Cancer Care Informatics: Challenges In Nuclear Medicine

Judging solely by the wealth of clinical trials devoted to exploring biomarkers in cancer care, the potential of nuclear medicine appears limitless—until one considers all the informatics issues and PACS pitfalls. Image management is still for many stuck back in the days of “unclear medicine.”  

There is a big gap today in reading nuclear medicine and molecular imaging exams, namely because standard radiology PACS lack the ability to fuse function and anatomic data. “This is a huge problem for PET/CT, SPECT/CT and PET/MR studies—which have become the most commonly performed studies in nuclear medicine today, ” explains Ryan Niederkohr, MD, chief of the department of nuclear medicine at Kaiser Permanente Santa Clara Medical Center in California.

Reading a nuclear medicine study on a standard PACS setup is a challenge, if not impossible. Greyscale displays from the early 2000s were still trying to imitate light boxes, but oncologic nuclear medicine studies are much more complex, require color and are a challenge to size correctly. Some specific complaints:

• Whole body images are typically 256 x 1024 pixels, which appears small on a standard diagnostic monitor.
• Renal scans include dynamic images and color graphs to distinguish measurements from the left and right kidney—a feature that doesn’t work on a black and white display.
• Even when color displays became more popular, they didn’t have the features or flexibility of dedicated nuclear medicine processing stations.
• While radiology PACS can display the separate images from a PET/CT, they may not correctly fuse the datasets. If both images can be displayed, traditional PACS may not be able to sync them as a nuclear medicine specialist scrolls through the images by slice because slice thickness and matrix size vary between PET and CT.

Hanging protocols—how images are arranged when viewed from the PACS—and windowing controls also can be a headache, says Jerold W. Wallis, MD, associate professor of radiology at Mallinckrodt Institute of Radiology, Washington University School of Medicine in St. Louis. When reading a nuclear medicine study, specialists generally want the lowest values (e.g., darkest blacks) unchanged while making adjustments to the highest values (e.g., the brightest portion of the color scale). “Trying to do that without inadvertently messing up the way the image is displayed on a system that is designed for CT is quite difficult,” he says.

Dedicated nuclear medicine workstations are often connected through a stand-alone network, and while they offer the full functionality specialists need, communicating data to other physicians through the hospital PACS represents a major hurdle.

“You may have a good environment for the physician who’s reading nuclear medicine and PET, but that oncologist who wants to see PET/CT images on PACS might just see a PET scan and a CT and have no way of correlating between the two just as if the patient had two different scans at two different times,” laments James R. Galt, PhD, director of nuclear medicine physics at Emory University Hospital in Atlanta.

Installing the full nuclear medicine reading environment on PACS workstations may not be possible. Hardware requirements for some PACS workstations are not nearly as intense as they are for nuclear medicine workstations, and even if the general PACS computers could handle the upgrade, it often costs big bucks for additional nuclear medicine software licenses. Major PACS vendors don’t prioritize inclusion of nuclear medicine functionality in their standard products.

“Because nuclear medicine and PET are a fairly small portion of the radiology market, most of the attention from PACS vendors has been in plain film, CT and MR,” says Galt. “They still don’t do a very good job of displaying nuclear medicine images in the basic PACS display software.”

Fix its

Organizations are forced to find ways to work around the various issues with sharing nuclear medicine studies with referring physicians. For example, Emory creates a file called a multi-frame secondary capture (MFSC, a DICOM compatible object) of each PET/CT or SPECT/CT, Galt says. The dynamic information of the original scan is relayed in what is basically a movie file that the referring physician can scroll forward and backward through, but other controls are limited. It’s not a format that would be easy to make initial interpretations from, but Galt says referring physicians are happy to have some type of image with the written report. MFSC is becoming fairly standard, but some workstations still may not be able to create such images.

Even sharing the report by itself is a challenge, notes Galt, since nuclear medicine workstations may not be tied into the hospital information system’s dictation tool. A physician must instead launch dictation on another system and select the patient there before moving to the nuclear medicine workstation and selecting the patient again to bring up the study. It’s a speed bump in workflow and creates the chance that the wrong patient name will be clicked and the wrong image read.

Some organizations are able to work with their PACS vendor to tinker with hanging protocols and windowing controls, but some organizations prefer to look for answers from the ground up.

DIY nuclear medicine PACS

When the nuclear medicine specialists at Beth Israel Deaconess Medical Center (BIDMC) in Boston wanted a system for viewing their studies, they took matters into their own hands. They developed their own, self-contained PACS database specifically for nuclear medicine, and they are freely sharing it with anyone who wants it. Dubbed OpenPACS, the shareware software runs on any PC running Windows 7, Windows Vista or Windows XP, and studies can be interpreted from the database using a host of specialized functions on an associated PET/CT viewer.

Gerald M. Kolodny, MD, director of nuclear medicine at BIDMC and supervisor of OpenPACS design, says the process for developing the software was special because with most other PACS, the level of input from nuclear medicine specialists is minimal. For OpenPACS, however, Kolodny and colleagues pushed the programmer to make a tool that was intuitive for nuclear medicine physicians to use. 

“What we wanted to do was more difficult for[the programmer], but made life much easier for us,” he says. “To get good software, you need to have a physician—a radiologist or nuclear medicine physician—involved.”

The final product allows users to see PET studies, either attenuation corrected or uncorrected, the corresponding CT and the fused studies all side by side. Linked cursors on each image allow for precise location selection. Fused slices can be linked and presented in transaxial, sagittal and coronal planes.

If a facility doesn’t want to install an entirely new PACS, but the current system doesn’t have the necessary functionality, a nuclear medicine add-on module from either a third party or the original vendor could be a solution. 

“One of the problems with integrating everything into PACS software is our field is changing so quickly, how can they keep up? The add-ons usually have a different team working specifically on nuclear and will probably be able to stay ahead, but the basic PACS software might not catch up,” says Galt. “The add-ons have the best potential to meet the need of the nuclear physician in the reading room.”

Each add-on has its own list of unique features, but image fusion and quantitative analysis tools are among the common major additions. Workflow efficiency stands to gain from automatic tumor segmentation tools for PET/CT. Add-ons that feature disease specific workflows can speed interpretation for different cancer types.

Preparing for the future

As clinical research leads to its own breakthroughs in cancer care, those on the IT front need to continue to make progress as well. An example of a clinical leap that has yet to be fully adopted by vendors is the PET Response Criteria in Solid Tumors (PERCIST) criteria. PERCIST provides criteria for quantitative PET evaluations of cancer patient response to treatment. One of the improvements it makes over previous criteria is the switch from measuring SUVpeak rather than SUVmax. The latter measures tracer uptake as a sign of tumor aggressiveness, but only looks at a single pixel within the tumor, while the former analyzes a larger volume. This makes SUVpeak less susceptible to noise, but it’s a measurement that might not be available on all systems. 

It’s impossible to know what the next prognostic indicator is going to be or what biomarker will prove meaningful in the future. Once new measures are validated, however, those on the front lines of the battle against cancer need the right equipment to begin tracking them.