Advanced Visualization: A Do-it-yourself Approach
With three-dimensional and other advanced visualization tools being used for a greater share of CT and other imaging studies, it is more important than ever for radiologists to have access to postprocessing software and equipment at the point of interpretation. It is equally important for radiologists to develop and keep up their skills in using these tools. This means that hospitals and radiology practices must decide which kinds of advanced visualization tools they will put in place and where and how they will deploy them. Will they opt for enterprise-wide, single-vendor solutions, or will they choose best-of-breed approaches that might limit image distribution in favor of the ability of one vendor’s tool to perform a task particularly well? At Albert Einstein Medical Center (AEMC) in Philadelphia, both approaches are used. Radiologists have access to advanced visualization tools embedded in their PACS software (Synapse from FUJIFILM) as well as a thin-client advanced visualization tool that has been interfaced with the PACS, tools powerful enough to handle the majority of the radiologist’s post processing needs. For other studies, like virtual colonoscopies, the hospital uses specialized post-processing tools. Terence A.S. Matalon, MD, FACR, FSIR, is chair of the hospital’s department of radiology. He has been at AEMC for five years and has seen the hospital progress, he says, from largely film-based imaging to the PACS-based operation of today. Matalon says that about half of his time is spent administratively. During the other half, he’s a clinician who splits his time equally between interventional and diagnostic radiology, with the bulk of the diagnostic work involving CT. He recently shared his approach to volumetric imaging with AEMC is a tertiary-care teaching hospital affiliated with Philadelphia’s Jefferson Health System and Thomas Jefferson University. AEMC has about 20 radiologists on its faculty, and each year, it admits 24 radiology residents. The radiology department performs about 250,000 diagnostic examinations annually, according to the hospital’s Web site. Start Thin Because advanced visualization is used more and more in CT imaging and, to a lesser extent, in MRI and ultrasound, Matalon says that it’s mandatory for radiologists who are going to be looking at multislice examinations to have access to postprocessing tools. He is also a stickler about radiologists doing—or at least being able to do—their own reconstructions. Matalon says that one core concept at AEMC is to start with thin CT slices and build from there. A second is to have ubiquitous access to post-processing tools that can be accessed through PACS. This is the foundation that broadens to more specialized postprocessing when needed. “We take the approach where we acquire the thinnest slices available on every patient that we image using CT,” Matalon says. “We send those images both to PACS (Synapse from Fujifilm) and to our 3D solution, which happens to be TeraRecon, as a default for every patient. That allows us, with a single click, to bring up 3D applications with that particular patient’s images already loaded right from the PACS, so the threshold associated with manipulating those studies is very low. Anyone interested in using a 2D or a 3D multiplanar reconstruction has very, very low work to be able to do that,” Matalon says. “The technologists will not only send thick and thin slices to the PACS, but will also routinely send coronal and sagittal reconstructions to the PACS,” he adds. “Whenever we bring up a patient, the axial, coronal, and sagittal reformats are immediately available to us as part of that patient’s default presentation.” For any off-axis views or 3D work, the radiologists can bring up the third-party application and produce added images that can be sent to PACS. Matalon makes the point that a big part of the value that radiologists add to the diagnostic process is their ability to create reconstructions themselves. “For CT angiography—for example, of the aorta and the lower extremities—the technologist will typically obtain routine 3D reconstructions and send those to PACS. My personal opinion is that, in general, these are not adequate for my purposes, and I will routinely produce my own. In the process of interpreting exams, I will use the 3D tool to evaluate the patient’s dataset. I will create what I call a movie for the clinicians and for anyone else who follows me. That allows people to see the pertinent lesions that a patient might have that are amenable (or not amenable) to vascular reconstruction,” Matalon says. He continues, “It is that ability to interrogate those images using the 3D tool that differentiates us and adds value to the exam, over and above the routine 3D views that a technologist or a so-called nonphysician imaging specialist might be able to render.” Matalon also point out that radiologists who do their own reconstructions build a knowledge base as they go along. “There are a number of very specific applications that we and others have found very useful in interrogating large datasets,” he says. “For example, using a thick-slab maximum-intensity projection (MIP) is extremely useful in bringing out subtle lung nodules. Far superior to scanning through 500 images of the chest is to use the thick MIPs to make subtle lung nodules more apparent.” He offers a second example of how a radiologist’s reconstruction expertise can help a patient, saying, “One of the tools that I like is to use is a curved planar reconstruction for the vertebral artery. The vertebral artery, as it travels in the neck, is protected by the neural foramina, or by the bone surrounding the neural foramina. Many of the bone-removal techniques are not adequate for that particular artery because they confuse the artery with the bone.” He continues, “By performing a curved planar reconstruction, one can look at the artery in its entirety, from the subclavian artery origin to the basilar artery. It is a somewhat tedious approach, but it is an approach that allows one to remove the bone somewhat confidently and accurately, as opposed to automated techniques that might not do that.” To handle its different advanced imaging needs, AEMC has deployed a number of tools, in addition to those available in the PACS and the thin-client application embedded in the PACS. These tools are used to process huge data sets, virtual colonoscopies, and coronary arteries. “Most server-based products are, by their very nature, less robust than a dedicated workstation,” he says. “They just don’t have the horsepower that a dedicated workstation has, to be able to perform some of the higher-end bone-removal techniques and so on, but a server-based product is a fantastic solution for doing limited 3D and coronal/sagittal and curved planar reconstructions without having to get up [from the PACS workstation. The Voxar, Vitria, Advantage, and Macintosh boxes have interfaces with the PACS, so we can upload images to those applications, produce 2D and 3D products, and send them back to PACS.” Master of None Just as he argues that radiologists should take a personal, do-it-yourself approach to reconstructions, Matalon says that institutions should tread lightly with vendors that try to be all things to all clients and fill all the niches in an enterprise solution. Institutions, too, should take a do-it-yourself approach to building their advanced-visualization arrays, he argues. “It’s hard to be everything,” he says. “It’s hard to be (the vendor for) PACS, RIS, 3D, and voice recognition. At some point, when you try to be a single-enterprise solution and ignore the best-of-breed concept, there is a point where you become a jack of all trades and master of none. You end up not having as much value as if you just focus on your primary business.” He says that one reason that enterprise-wide solutions are attempted, and have become popular, is that they are a natural fit for IT departments, where integration is a sort of holy grail. It’s a philosophy that AEMC’s radiology department has adhered to in mounting its own multivendor postprocessing array. Matalon offers another example of the reason for this, noting, “If you are just looking at spines or pulmonary emboli, or for lower-extremity peripheral arterial disease, those can generally be dealt with through an all-purpose tool…that is more than adequate to answer the clinical questions in those types of settings.” He continues, “Alternatively, if you are doing a virtual colon, there are specific applications that are optimized for it. The ability to filet the colon open electronically and bivalve it, if you will—so that one can see both the anterior and posterior sections of the colon artificially cut open for you—is a very unique tool that is inherent in a product that is best-of-breed and is dedicated to only that kind of examination. We take the approach that there is a need to be a jack of all trades, which we believe the server-based 3D tool is, augmented by these best-of-breed applications for specific, high-end, narrow indications.” Matalon also notes that the push to go filmless and install PACS in the past decade or so coincided with a huge flood of images coming from newer CT scanners and, to a lesser extent, from other modalities. As a result, he says, many of the early PACS were overwhelmed, and many institutions are into second-generation PACS now. It has become obvious to most radiology administrators that they now need some advanced-visualization, postprocessing solution, whether that is an enterprise solution, a best-of-breed solution, or the use of both systems in tandem, as at AEMC. Workflow At AEMC, workflow is handled in what Matalon calls a tiered format, with the less-demanding cases handled using the PACS-embedded tool. “For CT exams, we will use the [thin client tool] 50% to 60% of the time. Remember, in addition, that every one of our cases has sagittal and coronal reconstructions presented to the radiologist just as part of the case. We recognize the need to reformat in every case, to produce those images as a default in every exam. Even in the 40% to 50% where we don’t open [the thin-client tool], people are using reformatted images as part of the interpretive process without having to open a 3D or 2D tool.” He adds, “That’s probably one area where [Synapse] would be very useful and allow us not to have to do that work on the front end. Even though, now, it is generally done automatically by the modality, it still adds overhead to the network and adds cost to the archives. That’s one advantage that [Synapse] will offer in the future for us: not having to produce those images, send them to PACS, and then pay for the storage of those images (that is, the sagittal and coronal reconstructions).” Matalon reiterates the need to start with thin slices to maximize utility. “It’s not always apparent which cases will require additional manipulation,” he says, “so if one acquires images in less-than-optimal slice thickness, any subsequent manipulations will be suboptimal, versus the same data acquired at a thinner slice count.” When it comes to the financial viability of advanced imaging and whether it can be broken out as an income producer, Matalon says, “I don’t know the answer. It’s not an activity that we do to produce income. It’s an activity that we do to be more efficient and to interpret exams more accurately.”