Meeting Molecular Imaging's Challenges

Managing, distributing, navigating and storing molecular imaging studies is the next horizon to conquer in taking full advantage of these complex images that are increasing in clinical significance.

Bruce Line, MD, director of nuclear medicine at the University of Maryland (College Park, Md.), predicts that clinical molecular imaging, or PET-CT hybrid or fusion imaging, will fuel a tremendous jump in radiologists' clinical problem-solving capabilities with payoffs in oncology and nuclear cardiology. The molecular imaging umbrella can cover SPECT-CT and MRI as well. Regardless of the modalities employed, sites that have taken the lead in clinical molecular imaging are reporting some significant challenges in image distribution, management, navigation and storage. These challenges need to be overcome before molecular imaging can be seamlessly integrated in the radiology department. Consider:

Most PACS workstations are not designed to view molecular images. So how do clinicians view these information-rich datasets?Even with fast Ethernet networks, sending a 400 megabyte dataset across the hospital can be a bit taxing. How can sites route molecular images where they need to be now and in the future?

Radiologists need tools to rapidly navigate through molecular imaging data and mine the rich information available in the images, but the technology to enable sophisticated interrogation is not yet available.

"In general, the world is still catching up with the concepts of hybrid imaging and how to [distribute and] view these images," Line says. As molecular imaging moves into the mainstream, viable solutions to these challenges become more critical. Jeffrey Leal, research associate at Johns Hopkins University School of Medicine in Baltimore, explains, "Molecular imaging is really taking off because of fusion imaging, and it will be the central focus of radiology in the next 10 years."


Johns Hopkins University is on the cutting edge of molecular imaging with four SPECT-CT cameras, a PET-CT scanner and a PET camera. And clinicians are rapidly discovering and embracing molecular imaging. The hitch is image viewing. In an ideal world, molecular images would flow into the PACS environment so that clinicians could readily access the information for clinical decision-making. But Leal confirms, "Traditional PACS aren't designed to handle all of the components of molecular imaging [fusion scanners, SPECT, CT, etc.]." Line agrees and adds, "To date, there aren't good standards for fusion imaging that allow DICOM transfers of fusion images to a conventional PACS workstation." At Johns Hopkins, the nuclear medicine department receives daily calls from physicians who want to see molecular images but can't because their PACS workstations aren't capable of displaying them.

Johns Hopkins has resorted to old-fashioned solutions to this 21st century challenge. In some cases, the nuclear medicine department prints hardcopies of fusion images for physicians, which is hardly a timely or cost-effective option. In others, the physician treks to the nuclear medicine department or reading room to view images, which takes up valuable physician time, and once again, thwarts real-time decision-making. Finally, some software systems allow users to create a CD of molecular images. Leal notes, "These tend to be less than optimal. They're typically static images that don't give physicians the flexibility to peruse the data. Physicians should be able to peruse and quantify this data."

At the neighboring University of Maryland, the status of molecular imaging image viewing and distribution is similar. That is, radiologists can view the CT portion of PET-CT images via the PACS archive and infrastructure, but they must visit the nuclear medicine department to discuss and view findings of the combined scan. Furthermore, there are geographic distribution challenges. When the results of a full body PET-CT are distributed to various sections - neuroradiology, chest and abdominal radiology - radiologists find it difficult to see the same data as the reporting radiologist in nuclear medicine because they aren't viewing the images in the same way. "Radiologists need some way to view and manage fused data on a DICOM workstation in interrogative ways," Line says.


Line maintains that the current network infrastructure at the University of Maryland is sufficient for the upcoming onslaught of molecular images. "Speed is not a problem," he confirms. "Our distribution channels are wide enough to transfer hybrid-imaging information to workstations. Our network infrastructure also is probably sufficient for the traffic we're talking about in the future." Line is optimistic about the university's network capabilities because the network was designed for 16-slice CT scanners, and the 16-slice CT on its PET-CT equipment mirrors the stand-alone CT scanners. Transmitting the PET portion of a PET-CT scan doesn't place a significant burden on the network. Other sites, however, do report some network challenges.

Leal says pushing a 400-megabyte molecular imaging dataset across a fast-Ethernet network can take a few minutes. At Johns Hopkins, the fast Ethernet network lets the nuclear medicine department push molecular images where they are needed, but radiologists can't pull data on demand. "We rely on auto-routing tables to push data where it's needed ahead of time," Leal explains. "I'm hoping that in five years we'll have gigabyte networking to the desktop and that this will allow us to push and pull molecular imaging data on demand."

Mukesh Harisinghani, MD, director of body MRI at Massachusetts General Hospital in Boston, says bandwidth is a major issue for centers outside of the main hospital. In fact, just about one year ago it took up to 45 minutes for a radiologist to view a CT scan at an outpatient clinic six miles from the hospital. The situation has improved over the last year as vendors developed more efficient compression mechanisms and the hospital increased bandwidth by upgrading from a T1 to a T3 line, says Harisinghani, but further improvements are necessary for the hospital and its satellites to more fully realize the benefits of molecular imaging.


The image storage needs for molecular imaging have pushed beyond the previously small needs of nuclear medicine, going from small to large, quite quickly. Johns Hopkins' Leal says, "We're generating a lot of images and data; the magnitude is much greater than what we're used to in nuclear medicine. Nuclear medicine used to be the smallest producer of image volume. Now we are one of the largest producers, generating 0.4 terabytes of clinical data a year, and we're still ramping up molecular imaging." The department realized its need for dedicated storage and decided to install its own nuclear medicine archive with a 700-gigabyte RAID and 3.5 terabyte DVD jukebox to handle all of the data. Leal says the department initially estimated that the solution would provide a decade of deep storage and 18 months of online storage. Steadily increasing volume, however, has forced the department to recalculate the situation. Leal admits, "Now we think we have five years of deep storage and less than a year of online storage."

The molecular imaging storage challenge will continue in the future, predicts Leal. "As PET-CT and SPECT-CT begin to embrace the [multislice] protocols of dedicated CT scanners, the volume will increase even more," explains Leal. And the next-generation PET-CT scanner that the department plans to add in six months will take two to four times more disk space than the current PET-CT scanner. Leal sums, "It's become a serious management issue." The department is handling the colossal volume by preparing auto-routing tables and planning procedures in advance.

Massachusetts General Hospital faces a similar situation. Harisinghani says, "Image deposit and retrieval poses a significant challenge. We only have 12 months of local storage. Everything else is on CD ROM or DVD, and physicians are forced to take the time to load and pull images and data. This will be an issue until we develop computer solutions that can hold more locally."


Although molecular imaging carries a number of challenges, industry is stepping up to the plate. "On the positive side, most manufacturers are aware that the future will demand greater molecular imaging functionality," Line explains. "And suppliers of 3D software are interested in fusing the 3D world; several are working on [solutions for] importing molecular images."

In the short-term, how can departments handle molecular imaging viewing, navigation, distribution and storage concerns?

At the University of Maryland, the nuclear medicine department works closely with radiology to help both players develop an understanding of the new world of hybrid imaging.

Leal of Johns Hopkins says he now participates in camera purchase decisions since adding a new camera requires advance planning for image distribution and distribution.

Plan for molecular imaging demands on the network in upcoming capital expenditures cycles. Gigabyte networking to the desktop could be essential to truly realize the benefits of molecular imaging across the enterprise.

Imaging in the 5th dimension

Osman Ratib, MD, PhD, professor and vice chair of information systems for the radiology department at the University of California Los Angeles, describes the challenges of molecular imaging navigation as navigating in the fifth dimension. The first consideration is displaying and navigating through the tremendous static data generated by multislice CT scanners. Throw in a fourth dimension, like a moving, beating heart, and the demands increase significantly. Add yet another dimension, functional or molecular data, and the demands on the traditional workstation may be considered extreme.

Ratib admits, "Capacity was an issue for viewing these images in the past. Five year ago, most PACS workstations had a 120-gigabyte capacity. Today, you can buy that capacity in a laptop, so capacity isn't a challenge anymore. The challenge is to make the radiologist comfortable enough to rapidly navigate through these images. A lot of software development has yet to happen."

Ratib and UCLA researchers developed OsiriX, a new open-source platform to address some of the challenges of molecular image viewing. The OsiriX platform is tailored for large multidimensional, multimodality images like PET-CT studies and allows users to view and visualize these images. Ratib and his colleagues also are exploring alternate joystick configurations with multiple buttons designated for the x and y dimensions, contrast, molecular imaging or functional data and time to further facilitate 5D navigation.