Multi-Modality Imaging

The conceptual design of dual-modality imaging systems is quite simple. A typical dual-modality imaging has a CT scanner (X-ray tube and detector) and a radionuclide detector (PET or SPECT) mounted on a single gantry with a common patient table for X-ray and radionuclide imaging. CT and radionuclide scans are acquired by translating the patient from one detector to the other while the patient remains on the patient table. This allows the CT and radionuclide images to be acquired with a consistent scanner geometry and body habitus, and with minimal delay between the two acquisitions. After both sets of images are acquired and reconstructed, image registration software then is used to fuse the X-ray and radionuclide images in a way that accounts for differences in scanner geometry and image format (e.g., 128_128 vs. 512_512) between the two data sets.
Over the past decade, prototype PET/CT and SPECT/CT systems have been developed by academic researchers, most notably at the University of Pittsburgh and at UCSF. More recently, commercial dual-modality imaging system have become available from the major medical imaging companies, including GE Medical System, Siemens Medical Systems and CTI Incorporated, and Philips/ADAC Medical Imaging Systems. Finally, dual-modality systems for imaging small animals are being developed in both academic and corporate settings.

At UCSF, researchers in the Department of Radiology, and specifically at its Physics Research Laboratory, pioneered the development of dual-modality imaging systems that combine X-ray CT and SPECT. The concept of dual-modality imaging at UCSF was conceived by Christopher Cann, PhD, Robert Gould, ScD, and Bruce Hasegawa, PhD, faculty members in the Physics Section of the Radiology Department at UCSF, and by Eric Gingold, PhD, a graduate student in Bioengineering at UCSF. Over the past decade, research with dual-modality imaging at UCSF involved theoretical modeling (Soo Chin Liew, PhD), computer simulations (Susan M. Reilly, MS), and development of the first prototype scanner (Thomas Lang, PhD). Implementation of a second prototype scanner included development of image reconstruction techniques (J. Keenan Brown, PhD), and of new application-specific integrated circuits and detector characterization experiments (Joseph Heanue, PhD). These developments led to quantitative measurements of radionuclide uptake in a porcine model of myocardial perfusion (Kathrin Kalki, PhD) with the prototype scanner.

The development of these prototype systems led to the implementation of the first human-scale dual-modality system for clinical research (Figure 1) at UCSF (Stephen Blankespoor, MS) in collaboration with GE Medical Systems. The system was configured by siting a scintillation camera (GE 600 XR/T) adjacent to a commercial CT scanner (GE 9800 Quick), allowing the patient to be scanned first in the CT scanner, then moved into the SPECT system by simple translation of a common patient table. The first studies with this newer system included quantitative radionuclide assessments of myocardial perfusion (Angela J. Da Silva, PhD), again in a porcine model, and followed by the development of techniques for assessing cancer (H. Roger Tang, PhD), and for diagnosis of prostate cancer patients with CT/SPECT (Kenneth H. Wong, PhD).

Additional experimental studies have included development of new detectors and detector electronics (Koji Iwata, PhD, William Barber, PhD).
Scientists at the UCSF Physics Research Laboratory now are developing a compact dual-modality system for imaging small animals (Andrew H. Hwang, BS, Anne E. Sakdinawat, BS) for biological studies at academic and corporate research centers. Important clinical expertise and scientific insight have been provided by Randall Hawkins, MD, PhD, David Price, MD, Gary Caputo, MD, and Mohan Ramaswamy, MD of the UCSF Nuclear Medicine Program. This concise history underscores the important contributions of many talented individuals in the Department of Radiology who have been responsible for the growth and development of dual-modality imaging research at UCSF.

Figure 1: A prototype CT/SPECT system was configured at the UCSF Physics Research Laboratory within the Department of Radiology. It combined a GE 600 XR/T SPECT system with a GE 9800 Quick CT scanner for correlated anatomical and functional imaging.

GE Medical Systems (Milwaukee, WI) and Elgems Ltd (Haifa, Israel) have developed a dual-modality imaging system ("Millenium VH" or "Hawkeye") that combines SPECT and CT, similar to the approach developed at UCSF. The system performs planar scintigraphy or SPECT, and also performs coincidence detection of annihilation photons for imaging 18F-fluorodeoxyglucose (FDG) and other positron- emitting radionuclides, and incorporates a low-resolution CT scanner for anatomical localization and attenuation correction of the radionuclide data.

The Hawkeye system now is installed in the Nuclear Medicine Clinic at Moffitt Long Hospital (Figure 2). It is interfaced both into the Radiology PACS system and into the Nuclear Medicine local area network. This makes it possible to view images from the Hawkeye directly on the GE Integra workstation next to the camera, or on an additional Integra workstation in the Nuclear Medicine view room, and also on PACS terminals and on the Nuclear Medicine Pegasys (Philips ADAC) workstations. The system is used for both clinical and research studies. It has been used for routine clinical procedures (Figure 3) such as bone, renal and cardiac studies as well as for selected FDG coincidence imaging procedures. Single slice CT images having a slice width of 10 mm can also be acquired with the unit, and it is expected that in the future most nuclear medicine clinical studies will benefit from this level of co-registered CT acquired with nuclear medicine image sets. Areas of particular relevance of CT to clinical interpretation of nuclear medicine images include bone and whole body imaging.

A very promising capability of the technology is the improved attenuation correction in cardiac imaging. Because of the improved signal quantification the unit produces through more accurate attenuation correction and image segmentation approaches, it will be of particular value in new research protocols such as new radioimmunotherapy approaches being developed together with investigators in Medical Oncology and Pediatric Oncology at UCSF. This promises to make in vivo dosimetry with nuclear medicine radiopharmaceuticals much more accurate than with more traditional methods.

We prospectively enrolled patients with known head and neck cancer who are eligible for tumor and lymph node resection for sequential two-day pre-operative PET-CT imaging with 18F-fluorodeoxyglucose (FDG) and 18F-fluorothymidine (FLT). The patient’s position was repeated by the use of a thermoplastic mask with the neck supported on a Timo cushion mounted on a carbon fiber board. We acquired images in list-mode using a Siemens Biograph 16 PET-CT scanner (Siemens Medical Solutions USA, Inc., Malvern, PA) and used software provided by the vendor to reframe them as two dynamic series: 0-10 minutes post-injection over the heart and 15-60 minutes post-injection over the head and neck. A whole body PETCT scan was then performed, followed by a high-count head and neck PET-CT image. We calculated maximum standardized uptake values (SUVs) over specific regions of interest (TrueD software on a syngoMMWP system, Siemens Medical Solutions USA, Inc.). For Patlak graphical analysis, we performed all reconstructions using the same FORE+2D-OSEM algorithm provided by the scannermanufacturer with four iterations and eight subsets. We generated an arterial input function using a left-ventricular region, derived-time activity curve for 0-10 minutes post-injection and measured activity from five venous blood samples thereafter up to one hour post-injection. In the case of FLT, the FLT metabolite fraction was corrected using solid phase extraction and separation. This input function was used to perform Patlak analysis with the dynamic head and neck dataset of 15-60 minutes post-injection. Patlak Ki (slope) images were generated using OASIS, IDL-based tools developed at the National Institutes of Health, Bethesda, MD.

We measured the direct correlation of FDG and FLT uptake within primary tumors and lymph nodes and correlated them against tumor presence, Ki-67 index (positive cells/total cells), and the nuclear-to-cytoplasmic ratio using data from post-surgical pathology. These correlations are intended to relate the biological activity and the probability of tumor involvement to measured values of FDG and FLT uptake in the tumor and lymph node. Probability data generated from the PET-CT data will be used to draw theoretical radiation treatment plans for comparison against the formal treatment planning method.

To date, our team has used FDG and FLT with PET-CT to evaluate three patients (Figure 2). All patients tolerated the protocol well. Figure 3 shows FDG and FLT PET images in one patient by both standard uptake methods and by Patlak analysis. Corresponding pathology in this patient demonstrated an increased Ki-67 index in both the primary tumor and reactive lymph nodes (Figure 4). Primary tumor, metastatic lymph nodes, and reactive lymph nodes all showed increased FDG and FLT uptake (Figure 5).

Additional data are needed to determine probability and proliferation thresholds for application to radiation treatment planning. However, given the overlap in uptake values between tumor and reactive lymph nodes, this finding underscores the importance of determining thresholds based on pathologic correlation to improve the accuracy of radiation treatment planning volumes.

The overall project aims to address a fundamental limitation of current quantitative measures of radiopharmaceutical localization on PET-CT that leads to potentially significant variability in cancer therapies. In considering the integration of biologic tumor volumes determined by PET-CT into a radiation treatment plan for patients with cancers of the head and neck, there is potentially a significant inter- and intra-user variability in the implementation of PET-CT in the treatment plan. This is due to a combination of inaccurate quantitative PET-CT data, the current disconnect between quantitative PET-CT data and biologically relevant information, and the inability to display data regarding multiple biological markers along with anatomical data from CT on a single interface for the radiation oncologist. The value that biologic tumor volumes could contribute to the field of radiation oncology is promising, but could be rapidly discounted if arbitrary methods of inaccurate reconstruction and radiopharmaceutical localization persist. Making quantitatively accurate and biologically meaningful measurements of certain biologic characteristics with PET on

Stephen L. Bacharach, PhD, is an adjunct professor of Radiology. Nick G. Costouros, MD, is a clinical fellow in the Nuclear Medicine section. Benjamin L. Franc, MD, is a nuclear medicine physician with Radiological Associates of Sacramento Medical Group. Bruce H. Hasegawa, PhD, was director of the Physics Research Laboratory. Randall A. Hawkins, MD, PhD, is a professor of Radiology and chief of the Nuclear Medicine section. Youngho Seo, PhD, is an assistant adjunct professor. Henry F. VanBrocklin, PhD, is a professor in residence and director of Radiopharmaceutical Research at the Center for Molecular and Functional Imaging.