| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | RSS | TABLE OF CONTENTS |
|
|
|
||||||||
|
||||||||||
Clinical Investigations |
1 Charité, Universitätsmedizin Berlin, Campus Virchow-Klinikum, Klinik für Strahlenheilkunde, Berlin, Germany
2 Ernst-Moritz-Arndt-Universität Greifswald, Institut für Diagnostische Radiologie, Greifswald, Germany
3 Charité, Universitätsmedizin Berlin, Campus Virchow-Klinikum, Klinik für Gastroenterologie, Berlin, Germany
4 Department of Neurosurgery, Image Guidance Laboratories, Stanford University, Stanford, California
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
Key Words: PET • CT, spiral • image manipulation or reconstruction • pancreas • computer applications, detection
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
The primary task of diagnostic imaging is detecting any pancreatic lesion and differentiating between malignant and benign (e.g., inflammatory) changes in the pancreas. Furthermore, imaging should ideally be able to permit staging of the tumor. In the case of pancreatic cancer, any infiltration of the vessels and lymph nodes as well as possible distant metastases take on special importance due to their impact on the assessment of resectability of the tumor or the decision to initiate chemotherapy (2).
CT and MRI only provide morphologic information and thus fail staging in several pancreatic tumors. Since it permits the evaluation of a possible malignancy with a fairly high degree of sensitivity, the use of PET increased steadily over the past years. Limitations of PET are unspecific tracer uptake in inflammatory changes such as acute pancreatitis as well as the lack of anatomic information in these functional images. Quite often the necessary spatial correlation of a hot spot with anatomic structures is severely constrained. Digital fusion of PET with CT would open up the possibility of aligning the functional PET information with the spatial data provided by CT (3).
The objective of this study was to introduce retrospective PET/CT image fusion into clinical practice to improve the diagnosis of pancreatic cancer and to evaluate its impact on the detection, differentiation, and staging of such malignancies.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
CT Studies
The CT studies were performed using a triple-phase technique with a quadruple-line multislice CT scanner (Somatom Plus 4 Volume Zoom; Siemens AG). Patients were positioned in prone position with their arms raised above the head to avoid beam-hardening artifacts. To ensure similarity of CT and PET data, CT scans were performed in a respiratory midposition and without gastric filling. After planning the investigation on an anteroposterior scout radiograph, a noncontrast study of the upper abdomen was acquired (120 kV; 165 mA; collimation, 4 x 5 mm; table feed, 20 mm; pitch, 1; reconstruction interval, 8 mm). Eighteen seconds after initiating the intravenous administration of 100 mL iopromide (Ultravist 370; Schering AG), a contrast-enhanced acquisition was started during both the arterial phase (collimation, 4 x 1 mm; table feed, 4 mm; pitch, 1; reconstruction interval, 5 mm) and the venous phase 80 s after beginning with the intravenous administration of the contrast agent (collimation, 4 x 1 mm; table feed, 4 mm; pitch, 1; reconstruction interval, 8 mm). Image fusion was based on the arterial phase.
PET Scans
The PET scans were performed with an ECAT EXACT 47 scanner (Siemens AG). The blood glucose level of each patient was determined before he or she underwent PET scanning. Patients whose blood glucose level exceeded 110 mg/dL were excluded from the study. Scans were performed in prone position and no breathing commands were given throughout the scan. Intravenous administration of 5 MBq 18F-FDG per kilogram of body weight was followed by a 2-dimensional (2D) whole-body PET scan with integrated transmission measurement and iterative image reconstruction. An uptake period of 60–90 min was adhered to before the actual image acquisition. Contrary to the CT studies, the patients kept their arms alongside the body. Glucose uptake was quantified by the standard uptake value (SUV), with an SUV of >3.5 considered as indicative of malignancy (4). For definite assessment of a pancreatic mass, both the SUV value and visual aspects were considered. After inclusion of the last patient, a receiver operating characteristic (ROC) analysis of SUV values was performed to verify the cutoff point between benign and malignant lesions.
Image Fusion
Image fusion was performed with a software program based on the commercially available AVS Express (Advanced Visualization Systems) package, extended by custom-written software modules as part of this study. All calculations were run on an O2-workstation (Silicon Graphics). For this, the CT and PET images were transferred onto the workstation as DICOM (Digital Imaging and Communications in Medicine) images via the Intranet. The CT images were registered to PET images using a fully automatic voxel-based algorithm. Our method uses an independent implementation (5) of an algorithm for multiresolution optimization of the well-proven normalized mutual information image similarity measure (6). After registration, the coregistered images were reconstructed and visualized in the coronal and axial planes to perform plausibility testing—that is, a physician reviews the images and checks them for misregistration visually. Alignment in all 3 planes was assessed by checking the body outline and the position of metabolically active organs (heart, liver, kidney, and urinary bladder). Any registration not regarded as plausible was rejected from further analysis.
Once registered, the software allows the multimodality data to be viewed in different ways, either as a fused tomogram or as a full-fledged 3-dimensional scene using operator-defined combinations of volume and surface rendering. The preferred mode of presentation for this study is 2D transparent image overlay. Each pixel in the foreground image (usually a color-encoded slice from the PET image) is assigned a level of transparency (
-value) based on its intensity. High intensities are shown opaque, whereas low intensities are transparent. The PET pixels with varying transparencies are then overlaid onto the corresponding CT slice, which is encoded as gray values. The fused images were printed on a color printer as transparencies or paper hard copies.
Two radiologists evaluated the original CT and PET images as well as the fused images in a randomized order in 3 different settings with an interval of 2 wk each. Using a standardized questionnaire, the following aspects were assessed and compared with the gold standard:
50% (7)) to assess resectability: superior mesenteric vein; confluence of the superior mesenteric, splenic, and portal vein; superior mesenteric artery; splenic artery; common hepatic artery; celiac axis (8–10) |
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
In the 100 patients evaluated, the final diagnosis of a malignant tumor was made in 64 patients (64.0%), whereas 36 patients (36.0%) were assessed as having a benign pancreatic disorder. Within the malignancies, adenocarcinoma of the pancreas was most prevalent (n = 57), whereas the benign lesions were dominated by chronic pancreatitis (n = 28) (Table 1).
|
PET detected 54 of the 64 malignancies and 22 of the 36 benign lesions (true-positive, n = 54; false-positive, n = 14; false-negative, n = 10; true-negative, n = 22). This results in a sensitivity of 84.4% and a specificity of 61.1%. The retrospective ROC analysis confirmed an SUV of 3.5 as the best cutoff point for differentiation between benign and malignant lesions (Fig. 1).
|
|
|
Infiltration of Adjacent Tissue
Forty-seven patients of the study population had complete histologic analysis (resection or exploratory surgery) and were suitable for evaluation of tumorous infiltration of adjacent tissue. Statistical parameters of CT and image fusion were assessed but not discussed due to the high prevalence of infiltration (94%). If a pancreatic tumor is detected, an infiltration must be assumed.
Lymph Node Metastases
Of the 31 patients with confirmed lymph node metastasis (diameter, 0.4–2.1 cm; mean, 1.2 cm), CT and PET were able to identify 8 and image fusion was able to identify 10 cases. Image fusion was able to detect additional lymph node metastasis in 2 cases. In one case, CT detected a lymph node with an enlarged diameter of almost 2 cm, which did not show up on PET at all. In another case, a distinct paraaortic focal tracer hot spot identified on PET scans was aligned on CT scans with an obviously normal lymph node (Fig. 4). The other lymph node metastases remained undetected because of their small size (Table 2).
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
There are encouraging reports in the literature for selected organ systems dealing with the benefit of image fusion relative to the individual modalities. Schaffler et al. fused CT and PET images in 19 patients with abdominal or pelvic malignancies and demonstrated an improvement in the topographic correlation of the lesion as well as in the differentiation between necrosis and active tumor (13). In general, the most important advantage of fusing CT and PET images is the firm correlation of tracer hot spots in PET with the anatomic structures visualized on CT. In the present study, image fusion increased the sensitivity of detecting cancer of the pancreas by about 5%. Since no investigations have been published with regard to the benefit of image fusion in pancreatic tumors, this emphasizes the pilot function of our study. When looking at the infiltration of adjacent tissue, the sensitivity of image fusion improved by 20% compared with CT, but at the cost of a decrease in specificity. The diagnosis of distant metastases with the help of image fusion reveals the problem that usually CT only covers the region of interest, whereas PET is performed as a whole-body scan. The future routine use of fast low-dose CT may allow whole body acquisition.
Accuracy of Image Fusion
Accuracy of image fusion has been stated well in early publications (14,15) but, as far as possible, it must be ensured that both modalities image the same region of the body with no change in its condition (i.e., filling of stomach and bowel, position of upper extremities). For instance, CT scanning of the pancreas is often performed as hydro-CT, while PET imaging is run without gastric loading to avoid gastric contractions, which otherwise would result in increased glucose levels within the gastric wall (16). In CT studies, the fluid-filled stomach displaces adjacent organs, which could lead to misregistration or misalignment of PET activity when superimposed on the structures imaged by CT. To avoid these problems in the current study, CT was performed without gastric filling. the position of the arms in CT and PET is standardized and could not be changed. For CT studies, the arms had to be raised above the head to avoid beam-hardening artifacts; in PET scanning with the scanner used, the arms are kept alongside the body because of the long acquisition time required. Additionally, in PET, blurred organ outlines occur due to respiratory motion during the acquisition period of up to 40 min. The latter problem can be ameliorated by shallow breathing during PET and respiratory midposition during CT. Aside from this, the newest PET generation allows markedly shorter examination time (average bed stop, about 8–10 min) with advanced reconstruction techniques. In the present study, meaningful registration proved to be impractical in 4 patients because of the differences between the images due to severe patient motion (Fig. 6); for image fusion, this resulted in a technical feasibility of 96.2%. The problem might be solved in the future by so-called nonrigid or elastic registration methods, which are capable of correcting for local organ motion or deformation between 2 image acquisitions. The current disadvantage of these still experimental procedures is the long time required by the computations (17), but the use of a shared-memory supercomputer offers the possibility of performing complex nonrigid registrations in only a few minutes (18).
|
CT
CT is a well-established modality in diagnostic imaging and staging of pancreatic cancer. The introduction of fast multidetector spiral CT has made it possible to study the pancreas during different perfusion phases of the organ. Some groups consider CT scanning during the arterial phase of the perfusion as being advantageous in the diagnostic workup of pancreatic cancer, since a large percentage of the cancer tissue is fibrotic, thus resulting in hypoperfusion of this region at that point in time. Some studies demonstrated an advantage for multiphase CT (19). Although performance of hydro-CT (oral contrast enhancement with 800–1,000 mL of water) has become the practice in some institutions due to a better delineation of the pancreas from the posterior gastric wall (20–22), it was not used in the current study to achieve similarity to PET.
In the 100 patients investigated in this study, CT detected malignancies with a sensitivity of 76.6% and a specificity of 63.9%. Thus, our results are much lower than those in other publications. For instance, Catalano et al. studied a total of 46 patients and reported a sensitivity of 97.0% and a specificity of 80.0% (23). Possible explanations are the size of the lesions in the study population and the degree of awareness of the investigators assessing the images. Several studies have demonstrated that CT cannot be relied on to detect lesions with a diameter of <2 cm (20,24–27). However, the primary objective in the diagnostic workup of the pancreas must be the early detection of small malignancies, something that present-day CT seems hard pressed to do. The studies published so far do not break down the lesions according to size, but the percentage of small tumors was substantial and the investigators assessing the lesions did not have any knowledge of the clinical picture. In the literature, the resectability of pancreatic cancer usually is assessed according to presence of infiltration of adjacent tissue or vessels, lymph nodes, and distant metastases. Quite often, these 3 parameters are lumped together, which renders any comparison of the studies more difficult. With regard to the sensitivity, specificity, and accuracy in the detection of vascular infiltration, Furukawa et al. reported values of 83.0%, 100.0%, and 89.0%, respectively (28). Lu et al. reported a sensitivity of 84.0% and a specificity of 98.0% (7), which was confirmed by Bluemke et al. (29). The percentages for infiltration of adjacent tissue given in older publications state an accuracy of 37.0%–45.0%, which is much closer to the results of the present study (30,31). Despite optimum technology, CT still faces the perplexing problem of insufficient contrast between the lesion and the surrounding tissue, particularly when dealing with small tumors. Although some publications demonstrate that combined arterial and venous phase CT scanning will detect even small lesions, the probability of detection seems to be rather small (20,24–26). The discouraging results of lymph node assessment can be explained due to their small diameter (average, 1.2 cm), which is next to the cutoff size (>1 cm) usually used to differentiate between benign and malignant lymph nodes.
PET
PET offers the possibility of studying quite sensitively glucose metabolism, which exhibits different degrees of elevation in inflammation and malignancy. Glucose uptake in the tissue can be measured quantitatively by the SUV. However, glucose uptake depends on numerous factors, which may lower the sensitivity and specificity of the PET study. Differentiation between inflammatory and malignant processes is based on the SUV. The limit of 3.5 chosen in the present study conforms to the values chosen by other groups and our own experience. On the other hand, other limits have also been reported in the literature—for example, Zimny et al. regard an SUV of 6.4 as dealing with a malignancy but accept a range of ±3.6 (32). In a study population of 106 patients, the mean SUV in the case of inflammatory changes is given as 3.4 ± 1.7. Berberat et al. reported a mean SUV of 3.09 for cancer and 0.87 for inflammation (33). The reasons for the difficulty in defining a clear-cut value may be due to actual overlapping of inflammatory and malignant changes in the pancreas but also may be due to differences in the study protocol.
Compared with the present study, other publications have reported better results for FDG PET detection of pancreatic cancer. Berberat et al. found a sensitivity of 94% and a specificity of 88.0% in their 80 patients, whereas Bares et al. reported a sensitivity of 92.0% and a specificity of 85.0% for their group of 40 patients (33,34). Unfortunately, no details are given for the study protocol, which makes comparison of the results more difficult. But there is the note that any false-negative result was diagnosed either in hyperglycemic patients or in those with early tumor stages. The study by Bares et al. primarily enrolled patients with enlarged pancreatic lesions suspected on CT as being malignant. This preselection could explain the excellent results. For their study population of 106 patients, Zimny et al. reported a sensitivity of 89.0% and a specificity of 53.0%, but they evaluated their data only visually (32). The large number of false-negative results was explained as being due to patients with an elevated blood glucose level. After exclusion of this group of patients from the study population, PET exhibited a sensitivity of 98.0%; from this it was concluded that the blood glucose level must be considered as well when evaluating PET. In the present study, the blood glucose level was already accounted for as part of the inclusion criteria. The large number of false-negative patients can be explained with the appearance of tumors in early stages with limited extent. There is no general agreement on the benefit of quantitative assessment by means of the SUV; for instance, Zimny et al. dispute the benefit of the SUV, whereas other authors consider the SUV a valid parameter (32,35). Zimny et al. reported data on PET detection in lymph node and distant metastases, with a correct result of 40.0% in lymph node metastases and 52.0% in distant metastases. Bares et al. compared PET with CT and ultrasound with regard to possible lymph node metastases and found a sensitivity of 76.0% for PET, 42.0% for CT, and 8.0% for ultrasound. A subsequent study of a larger number of patients lowered the value to 61.0% (34,36,37). The poor detection rate of lymph node metastases presumably has several causes. Most of the lymph node metastases that were found were located in the peripancreatic space and could not be differentiated from the pancreatic tumor. Additionally, as mentioned for CT, the mean size of infiltrated lymph nodes must be considered.
|
CONCLUSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
For correspondence or reprints contact: Arne-Jörn Lemke, MD, Universitätsklinikum Charité, Campus Virchow-Klinikum, Klinik für Strahlenheilkunde, Medizinische Fakultät der Humboldt-Universität zu Berlin, Augustenburger Platz 1, D-13353 Berlin, Germany.
E-mail: lemke{at}charite.de
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
Related articles in JNM:
This article has been cited by other articles:
|
|
 |
|
  P Ghaneh, E Costello, and J P Neoptolemos Biology and management of pancreatic cancer Postgrad. Med. J., September 1, 2008; 84(995): 478 - 497. [Full Text] [PDF]
|
|
|
|
 |
|
 C. von Forstner, J.-H. Egberts, O. Ammerpohl, D. Niedzielska, R. Buchert, P. Mikecz, U. Schumacher, K. Peldschus, G. Adam, C. Pilarsky, et al. Gene Expression Patterns and Tumor Uptake of 18F-FDG, 18F-FLT, and 18F-FEC in PET/MRI of an Orthotopic Mouse Xenotransplantation Model of Pancreatic Cancer J. Nucl. Med., August 1, 2008; 49(8): 1362 - 1370. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. Ghaneh, E. Costello, and J. P Neoptolemos Biology and management of pancreatic cancer Gut, August 1, 2007; 56(8): 1134 - 1152. [Full Text] [PDF]
|
|
|
|
|
|
A G Schreyer and R Kikinis Combined PET/CT colonography: is this the way forward? Gut, January 1, 2006; 55(1): 10 - 12. [Full Text] [PDF]
|
|
|
|
 |
|
 A. Rosset, L. Spadola, L. Pysher, and O. Ratib Informatics in Radiology (infoRAD): Navigating the Fifth Dimension: Innovative Interface for Multidimensional Multimodality Image Navigation RadioGraphics, January 1, 2006; 26(1): 299 - 308. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. J. To'o, S. S. Raman, N. C. Yu, Y. J. Kim, T. Crawford, B. M. Kadell, and D. S. K. Lu Pancreatic and Peripancreatic Diseases Mimicking Primary Pancreatic Neoplasia RadioGraphics, July 1, 2005; 25(4): 949 - 965. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M.-J. Ribeiro, P. De Lonlay, T. Delzescaux, N. Boddaert, F. Jaubert, S. Bourgeois, F. Dolle, C. Nihoul-Fekete, A. Syrota, and F. Brunelle Characterization of Hyperinsulinism in Infancy Assessed with PET and 18F-Fluoro-L-DOPA J. Nucl. Med., April 1, 2005; 46(4): 560 - 566. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | RSS | TABLE OF CONTENTS |
| JOURNAL OF NUCLEAR MEDICINE TECHNOLOGY | THE JOURNAL OF NUCLEAR MEDICINE |
