The marriage of anatomic data from computed tomography (CT) and magnetic resonance (MR) imaging to the metabolic data from positron emission tomography (PET) and single-photon emission computed tomography (SPECT) is an ex-ample of the value of the whole being greater than the sum of its parts. Such fused images provide radiologists, nuclear medicine specialists, and other physicians with more meaningful data that provided by either image individually. Although the practice of merging images has been around for some time, at least in a rudimentary form, recent advances in both hardware and, particularly, software technology have taken this process to a new, more practical and efficient level.

Methods of image fusion

The first and most common method of image fusion is "mental fusion," in which two images are placed side by side and the radiologist attempts to mentally overlay the images in his or her mind. This is the most basic and inexpensive form of image fusion. "The only expense involved in this method is the time of the individual performing it," said Thomas J. Hook, senior vice president, CTI Molecular Imaging, Inc., Knoxville, TN. "The true costs of this method, however, are the errors that are made in trying to reach accurate conclusions by looking at the images without overlaying them. Obviously, the cost can be quite severe if there is a missed diagnosis or a misreading of the exam."

The next method of fusing images involves overlaying the images either manually or through the use of simple software programs that scale and align the images. The difficulty with this method is the ability to accurately align two images acquired at different times. "A human is not a static object," noted Hook. "Organs within the body move, the person moves. So software that does simple rigid registration typically allows you to do a better job than the mental fusion, but it's still a relatively simplistic approach and you still end up with mismatches in the two images."

The third level of fusion is one in which a combined image is obtained on a hybrid imaging device, such as a PET/CT scanner. The benefits of such a system include the ability to acquire both images in the same plane at the same time. The drawbacks are the considerable capital costs associated with purchasing a hybrid imaging system and the fact that the images are still fused based on a rigid software algorithm.

There are times, however, when such hardware-based fusion imaging is the method of choice. "There are certain clinical applications in which it is beneficial to have all the information gathered at the same time, for example, in cardiac applications where it is difficult to synchronize the images if you don't acquire the information at the same time." said Christian P. Behrenbruch, PhD, CEO of Mirada Solutions, Ltd., Oxford, UK.

Hardware fusion can also, at times, increase clinical throughput. "Having the fusion on hardware can mean that the clinical throughput is faster and the image quality may be better," said Behrenbruch. "Scheduling a greater number of patients for imaging in a shorter period of time is also beneficial."

For patients, as well, hardware fusion can be beneficial. "Hardware-based fusion brings benefit to the patient," said Clare Jones, vice president of marketing at Mirada, "because they only have to undergo one scan rather than two." Image distortion and decreased resolution are still a possibility with simultaneous image acquisition. This can cause misalignment between the center and peripheral areas. The use of sophisticated fusion software can address this problem.

A new level of fusion

"Those are the three initial levels of fusion," said Hook. "The Mirada Reveal MVS workstation (Mirada Solutions and CTI Molecular Imaging) adds a whole different level to fusion imaging because Mirada brings what's called 'non-rigid' or 'non-linear' fusion capability to the table."

With a nonrigid or deformable technique, one image is manipulated to match the other. "Rather than trying to match up entire datasets--for example, the entire PET dataset with the CT dataset--you try to line up various regions or more localized densities or areas of information that match up and then deform the shape of the PET image to make it fit the shape of the CT image," said Todd Deterding, senior product manager for CTI. " The image quality of the PET image remains the same, but it is placed in the context of the CT, where it makes the most sense."

There are a variety of benefits associated with the use of this nonrigid, software-based fusion system. First, it can be used with images acquired at any time and on all types of modalities. "Ultrasound, X-ray, MRI, PET, CT--you can bring in all these modalities," noted Behrenbruch. "For example, you don't find an MR system bolted onto a PET system. The physics of creating such a system are challenging. Yet there are all kinds of applications, such as in neurology, where a radiologist would love to compare the results of MR and PET images."

"The Reveal MVS workstation can give you a 'pseudo device' even though those things don't exist," agreed Hook. "Obviously the more images you can 'morph' and overlay, the more power it gives the radiologist or nuclear medicine physician reading the exam to actually make a more definitive diagnosis."

Another advantage of software solutions is the ability to perform longitudinal comparisons. The user can compare images over time to assess treatment response or any other variable. "There are all kinds of possible applications," Behrenbruch explained. "When you have an oncology patient being worked up and they are about to be sent to surgery, by the time they are sitting on the operating table, hundreds of thousands of dollars have been spent on that person, probably including a PET scan, an MR, and a CT scan. The more information you can integrate to build a complete picture of what's happening inside the body, the better the benefit for the patient."

Other applications of deformable software-based fusion include use in comparing the results of contrast-enhanced breast MR and the metabolic activity from a PET image, particularly when assessing response to therapy over time. "If you are looking at liver lesions, a helpful modality for that is ultrasound," noted Behrenbruch. "So if you could do ultrasound/CT fusion, that would have a lot of practical applications, particularly when you want to start doing ablation."

Another benefit to the software-based solution is that it sits in your information domain and, therefore, provides flexibility when upgrading hardware components.

"If your PACS evolves, for example," said Behrenbruch, "you add a new plug-in and you get better DICOM connectivity, or if you switch to Web-based PACS, information-type applications can move with it. In the case of hardware-based fusion, the obvious disadvantage is that if you want to upgrade your hardware system-for example, if you want to use a different detector material or you want to upgrade to a 132-slice CT scanner when it becomes available-you would then have to have someone upgrade part of your system or send it to the factory for a couple of weeks. So software fusion has intrinsic value."

Economies of time and money

There are also economic benefits to this system in terms of both time and money. "The economic reality is that, sometimes just for dollars and cents reasons and sometimes for space considerations, a PET/CT is not an option for a provider," said Hook. "The only option may be to use a workstation with very sophisticated software to get a fused image."

"There are a lot of people doing functional imaging who don't have the economic clout to get a PET/CT scanner," said Behrenbruch. "There are also some very good operational examples as well in which PET/CT scanners are not practical, such as mobile machines." The cost of outfitting a coach for mobile PET/CT and the cost of driving an extra 10 tons of weight can be prohibitive.

The workstation-based system is also time efficient. "Fusion software used to be incredibly slow and labor-intensive," noted Behrenbruch. "You had to click lots of points down, or if it was a more sophisticated type of fusion, you had to wait-sometimes for a couple hours-for it to register the two images together. Today, it's instantaneous. With our platform, the average time to complete image fusion is about 20 seconds."

"In many cases it's not actually the radiologist performing these functions," added Jones. "With our fusion software it very well may be the technologist who performs the fusion and passes the case to the radiologist for review."

"That has been one of the challenges of fusion software," continued Behrenbruch. "Unless you can make the fusion really robust, you can't leave it in the hands of a technologist."

Combining software and hardware

Hardware- and software-based fusion are not mutually exclusive. "It's also equally powerful that that same MVS Reveal workstation can be used in a PET/CT environment," said Hook. "You can collect the same dataset on a PET/CT and still use the REVEAL MVS to remove even more errors from the fused images. It's really a tool to be used across the industry."

"Software is totally complementary with a hardware solution," concluded Behrenbruch. "There are very different reasons why you do each one. In the future, what we will see is software-based technologies augmenting the capabilities of both systems in terms of accuracy, longitudinal comparison, and in really integrating other modalities. It's like glue in a multiple imaging environment."