy down-conversion, according to the article. Additionally, recent CSS3 standards are needed for certain display techniques that could be needed for prior image comparison. Thus, a web-based solution of the type they propose was impossible until recently. These standards along with cloud computing's current ubiquity makes implementing web-based PACS more attractive because hardware costs are abstracted away from the solution.
The article outlines and describes in considerable, sometimes technical, detail how a web-based solution could address each of the shortcomings of traditional PACS described earlier.
- Image management and display
- The use of AJAX techniques allows a web client programmer to specify that images be loaded partially, which can speed up loading times overall — that is, the first image in a series might load immediately and be available for the user while other images are still loading in the background, reducing the user's perception of system latency, and thereby eliminating the requirement of a low-latency LAN as a core component of a PACS. Advanced AJAX techniques taking advantage of features of HTML5 can also be used to load a lower resolution version of an image very quickly while in the background the full resolution version is loaded for display, reducing perceived latency even further.
- Streaming technologies mentioned above reduce image presentation latency significantly.
- Querying traditional PACS is generally limited to using DICOM DIMSE protocols for image retrieval, whereas web-based applications can use a number of options. The authors focus on REST APIs as the presumed mode of operation, noting their strengths as being able to completely replicate functionality of DIMSE protocols and to provide a robust authentication/authorization capability.
- Cloud computing is cited as beneficial for a number of reasons:
- The "pay as you go" model for storage and compute capacity offered by cloud providers provides a cost savings in that dedicated storage/compute need not be provisioned beforehand; fixed costs are therefore reduced.
- Data replication between a number of global cloud data centers could provide for reduction of latency for images that must be accessed in a number of locations around the globe.
- Cloud provisioning can provide built-in High Availability, data backups, and disaster recovery capabilities, reducing the IT workload for an institution.
- The authors cite the availability of at-rest and on-wire encryption as a reasonable substitute for physical custody of data and devices.
- They state that authentication models such as OAuth and two-factor authentication can be leveraged to provide a more distributed security model.
- The authors note that the new HL7 FHIR standard uses XML natively, and as such could be readily used for data interchange in the context of a web-based PACS built on RESTful services and web protocols.
- Additional Capabilities
The authors assert that a web-based PACS could also provide a number of services traditional PACS cannot.
- Notification - a web-based PACS could provide email notification based on any number of conditions, or when an item is shared.
- Hyperlinking to images in PACS from other applications
- Image sharing via emailed hyperlinks
- Server-side image processing - the authors provide the example of creating edge detection overlays for images in the PACS that could be displayed on the PACS client on demand.
- Image annotation - data added by radiologists could be stored as image metadata.
- Image anonymization - The authors assert that "a web-based system can offer two type of anonymization techniques" but no details are provided. The phrasing seems to indicate that a portion of text was left out of the published paper.
- Radiology report template matching based on image/study type could allow doctors to automatically load the appropriate interpretation report template for a given study based on the image metadata.
The authors conclude that new technologies will enable PACS to be deployed in a cloud environment, allowing developing countries to benefit from the associated cost savings, but they also predict that web-based cloud-enabled PACS deployments will generally supplant traditional PACS as well.
I couldn't understand the authors' argument for the change in security model. While it's clear that a different model would be needed for a web-based PACS system vs a traditional one, it's not clear from their description that an OAuth or two-factor system would adequately address an institution's security concerns; they didn't clarify how such a model would serve business requirements. A federated authentication or single signon scheme could be implemented for a PACS such as what they describe; however, this would not address any of the concerns an institution might have about how to administer sharing of image resources, which is one of the benefits that the authors are touting with this piece.
It was startling to encounter a plainly missing section of text; image anonymization is an important feature, but it is not described at all. That section of the paper ends with a colon, indicating that additional text was intended to be inserted but was not.
The authors assertions about traditional PACS not having a variety of features that would be available in the proposed web-based system strike me as odd. I am somewhat familiar, at least, with two of the major mainstream PACS systems, specifically Philips iSite/IntelliSpace and the GE PACS product. Both of these use RDBMS, and have for many years. They have a number of features for dynamic loading of images in their (web-based!) clients; they can annotate images at the workstation and save those annotations as image metadata; the Philips PACS can use its Standard URL-Based Integration (SUBI) protocol to link directly to images in the PACS via web services. Because of all these factors, it seems to me that some of the authors criticisms of "traditional PACS" are misplaced.
That being said, I agree with their conclusion; in fact, the current state of PACS such as Philips' offering goes some way to support it. The Philips PACS (as of version 3.6, which has been available for many years as of this writing in 2015) already has a web-based GUI, including DICOM image viewing; has a RDBMS backend; is capable of using federated authentication; streams images to the workstation; handles bit overlays similar to the edge detection case proposed by the authors.
I concur that it would be a major leap, though, for commercial offerings such as theirs to move to the cloud, and to using shareable URLs and authentication protocols such as OAuth. That being said, the authors of this article seem to regard the cloud as the inevitable final state of PACS; however, they do not seem to have a firm grasp on the technical bottlenecks that come with running a distributed PACS system. Even with a large amount of dedicated specialized hardware and high-speed LAN, a PACS can still be slow for a variety of reasons. I suspect that if the product these authors are offering gains traction in the market, they will be confronted with some technical headaches they had not foreseen, but that an experienced traditional PACS administrator would easily have predicted.
Commentary from Other Reviewers
- Could having PACS in a cloud environment will allow improved working conditions for the radiologists? A study that is done at night time here can be read by a day time radiologists across the globe in a timely manner.
- Some vendors' PACS, such as Fuji Synapse PACS, offer a mobile solution in addition to the web-based client application. This requires certain criteria for installation. Notably, functionality of the application is more robust in the client than the mobile solution. In order to make the best possible diagnosis, radiologists need an application with a full set of tools and functionality.
- Parikh, A. & Mehta, N. (2015) PACS: next generation. Proc. SPIE 9418, Medical Imaging 2015: PACS and Imaging Informatics: Next Generation and Innovations. 9148(94180G), 1-8. doi: 10.1117/12.2081987. Retrieved via ILLiad Interlibrary Loan.