Applied Radiology

Dr. Leong is the Executive Medical Director, Mirada Solutions, Oxford, UK; http://www.mirada-solutions.com. Dr. Siegel is the Director of Nuclear Medicine in the Department of Diagnostic Radiology, Dartmouth-Hitchcock Medical Center, Lebanon, NH.

Image fusion refers to the merging and visualization of image data from more than one source to present the same information in a more meaningful and synergistic manner. ¹ Within the medical imaging environment, these data sources include modalities such as PET, SPECT, CT, and MR imaging. The data being fused may be from different imaging modalities or from the same modality acquired at different times (such as pre- and posttreatment images).

While the concept is simple, the process is not straightforward, particularly when the image data is in three dimensions and may differ in resolution, definition, matrix size, or even shape and structure. Fusion is a problem that has both mechanical and mathematical components, and, consequently, there have been both hardware (so-called "combined" or "dual-modality" or "hybrid" scanners) and software solutions. That the two should be mutually exclusive has largely been due to the competitive nature of business. In fact, as many are now realizing, hardware and software approaches to fusion are a complementary, rather than adversarial, combination. This article is intended as a brief overview of clinical applications of fusion imaging.

Fusion plays an important role in every radiological practice

Fusion can play a role in nearly every type of radiological practice, from those with mobile PET facilities to those employing dual-modality scanners. We have seen clinical sites that regularly fuse PET data from mobile PET with CT or MR images imported from another hospital. In nuclear medicine departments, such as at Albert Einstein Medical Center (Philadelphia, PA), SPECT data is fused with separately acquired CT or MR images, sometimes revealing pathology that was missed in the earlier CT or MR report. At Children's Hospital (Boston, MA), preoperative pediatric MR brain datasets have been fused with postoperative brain CT scans. At Dartmouth-Hitchcock Medical Center (Lebanon, NH), a variety of fusion imaging is performed with mobile PET, CT, MR, and, most recently, dual-modality PET/CT.

Much has been written about dual-modality scanners, which provide combinations of PET/CT and SPECT/CT and benefit from consistent patient positioning between scans. ^2,3 PET/CT scanners have the advantage of using the CT data for attenuation correction, thus reducing the PET scanning time by eliminating transmission image acquisition. ² Even so, this does not completely eliminate movement (both physiological and voluntary) during the PET or SPECT study, and the publications discussing modified breathing protocols to minimize this are an indication that effort still has to be made to ensure reasonable alignment between scans. ^4,5 Slight but noticeable misalignments are not uncommon with hybrid PET/CT scanners at the lung bases or in and around the heart-areas that typically have a large amount of motion.

Sophisticated software fusion, such as that provided by Mirada Solutions/CTI Reveal MVS (Oxford, UK; Knoxville, TN) is able to compensate for this movement and, if necessary, "deform" the image data to provide a better match. It also has the ability to fuse any combination of CT/MR/PET/SPECT. Software further complements dual-modality scanners by extending the fusion possibilities to the analysis and fusion of multiple datasets over time and quantification of the degree of change between these studies.

Benefits of fusion

Providing structural and functional information in the same image

Various authors have demonstrated the clinical utility of ¹⁸ F-FDG-PET imaging in lymphoma staging ^6,7 and monitoring of therapeutic response, ⁸ as well as in colonic carcinoma staging. ⁹ Melanoma has an extremely high level of FDG uptake, providing an excellent target-to-background ratio. ¹⁰ In the more controversial area of breast cancer, ¹⁸ F-FDG-PET has the advantage of being independent of breast density, ⁷ something yet to be fully exploited in medical practice. ¹¹ In each of these clinical situations, fusion provides the tool to anatomically localize the areas of interest seen on the PET scan. In some circumstances, it is the absence of FDG uptake that is important (Figure 1).

Yet we also know that ¹⁸ F-FDG is not tumor-specific and will accumulate in any benign disease with increased glucose metabolism (such as pseudomembranous colitis, acute enterocolitis ¹² and sarcoidosis ^13,14 ). Fusion can improve the specificity of ¹⁸ F-FDG-PET ¹⁵ through correlation with anatomical structure and/or abnormalities present in the CT or MR study (Figure 2). The same can be also be said for SPECT/CT fusion. ^16,17

Fusion also has utility in the investigation and management of nontumorous disease processes (such as Parkinson's disease, ¹⁸ Alzheimer's disease, ¹⁹ and epilepsy) (Figure 3). More specific radiopharmaceuticals, such as FDDNP (2-(1-{6-[2-[ ¹⁸ F]Fluoroethyl) (methyl) amino]-2-naphthyl} ethylidine)malononitrile) for imaging plaques and neurofibrillary tangles in Alzheimer's disease, increase the need for fusion with CT or MR to provide an anatomical context for data interpretation.

Improving reading efficiency

Fusion involves two components: registration (alignment or matching) of images and visualization of the fused data. Without the latter, there is no appreciation of the former. Sophisticated fusion software provides different methods of visualizing the data with some overlap into the functionality provided by 3D visualization software.

The improved resolution of imaging hardware and the range of possible imaging modalities have resulted in a clinician being faced with an increasing volume of image data to assess. Fusion is one way of managing and presenting this data in a manner that represents a synthesis of all the available data, thus improving the efficiency of assessment.

Improving confidence in diagnosis when one modality alone is not definitive

The premise that fusion solves a diagnostically challenging case is inaccurate. The clinician remains responsible for the final diagnoses. Fusion is only a tool to facilitate this process-it does so not by creating new evidence, but rather through presenting existing evidence in a manner that can aid in diagnostic interpretation.

Fusion can improve the confidence of the diagnostician by utilizing all the patient studies in combination rather than in isolation. A subtle distortion in soft tissue seen on a CT becomes more significant when coupled with increased FDG uptake in the same area. An abnormality seen with MR can be determined to be necrotic tissue when fused with a PET study (Figure 1). With fusion, radiologists are finding that they are able to review previous studies and detect abnormalities that were missed on initial appraisal.

Quanti fi cation of the difference between scans

Mirada Solutions/CTI Reveal MVS software has quality-control tools that allow the user to visualize the regions where there has been the greatest amount of change (deformation) following a fusion process. These are indicated as either a grid or a colormap overlay (Figure 4). The cursor provides more detailed information with the displacement vector of each point.

Although initially intended as a means of quality control, this is also an effective way of quantifying how different two image datasets are, and it has potential utility as a means of assessing how well a patient has managed to keep still (Figure 5). This information may temper a clinician's diagnostic judgment.

New applications

Fusion imaging blurs the boundaries between the different medical specialties. Nuclear medicine physicians are now able to see their SPECT studies in the context of a CT or MR study with greater anatomical detail. Fusion between 2D and 3D modalities is also possible. ²⁰ Surgeons and physicians can view all of the imaging studies they have ordered in a more holistic manner. Radiologists can extend their practice into functional imaging studies. This is threatening to some, but this improved utilization of patient data will ultimately benefit quality of care.

Radiation therapy planning

It has been recognized that a CT scan alone underestimates the true margins of a tumor ^21,22 and that using FDG-PET/CT fused images can alter radiotherapy volumes between 26% ²³ to 49% ²⁴ of the time. Fusion of CT and MR images has also been used to improve lesion delineation. ²⁵

A fused dataset, therefore, has value in defining the boundaries of a tumor more accurately than a single imaging study alone. Some radiation therapy planning (RTP) software provides software fusion, albeit in a simplistic fashion. A better approach is for the initial tumor contouring to be performed within dedicated fusion software, which is then exported as a DICOM RT structure set to a dedicated planning system for refinement. Often it is difficult for the radiotherapist to determine the significant lesions. In such a situation, a radiologist using fusion software could contour and send an initial volume of interest for the radiotherapist to refine.

Longitudinal fusion and fusion of multiple datasets

It is often desirable to fuse image studies taken at two different times to determine if there has been significant change in pathology-for example, following chemotherapy. A software-based approach to fusion has the scope to handle image studies acquired at different times and compensate for differences due to patient movement and changes in shape. It is only a slightly more advanced step to be able to fuse PET/CT to PET/CT studies, functionality that further builds upon the complementary nature of software fusion with hardware fusion solutions.

This leads to the idea of fusing multiple datasets, the best example being the use of serial PET/CT studies to track tumor regression in response to therapy. It would be conceivable that fusion of three datasets would be clinically useful. For example, a pre- and postcontrast MR breast study with PET; or a PET/CT study with an earlier PET study.

Quantification of therapy response

While not being a direct feature of fusion, quantification tools impart a degree of objectivity to observations in oncology and disease management. Simpler quantitative measures include metrics for user-defined volumes, but there are many more. To be able to track and quantify the size of multiple lesions in the body over time has great value as a measure of treatment efficacy. There is insufficient scope to discuss this area in greater detail here other than to assert that quantitative measures lend objective meaning to visualized pathology and also facilitate the introduction of computer-aided detection.

The value of fusion in pharmaceutical drug trials

The impact of longitudinal fusion combined with quantification tools for in vivo monitoring of molecular expression or tumor response has not gone unnoticed by astute pharmaceutical companies. These tools have considerable utility in drug trials in both animals and humans.

Conclusion

The principles of fusion have existed for many years, and the application of this technique is not confined to medical imaging. The military has used fusion of infrared with visible light images to detect camouflaged targets. The daily weather report comprises a satellite image fused with a geographical map. Software fusion is like glue, drawing together images from multiple devices, including dual-modality scanners and PACS systems and, now, even radiotherapy planning systems. The relationship between software and hardware fusion will become more intertwined in the future, and radiological practices will only benefit from this.

There are, of course, limits to what is possible with any fusion methodology. Extreme differences between image studies will always create problems. Use of deformable (nonlinear) techniques introduce a degree of flexibility and extend the range of possibilities; but the fact remains that the quality of fusion imaging is dependent upon the quality of the data being fused. We should remember that fusion imaging is not infallible, and neither are the underlying imaging modalities.

Nonetheless, this is an exciting tool for the diagnostician and the researcher. As more realize the role of PET as the method of in vivo molecular imaging, ^26,27 it is likely that more possibilities of fusion imaging will be discovered.