Marrying form and function — that’s what image fusion is all about. This evolving technique — also referred to as fused image tomography — digitally joins anatomical images from CT and MRI with physiological images generated by a gamma camera or PET camera. It is proving beneficial in brain and prostate cancer imaging, as well as in surgical planning.
The best examples of what the future holds for image fusion were seen in June at the 47th annual meeting of the Society of Nuclear Medicine (SNM of Reston, Va.). Advances in this rapidly developing methodology were evident in the “Image of the Year” honor, which was presented to the University Hospitals/Case Western Reserve University (CWRU of Cleveland) for its 3D renderings of fused PET/CT studies of prostate cancer.
Inset: Prostate image given “Image of the Year” award at the Society of Nuclear Medicine meeting, from Drs. Zhenghong Lee and D. Bruce Sodee of the University Hospitals of Cleveland/Case Western Reserve University. Background: CT localization map (top), CT generated attenuation map and attenuation corrected PET data (lower left), and fused CT and PET studies (lower right) from SMV America’s PosiTrace system.
“We’re moving into a marriage of anatomy with function and physiology, and to me that certainly is the best for the patient in terms of diagnosis, management and follow up of a variety of diseases, including cancer,” says Robert Caretta, M.D., immediate past SNM president and medical director of nuclear medicine at Sutter Roseville Medical Center (Roseville, Calif.). “It gives us a chance to get true attenuation correction for each individual patient.”
Michael F. Hartshorne, M.D., chief of imaging service at New Mexico VA Health System (Albuquerque, N.M.) and vice chair for radiology at the University of New Mexico (Albuquerque), describes image fusion as providing more than just a comparison of the individual studies. “It’s not that it just makes abnormalities more conspicuous,” he adds, “they’re easier to understand.”
Image fusion involves co-registering data sets from two different imaging modalities, so the finished product appears as an overlay of images.
“We usually leave the anatomy image in black-and-white and we use the nuclear medicine images like a color wash, so you can tell which is which,” Hartshorne explains. “The kind of fusion I’m doing is to take the same data from two wildly different modalities and register them pixel for pixel, voxel for voxel, so that they absolutely match. Then, we’re able to display them so we can see one or the other or both at the same time.”
Two approaches
Image fusion can be accomplished either during a single examination at the point of image acquisition with a hybrid PET/CT scanner or after the image data has been generated in imaging studies from nuclear medicine and anatomical modalities, such as CT and MRI.
A joint venture between Siemens Medical Systems Inc.’s Nuclear Medicine Group (Hoffman Estates, Ill.), CTI PET Systems Inc. (Knoxville, Tenn.), the University of Pittsburgh Medical Center (UPMC) and the National Cancer Institute (Bethesda, Md.) has resulted in the development of a hybrid CT/PET scanner. The group is in the process of submitting a 510(k) application to the the FDA. Their goal is to have a commercial version — as yet unnamed — available by this November’s annual meeting of the Radiological Society of North America (Chicago).
This works-in-progress prototype installed at UPMC has been used primarily in imaging patients where internal organ movement presents challenges.
“If you have a combined device, you scan the patient and you have everything there,” says David Townsend, Ph.D., professor of radiology in the UPMC’s radiology department and co-director of UPMC’s PET facility. “The fusion is trivial, because it is just overlaying images, since you know exactly where they are relative to one another.”
Another benefit to hybrid scanners is that the patient is required only to endure one imaging study. This benefit is particularly important when patients must travel long distances to an imaging center.
“We have a large catchment area in Pittsburgh,” adds Townsend. “Some people drive 50 miles to come to the hospital. Given our heavy patient load, it’s almost impossible to schedule a nuclear medicine study and a CT at the same visit, so patients have to make two trips.”
Another hybrid scanner is the PosiTrace, a works-in-progress from SMV America (Twinsburg, Ohio). Introduced in December 1999, the first clinical studies were obtained in May of this year in Rennes, France. FDA 510(k) clearance is pending.
The combined PET/CT scanner does not require image registration. “We are fusing images during acquisition, as opposed to during processing,” says Lonnie Mixon, SMV’s vice president of worldwide marketing. “That really gives a distinct advantage — the confidence that the data sets are perfectly matched, which physicians need to know” in order to plan surgical procedures.
The second technique involves fusing the images when the data is processed. Originally, image fusion studies were accomplished with manual co-registration of image data, usually in brain studies fusing PET and MRI studies. Dennis Nelson, Ph.D., associate professor of radiology at University Hospitals/Case Western Reserve University, was involved in accomplishing the manual manipulation of image data sets, which took approximately an hour to complete. Today, software applications have improved so that automatic co-registration is possible.
IDL connection
Among the developers of automatic co-registration software is Research Systems Inc. (Boulder, Colo.). The company developed Interactive Data Language (IDL) as a programming language designed specifically to work with images and other technical data. “For example, many images from the Hubble space telescope are processed by IDL,” says Lawrence White, Research Systems’ medical product manager.
“Researchers are using IDL within the medical domain for a number of things; co-registration just happens to be one of them,” says White. “IDL is not a fusion application, but it enables those types of applications to be created in a much shorter amount of time than if you were developing them in C++.”
C++ is a high-level computer programming language designed and implemented by Bjarne Stroustrup, Ph.D., at AT&T Labs/Research (Florham Park, N.J.). Currently, C++ is one of the most popular programming languages used for graphical applications for a variety of computer platforms.
“We use IDL in programs because it works on many different platforms,” says Nelson. “It is a programming language that makes it easy to work with images. For imaging, the amount of code I write is about one-fifth that I would write if I were writing in C.”
The fusion application the Case Western Reserve team is designing is called Medical Image Merge (MIM). FDA 510(k) clearance is pending.
Another member of the Case Western team, Zhenghong Lee, Ph.D., assistant professor of radiology, has combined IDL with C++ to produce 3D image visualization capabilities.
“This will greatly help us for brachytherapy planning,” Lee says. “Traditionally, [physicians have] used 2D images and constructed in their minds where to place the radioactive seeds three-dimensionally. What we did is to display [the image] for them, and we can rotate it to any angle.”
Another aspect of Lee’s work is developing automatic alignment between SPECT and MRI images. Using an optimization routine to search and find the best aligned position, the need for manual co-registration is not necessary. “Instead of human intervention, the computer does the work,” Lee concludes.
Siemens has incorporated IDL into its workstations, with ICON-IDL as its Macintosh-based product and a Windows-based product now under development.
Image flexibility
One of the benefits to image fusion during processing is flexibility.
“I’m a radiologist in a hospital that does 85,000 cases per year, and my small group runs an imaging center,” says New Mexico’s Hartshorne, who uses image fusion software from Marconi Medical Systems Inc. (Highland Heights, Ohio). He explains that sometimes he may want to fuse images from modalities other than just CT and PET.
“Let’s say with a bone tumor, I’d like to be able to fuse the CT of the bone tumor with the MRI image,” he adds. “The MRI shows the soft tissue components exquisitely, but the CT shows the calcified matrix of the tumor that is basically invisible on the MRI. I make up a bone window CT scan, which we usually paint orange to make it visible, and then paint that on top of the MRI scan.”
Not only can Hartshorne see the location of the tumor, he can see “where the reactive calcified stuff is in it, around it, etc. It is easier to read than holding both of them up and looking back and forth.”
Clinical applications
Most clinicians who use image fusion techniques engage in the diagnosis and management of patients with cancer. At Wake Forest University Baptist Medical Center (Winston-Salem, N.C.), for example, the primary clinical application for image fusion is brain tumor imaging.
“We have patients who have been treated for their primary tumor and now they’re trying to determine if there has been recurrence, which can be a very difficult task to do using just MRI,” says Beth Harkness, M.S., assistant professor of radiologic sciences. “We register our FDG/PET scan to an MRI image. It makes the physician’s level of confidence in interpretation much higher.”
The Wake Forest radiology group uses automatic image registration (AIR), which was developed by three researchers at the University of California-Los Angeles School of Medicine, to perform image registration.
AIR was developed by John Mazziotta, M.D., director of UCLA’s division of brain mapping, vice chair of neurology and professor of neurology, radiological sciences and pharmacology; Simon Cherry, Ph.D. associate professor and associate director
of the Crump Institute for Biological Imaging at UCLA; and Roger Woods, M.D., assistant professor of neurology at UCLA.
Hartshorne explains that the facility’s brain scans employ Thallium 201 or fluorodeoxyglucose (FDG) as the nuclear medicine imaging agent to locate brain tumors and differentiate them from brain necrosis. “To understand those nuclear medicine scans is easy once you have them co-registered with the MRI scan that showed something questionable” he says.
Fused images also are utilized in cancers of the liver, lung and pancreas. Such is the case for Peter Faulhaber, M.D., assistant professor of radiology at Case Western and director of nuclear medicine at the Louis Stokes VA Medical Center (Cleveland).
“We’ve had cases where no other study demonstrated a lesion in the liver, but they were able to find it in the operating room with intraoperative ultrasound correlated with where we told them [the lesion] was, based on image fusion,” Faulhaber says. “The same is true for lung cancer. If I find a mediastinal lymph node with a very small finding on CT scan, we’ve used the fusion to direct biopsy by mediastinoscopy or even bronchoscopy by telling them where the node is relative to the bronchial tree.”
The diagnosis and treatment of prostate cancer is a primary focus for D. Bruce Sodee, M.D., associate professor of radiology/nuclear medicine, University Hospitals/Case Western Reserve University.
“We fuse with five millimeter sections of CT and MRI co-registered with appropriate ProstaScint SPECT images,” Sodee says. ProstaScint is a monoclonal antibody tailored to attach to prostate-specific membrane antigen. The fused images have a direct impact on therapy planning, including which intervention to employ: prostatectomy (surgical removal of the prostate), brachytherapy with radioactive seeds, or external beam radiation therapy.
“We have worked closely with a brachytherapist for the past three or four years,” Sodee continues. “He places radioactive seeds, either I-125 or Palladium, and uses our information to place the seeds where the cancer is in the prostate.” Looking at a volumetric image that can be rotated to visualize all planes allows precision in brachytherapy not possible before.
Hartshorne describes the use of image fusion to diagnose and manage lymphoma. CT scans can detect enlarged lymph nodes, but cannot differentiate between ones that are sterile and ones that are tumorous. With a gallium PET scan, tumors are illuminated. A fused image “makes the gallium scan easier to understand and the CT scan much more specific,” Hartshorne adds.
Future Directions
Given the work load of most radiology departments, automatic co-registration is essential for image fusion to prove a practical adjunct. Most radiologists simply do not have the time to perform manual registration of two image data sets.
Triple fusions are part of the equation at CWRU and may lead to further applications. “We are doing triple fusions with ProstaScint for prostate cancer with FDG/PET and CT and MRI. One is for antibodies to the disease; one is for metabolism of the disease; and the third is the anatomy,” explains Faulhaber.
CWRU’s Lee suggests that studies to determine uptake of medications may be another application for image fusion in the future.
“When people want to study receptors, such as pharmaceutical companies that want to do drug studies, they want to know in vivo biodistribution,” Lee says. By radiolabeling the medication and performing perfusion studies fused with anatomic information, they could determine pharmokinetics to understand drug effects.
Caretta suggests that as new treatment modalities are developed, image fusion may play a vital role if radioactive antibodies or peptides were used to treat cancer. “You could actually do modeling of dosimetry to the individual tumor sites,” he notes.
Future standard
So far, image fusion techniques have provided valuable new information for clinicians working in several settings. As these techniques proliferate, additional applications in other healthcare settings no doubt will be developed and many nuclear medicine observers see image fusion is becoming the standard, rather than the exception, in medical imaging.
Henry Wagner, M.D., one of the founders of nuclear medicine and a past SNM president, predicts that within five to 10 years, “60 percent of all imaging studies will be fused images.”
And Wagner is not alone in his prognostication.
“Last year, I said that in five years, if you weren’t doing image fusion, you would be making excuses for why you weren’t doing it,” says Hartshorne. “For practical purposes, I’ll bet you that 80 percent of the utility of fusion is going to be taking advantage of what I call the ultimate contrast agent: radiopharmaceuticals.”
