In 1993, radiologists working at the Baltimore Veterans Affairs Medical Center (VAMC) did something revolutionary. They moved into a new hospital facility that had no permanent film file room. Why? Because the Baltimore VAMC had purchased a PACS, the very first full-hospital PACS to be deployed.
After 18 years of nonstop use, it's time to replace the PACS, and the radiology department is preparing to go shopping. Dr. Eliot Siegel, chief of imaging services at the VA Maryland Health Care System and the radiologist who orchestrated that first PACS purchase, described the features he'd like to see in a new PACS capable of lasting another 18 years to attendees at the New York Medical Imaging Informatics Symposium held in Manhattan last week.
Siegel's requirements are tempered by reality but also look ahead to the future. He presumes the PACS will have features that include high performance, reliability, safety, and security. He also wants an infrastructure and architecture designed to accommodate new, as-yet unknown technological innovations that will materialize over the next 15 to 20 years. And the next PACS must be affordable!
Hassle-free, interchangeable archiving
Having a user-transparent, scalable, hassle-free, truly vendor-neutral archive is at the top of Siegel's wish list. He explained that when the Baltimore VAMC spent $800,000 for the best long-term DICOM image archive money could buy in 1993, it acquired a 1-TB Kodak optical jukebox. (Last week, the cost to purchase 2 TB of data for home personal use cost him a whopping $65, he noted wryly.)
"By applying heavy image data compression, the jukebox capacity lasted for four years," Siegel said. "Then we switched over to [a Plasmon archive], but our PACS vendor was never able to migrate the data in the jukebox to this new system. Then Plasmon went out of business, and we switched to our current EMC content-addressable storage with spinning disk technology. Data migration was still a hassle. It seemed slower to migrate images than it was to acquire them."
— Dr. Eliot Siegel, VA Maryland Health Care System
"So although we've been using the same PACS for 18 years, we are still on version 1.x of the software," he said. "Every time we asked our vendor to upgrade us to the next generation of software, our vendor would say, 'But you still have those old Kodak jukeboxes that are still running.' This created huge issues for us. We've spent thousands of hours performing archive-related activities."
Siegel wants a universally accessible, scalable, intelligent system that can store images agnostic to the hardware and that will enable anyone to use archival storage as a commodity. He would like a single archive with automated backup capability that can store any type of image -- DICOM and non-DICOM -- even if the archive doesn't understand the image format. The archive needs to be able to acquire data not only from all types of PACS and subspecialty PACS, but also from advanced visualization systems. Specifically, the archive should store studies for specialized PACS in their proprietary formats, so that the PACS perceives the archive as being native to their systems.
It should be able to store raw data from all modalities, such as thin-section data from CT, tomosynthesis modalities, and potentially sinogram CT. The goal is to enable data to be reconstructed differently, if so desired.
A replacement storage system needs to be able to retain annotations. "We've been annotating for years, and I want to be able to preserve these annotations, both for newly acquired images and for ones that have been migrated," Siegel explained. "Surprisingly, most of the vendor-neutral archives sold today don't have this ability. But annotations are really important to highlight important findings and document quantitative measurements, especially when looking for change over time."
Going green, and more-intuitive interfaces
Siegel and colleagues have conducted a significant amount of research on optimizing reading room conditions, and they've put what they learned into use. To reduce electricity consumption, with respect to both powering workstations and cooling the rooms in which they reside, Siegel would like to see workstations and monitors that can turn themselves off when they aren't in use.
He'd also like to see highly efficient, more-ergonomic user interfaces that are designed to work with large datasets. The mouse that was designed decades ago for word processing and spreadsheets can cause repetitive stress disorders. Newer products, such as a jog shuttle and a programmable cylinder substitute for a mouse, have been proved to increase the efficiency of radiologists reviewing large CT datasets.
Siegel showed examples of multitouch tablets, handheld guidance devices, thought-reading headsets, and electronic gloves, any of which may someday morph into use in a diagnostic workstation. Radiology could learn a lot from the gaming industry in terms of developing intuitive user interfaces that can be customized to task or preference, with improved navigation and user feedback, and that are fun to operate, he said.
Workstations should be able to run on a variety of operating systems or use multiple operating systems. PACS architecture should support server-side rendering, which would eliminate the cost of workstations that need to be equipped with a lot of power or memory. A supercomputer in the PACS would enable a workstation to interact with it and receive just-in-time downloads of information. From a consumer perspective, this would be comparable to requesting a specific street view of a location on Google Maps.
A new PACS should enable multiple software programs and applications to run on a single host workstation and different programs to transparently exchange data.
"Right now, I have to walk from one workstation to another to do my job," Siegel said. "This is not efficient. I'd also like to be able to use best-of-breed software, and swap out an app when I wish to do so with a better one."
Intelligent data sharing
PACS software and related healthcare IT applications should be smart enough to share data where the information needs to be shared. Applications also should be able to extrapolate data from a measurement or annotation, and be able to automatically populate this into standardized fields that can be data mined.
Data should be accessible from any location and with any device, whether locally or in the cloud, by any authorized user. Robust security for this activity is presumed. With universal image access, heavy reliance on DICOM CDs will end. Siegel believes that CDs are an interim solution that will be phased out, much like film digitizers have been.
Having access to a patient's clinical data, intelligently presented, with hyperlinks, is a futuristic feature that will confer huge benefits to radiologists. "There are major opportunities to healthcare IT and PACS vendors for improvement in delivering contextual clinical information to the point of diagnosis," he said. "I am convinced that 20 years after the introduction of electronic medical records [EMRs] and PACS, the clinical history that I get is really no better than it was two decades ago."
 |
| Dr. Eliot Siegel believes that in the future radiologists could have access to workstations like Watson, an IBM supercomputer that could provide radiologists with real-time feedback on clinical analytics. Image courtesy of Dr. Eliot Siegel. |
"I am lucky enough to have residents to review a patient's EMR and call her doctors if necessary, so that I can be more effective with respect to image interpretation," he said. "But there should be analytic systems that collect this information. A smart system could automatically look at a patient's chart and summarize the information for me to facilitate interpretation of a specific exam. This information would be different for a head CT compared to a chest or abdomen study."
"If a summarized clinical history containing information that I defined as relevant was displayed in conjunction with the study I was interpreting, it could impact how I interpreted the study," he said, and provided several examples of clinical information that could change findings in a report, if known.
Vendors are developing business analytic software functionality that focuses on billing and productivity rather than clinical summaries, he noted.
"Analytics that look at report turnaround time, patient dose, relative value unit productivity, study volume, and physician referral patterns are really important, but what we don't have yet are analytics to analyze a patient's medical record, and help us as radiologists make a decision," he said.
Siegel would like to see vendors develop clinical hanging protocols that include a personalized list of what radiologists would like to know for a specific type of exam, and how they would like it presented on a monitor. This could include lab studies, prior surgeries, patient medications, previous recommendations, abnormal findings from a physical exam, genetic information, and a photo of the patient.
Another capability Siegel would like to see is the integration of one-click request, just-in-time knowledge delivery. This would include the ability of the workstation to provide instant access to comparable cases in the PACS database or in external image libraries and relevant peer-review articles, for example.
The diagnostic workstations on his new PACS should also be equipped with basic advanced visualization functionality: multiplanar reconstruction (MPR), maximum intensity projection (MIP), volumetric rendering, arithmetic summation of thin images, and the ability to combine thin slices into thicker ones for viewing, he said. He also would like future workstation software to contain intelligent computer-aided detection (CAD) systems that search for and label anatomy and look for pathology.
Use of structured reporting templates, such as those developed by the RSNA, will be beneficial. They allow a radiologist to see critical features of what should be reported, which, when entered into a standardized field, can be data mined. Siegel predicted that in the future there will be better templates to facilitate interaction and improve patient care. One example he suggested was software that measures the size of nodules over time.
He also wants the PACS to be able to support near-future innovations, such as the next generation of contrast injectors. These will be able to send information that allows a radiologist to correlate between the injection curve and images to determine why images are of good or bad quality. PACS and contrast injectors should be able to communicate with each other to identify previous injections and to help radiologists optimize injection protocols.
Another application on Siegel's wish list is software that follows up on the outcome of recommendations made to referring physicians. He'd also like software that that supports the Integrating the Healthcare Enterprise (IHE) teaching file and clinical trial export. Siegel said that vendors have been slow to implement this capability, but once it has been implemented, a radiologist can capture interesting images, write a quick tag, and have them exported and saved for teaching file or research purposes.
Diagnostic workstations have the potential to function someday as a clinical or radiology assistant, Siegel said, but he predicts it will be many years before artificial intelligence is at a level where a computer can be taught to make visual decisions the way a radiologist does.
In conclusion, he advised that "when shopping for a PACS, remain aware that there are great possibilities of improving what we have. Don't be complacent! There's a lot that we expect the system to do now and be able to configure itself to do in future years. Most importantly, tell vendors to design PACS that are smart, fun, interactive, and have a great personality."
