Radiology Takes to the Cloud

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon
cover.jpg - Radiology Cloud

Radiology appears to have reached a tipping point in its adoption of cloud computing, with economic and availability issues sending many applications to the cloud.

From retail and manufacturing to the government and public sectors, the cloud computing movement has left its mark on a wide variety of industries, with heavy-hitters like Apple and Amazon touting it as the Holy Grail in business and life alike. While the healthcare segment as a whole remains a bit slower to adopt a cloud-based model, radiology is finding multiple rationales for moving forward on this front and has begun to embrace cloud computing for a range of clinical and operational applications.

Cloud and cloud computing applications may vary from industry to industry, but the definition of the terms themselves is straightforward. “The cloud,” as it is known, is a cadre of remotely-located technology tools, connected by the Web. As for “cloud computing,” the American College of Radiology (ACR) follows a definition set by the National Institute of Standards and Technology (NIST).

ACR executive vice president and CIO Mike Tilkin. According to NIST, “cloud computing is a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider action.”1

More specifically, notes Rasu B. Shrestha, MD, MBA, chief innovation officer, University of Pittsburgh Medical Center (UPMC) and president, UPMC Technology Development Center, cloud computing comes in several flavors, including software-as-a-service, wherein specific software applications are run through a cloud; platform-as-a-service, in which users leverage a suite of virtual applications, programming languages, and tools); and infrastructure-as-a-service (reliance on remote data storage networks). The menu of cloud types varies as well, encompassing public (used by multiple entities, with services available through such entities as Amazon) private (dedicated to a particular organization), hybrid (a combination of public and private clouds), and community (shared by a number of organizations).

The Case For Adoption

Whatever its category, there exists strong impetus for bringing the cloud and cloud computing to the radiology table. Under a cloud–radiology umbrella, images are no longer stored on local IT infrastructure, nor are they shared through such physical means as compact discs. Rather, image storage and transmission occur via an off-site network of computer servers, facilitating file backup and allowing access to data through multiple devices, without concern about data loss should an individual device be compromised.

Economics ranks among the top motivators for pushing imaging into the cloud. Transitioning to the cloud eliminates significant capital expenditures associated with in-house IT infrastructures—ie, large (servers) and small (routers and switches) hardware components as well as the cost of data center operation and maintenance, points out Jon Copeland, former CIO, Inland Imaging, a tech-savvy practice based in Spokane, Wash, and the fourth largest private radiology practice in the U.S., according to Radiology Business Journal’s 2014 ranking.

Copeland also is CEO of a full-service eHealth system development and integration company with offices in Spokane, Seattle and St. Louis, Nuvodia, a subsidiary of Integra. In addition to reselling one vendor’s PACS, the company implements that system for healthcare providers throughout its service network and the referring community, utilizing its eRadiology secure, private cloud platform to host and maintain the system database at vendor-operated data centers around the U.S.

All customer sites have on-premise image storage; unlimited web-deployable enterprise and diagnostic workstations yield full access to all cloud technology features. Additional data storage initiatives are under development, including archival storage on a public cloud and an interactive third-tier storage solution.

“Healthcare is under extreme economic pressure, and the fact that cloud takes IT from a capital expenditure to an operating expenditure adds greatly to its appeal,” Copeland says. Hospital systems, the executive adds, would much rather spend their capital on integrating physician practices and acquiring other healthcare systems than on operating data centers.

Shrestha corroborates Copeland’s comments, adding that although cloud technologies in imaging have the potential to decrease capital expenditures on hardware, there can also be a savings on the software and services front, particularly if deployment occurs on a wider scale. Cloud deployment in general necessitates a low initial investment, and although additional IT investment is needed as system use increases, costs can drop if usage decreases.

“The final product here is a closer match between cash flows and total system cost,” Shrestha asserts. “Subscription and pay-per-usage models are also financially attractive.”

Multiple efficiencies

Along with other gains, some financial savings can be derived from related, equally compelling advantages of, and justifications for, migrating to cloud-based radiology applications: simplified, more logical data storage; rapid data retrieval; efficient image-sharing, and a streamlined workflow. According to Shreshtha, cloud computing is a better means of grappling with the large imaging data sets, complex algorithms, pre- and post-processing requirements, and increasingly distributed environment with which radiology must now grapple.

The distributed data grids inherent in cloud architecture perform with greater efficiency than do hard drives, in large part because of memory-sharing between data centers. This, Shrestha says, means faster access to data.

“By allowing us to establish storage hierarchies, cloud offers the opportunity to better organize (data) retention schemes and use a tiered storage hierarchy that makes sense from a standpoint of what should be retained and for how long—six months, two years, seven years, etc.,” observes Richard “Skip” Kennedy, MSC., CIIP, technical director, imaging informatics, Kaiser Permanente Northern California. Kaiser Permanente Northern California currently is piloting cloud-based storage for PACS for visible light (VL) imaging; other cloud-storage tests centering on PACS are on the drawing board.

Kennedy says he can easily envision that over the next two years, many PACS implementations will effect a transition wherein primary back-end server architecture lies in the cloud. “We will always have some internal storage, but is there cloud in our future? Absolutely.”

For its part, UPMC utilizes a private-cloud platform to execute image exchange between 18 sites as well as among outside institutions and has realized a mid-double-digit decrease in enterprise-wide imaging study management expenditures. “Accelerated image exchange and heightened workflow efficiencies have a lot to do with it,” Shrestha states. “Private cloud computing allows us to take advantage of speed and processing power, pushing intelligent pre- and post-processing of imaging data that are directly integrated into clinical workflows.”

“Just as importantly,” he continues, “when images are in the cloud instead of on a CD, we can go beyond accessing them at the point of care, efficiently doing our jobs from wherever we happen to be—armed with complete imaging data sets, priors, critical clinical information and the correct set of diagnostic tools.”

Patient-care benefits

Then, there is the benefit of cloud-driven patient care enhancements. Indiana University Health (IU Health), which has transitioned from CD-based to Internet image transfer using a proprietary platform known as IU Health Radiology Cloud, ranks among institutions that are reaping those benefits.

The main referring hospital group in Indianapolis and all of Indiana, IU Health provides Level 1 Trauma, Level 1 Stroke, Level 1 Vascular, and several other acute services. As such, it employs IU Health Radiology Cloud platform as a conduit for transferring imaging exam data generated and uploaded from the PACS at 86 area hospitals (83 in Indiana, two in Wisconsin, and one in Illinois) to its own radiology PACS. An additional 15 hospitals are expected to start availing themselves of the IU cloud platform in 2015, surpassing an original target of 100 hospital users.

“What is very positive is that three or maybe four of the other biggest hospital groups are becoming hubs, meaning we will see more of a grid rather than just an IUH Radiology Cloud and are perhaps heading toward an ‘Indiana Cloud,’” Jonas Rydberg, MD, professor of clinical radiology at Indiana University School of Medicine and Chief of Radiology at IU Health Methodist Hospital, observes. He adds that some 800 to 1,000 imaging studies are sent to the healthcare provider each week, with a total of about 45,000 exams received from January through December of 2014.

Two parallel databases reside on IU Health’s radiology PACS. One repository, known as the “IU Health Database,” houses data from 10 different IU hospitals, including Methodist Hospital, University Hospital and Riley Hospital for Children in downtown Indianapolis; two other hospitals in the network maintain their own databases. Imaging exam data from hospitals outside the network, along with exams conducted at local freestanding imaging centers, physician offices, and other non-IUH facilities as well as those downloaded from CDs, is stored on the second repository, called the “Outside Exams Database.”

The databases are linked, with two components of the PACS allowing access to the contents of both. One component puts clinical information for a particular study—specifically, priors, study information, reports, notes, documents, and series—at clinicians’ fingertips. The other displays studies for all patients with multiple identifiers within the same database and can use MRNs or MPIs to find exact matches. It also displays near-matches by first and last name, gender, and date of birth.

Rydberg says the anytime, anywhere access to current and prior studies afforded by the cloud platform improves patient care in part by reducing the need for re-scans; IU Health’s re-scan rate for acutely transferred patients has plummeted from 10% to 15% to a mere 1% to 2%  since it boarded the cloud radiology train. Patients consequently receive lower doses of radiation, which, Rydberg asserts, is a plus for all individuals but of extreme importance in pediatric imaging.

Speedy transfer

Additional patient care enhancements stem from the speed and efficiency of cloud-based data transfer. Treatment decisions, Rydberg states, can be made sooner than ever before—in certain situations, before patients arrive from the referring hospital.

“We can prepare accordingly and not risk complications or loss of life in the ER from delays on arrival because how-to-proceed hasn’t been determined,” Rydberg said in a prior interview with Radinformatics.com2, a sister publication Radiology Business Journal. “We can better determine when it is not necessary to transfer a patient—something ER physicians in small hospitals may not be able to do. Clinicians treating outpatients referred to us by other hospitals can review images in advance of appointments. All of this has a bearing on the caliber of care—plus, it is a great timesaver all around.”

Rydberg also points to ancillary benefits afforded by IU Health’s move into the cloud. For example, the hospital system saves about $38 each time an exam is sent to its destination through the cloud rather than via FedEx ® overnight courier.

Not surprisingly, the ACR, too, is promulgating the cloud as an enabler of better patient care. “It is our intent to facilitate connectivity with and among our members to obtain and push data [into clinicians’ hands] in the name of increased patient care, while making the flow of that data as seamless as possible,” Tilken explains. “A cloud infrastructure,” by virtue of the fact that it enables real-time connectivity to real-time data in clinical repositories, “is key.”

The ACR maintains a collection of data registries under the umbrella of the National Clinical Radiology Database; the latter was recently designated a Clinical Data Repository by the Centers For Medicare & Medicaid Services (CMS), permitting radiologists use it to submit data for the Physician Quality Reporting System (PQRS) program. The society is working on establishing cloud-based access to the registries.

Also underway is a collaborative effort among the ACR, Mass General Hospital and a vendor of speech recognition technology to deliver clinical decision support to radiologists as they interpret studies. Data used to provide this support will be derived from a variety of sources, including, but not limited to, white papers and actionable findings generated by ACR committees, and will reside in central cloud-based repositories.

“The idea is to offer clinical guidance in whatever form, in real-time—again in the name of better patient care,” Tilken states. “We envision that down the road, clinicians will, as they dictate reports, be naturally directed to a set of clinical guidelines that pertain to findings they cite—for example, a particular lesion or a condition—in a cloud-based repository.”

On FHIR

Other catalysts will push the cloud adoption envelope going forward, specifically, the introduction of the Fast Health Interoperable Resources (FHIR) interoperability standard, created by HL-7. FHIR uses application programming interfaces (APIs) that call for specific elements, rather than the entire message as required of other HL-7 standards.

These APIs leverage a Representational State Transfer (RESTful) architecture that allows resources to be exchanged on an as-needed basis; communication typically occurs via the widely used Hypertext Transfer Protocol (HTTP) or Hypertext Transfer Protocol Secure (HTTPS) protocols for moving files from point to point on the Web. A FHIR implementation standard is currently in the draft standard for trial use (DSTU) phase and has yet to receive full approval; a second version of the DSTU is slated for release in July of 2015.

The RESTful architecture makes it easy to transfer resources between systems because they can be accessed through a uniform resource locator, or URL, Shrestha explains, thus rendering FHIR a good fit for cloud-based platforms. In turn, the enhanced, more manageable interoperability, medical device integration and implementation of customized workflows afforded by FHIR will spark heightened interest in cloud computing among radiology constituents, Shrestha contends, adding that many vendors and institutions are already using  Web services in programming.

Developments related to the Digital Imaging and Communications in Medicine (DICOM) standard, published by the Medical Imaging & Technology Alliance (MITA), should also play a role here. In May, MITA released an updated version of the standard that includes a machine-readable extensible markup language (XML) representation; this machine-readability, according to MITA, opens doors for the use of appropriate standards to connect healthcare systems and, consequently, to permit data to flow and be understood between different systems. Additionally, the new addition incorporates a family of DICOMweb services intended to facilitate wider image sharing and enhance accessibility for clinical interpretation.

“The ability to make use of standards and to be more agile in developing client applications,” as supported by DICOMweb, “will unquestionably drive more activity in the cloud as it relates to radiology,” Kennedy asserts.

Questions and Challenges

Benefits notwithstanding, cloud platforms and computing present their share of issues and challenges, and data security tops the list. In order to minimize the potential for data breaches, sources say, hospital systems have taken such precautions as engaging server vendors to conduct annual security audits (and perform remediation), as well as arranging for monitoring by PACS vendors and others.

Other steps encompass encrypting data during storage and transfer, storing data in virtual “compartments” within storage media for an added layer of protection, connecting with servers using secure URLs only (HTTPS is an example) and restricting access to data via biometric markers, such as fingerprints.

Whether to opt for a private or public cloud platform—or some manner of hybrid—also remains a burning question. Copeland counts himself among those who believe there is much to be said for cloud services delivered by Amazon and its ilk because of the airtight security other clients have dictated that they provide. Amazon also operates a private cloud for the Central Intelligence Agency, Copeland notes.

Amazon’s public cloud is compliant with many federal and third-party assurance frameworks, including ISO27001, ISO9001, PCI Security Standards Council, Service Organization Controls 1, 2 and 3, HIPAA, the Federal Risk and Authorization Management Program and, most recently, the Department of Defense Cloud Security Model, Levels 1 and 2.

Others are more comfortable going private. There is, sources say, significant work being done by vendors and other entities to ensure the security of data stored in private clouds. Moreover, some purport, opting for a proprietary (private) cloud platform increases the likelihood of being able to maximize storage flexibility to suit individual institutions’ needs.

Still, there is no question that the forecast for radiology is cloudy. Security-related snafus, and likely other twists and turns, may await. However, the platform seems firmly in place.

Julie Ritzer Ross is a contributing writer for Radiology Business Journal.

References

  1. Mell, P., Grance, T. The NIST Definition of Cloud Computing. NIST web site. http://bit.ly./NISTCloudComputingDef.  Accessed December 19, 2014
  2. Ross JR. IU Health: achieving data ubiquity with a little help from the cloud. Radinformatics. April 12, 2014.  http://www.imagingbiz.com/portals/radinformatics/iu-health-achieving-dat.... Accessed December