| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Gastrointestinal Imaging |
1 From the Division of Abdominal Imaging, Department of Radiology, University of Pittsburgh Medical Center, 200 Lothrop St, Pittsburgh, PA 15213. Received February 28, 2000; revision requested April 10; final revision received December 21; accepted January 4, 2001. Address correspondence to M.P.F. (e-mail: federle@pitt.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Radiologists reviewed multiphase helical CT liver images in 86 patients with 37 hemangiomas and 49 malignant liver tumors. They evaluated lesion type and degree of enhancement for change from arterial to portal venous phase. They rated their confidence in the discrimination of hemangiomas from malignant tumors.
RESULTS: At arterial phase CT, enhancement similar to aortic enhancement was observed in 19%–32% of hemangiomas and 0%–2% of malignant tumors; globular enhancement, in 62%–68% and 4%–12%, respectively. At portal venous phase CT, enhancement similar to blood pool enhancement was observed in 43%–54% of hemangiomas and 4%–14% of malignant tumors; globular enhancement, in 46%–49% and 0%–2%, respectively. For all readers and all phases of enhancement, the area under the receiver operating characteristic curves was 0.81–0.87, indicating that inherent accuracy of CT is high and that there was no significant difference (P > .28) in overall accuracy. Readers diagnosed hemangiomas with 47%–53% mean sensitivity with all enhancement phases and diagnosed malignant lesions with 95% mean specificity.
CONCLUSION: Small hemangiomas frequently show atypical appearances at CT. Two-phase helical CT does not improve sensitivity but does improve specificity for differentiating hemangiomas from hypervascular malignant tumors.
Index terms: Angioma, 761.3194 • Liver, CT, 761.12112, 761.12114, 761.12115, 761.12119 • Liver neoplasms, 761.32, 761.33 • Neoplasms, metastases, 761.33
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Most cavernous hemangiomas are easily distinguished from malignant hepatic tumors due to characteristic features, such as near isoattenuation with blood on nonenhanced images and globular or nodular peripheral enhancement similar to attenuation of blood vessels, at rapid CT with bolus administration of contrast material. However, diagnosis of small hemangiomas is more difficult because they may not demonstrate nodular enhancement but often enhance homogeneously during the hepatic arterial or portal venous dominant phase of a multiphase helical CT examination (7–10). These features may simulate those of other hypervascular liver tumors, including hepatocellular carcinoma (HCC) or metastases. Because hemangiomas are encountered frequently, distinction from hepatic malignancy is an important and common challenge.
The goals of our investigation were (a) to determine the CT characteristics of a large number of small hemangiomas at nonenhanced and dual-phase helical CT and (b) to evaluate the use of CT to distinguish between small hemangiomas and small hypervascular malignant liver tumors.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
One experienced abdominal radiologist (T.K.), who served as the study coordinator, reviewed clinical, radiologic, and histopathologic reports and follow-up imaging studies and identified 86 patients (40 women, 46 men; age range, 23–78 years; mean age, 57 years) who satisfied the inclusion criteria. One hepatic lesion, which was depicted on at least one CT imaging level, in each of 86 patients was then included in this study. If a patient had multiple such lesions, the largest one was included. The included hepatic lesions were hemangiomas (n = 37), HCCs (n = 28), and metastases from renal cell carcinoma (n = 10), neuroendocrine tumor (n = 6), breast carcinoma (n = 2), leiomyosarcoma (n = 1), angiosarcoma (n = 1), and thyroid carcinoma (n = 1). The mean diameter of the lesions was 1.9 cm (range, 0.5–2.9 cm) for the hemangiomas and 1.8 cm (range, 1.0–2.9 cm) for the malignancies.
Diagnosis of hemangioma was established by means of histologic findings in specimens obtained at percutaneous biopsy (n = 3), open biopsy (n = 1), and transplantation (n = 3). It was also determined by means of typical findings at ultrasonography, magnetic resonance (MR) imaging, including dynamic contrast material–enhanced imaging or scintigraphy with the use of technetium 99m–labeled red blood cells, with no change in the size of the lesions for 6 months or more at serial imaging (n = 30).
Diagnosis of HCC was proved at histologic examination of specimens obtained at percutaneous biopsy (n = 17) or at liver transplantation (n = 11). Diagnosis of metastases was determined with percutaneous biopsy (n = 7) or surgery (n = 2) or with evidence of definite lesion growth on serial cross-sectional radiologic images obtained during a period of 6 months or less in patients with a known primary tumor (n = 12). In patients with multiple hepatic lesions, we required proof, as detailed before, of the nature of the specific lesion that was analyzed by the CT readers (including M.P.F., M.S.P., Y.K.) in this study. No patients underwent chemotherapy or other treatment for the hepatic lesions prior to CT evaluation.
CT Examination
All examinations were performed with a helical CT scanner (HiSpeed Advantage; GE Medical Systems, Milwaukee, Wis). All nonenhanced scans were obtained incrementally with a 2-second scanning time and 5-mm collimation at 8-mm intervals. Nonhelical scans were used to avoid excessive tube heating that might cause tube-cooling delays during the arterial and portal venous phases of the studies. The arterial and portal venous phase images were obtained with the helical technique with 7-mm collimation and a table pitch of 1:1 to 1:1.5. A window width of 150 HU was used for the nonenhanced images, whereas a window width of 350 HU was used for the arterial and portal venous phase images.
For the contrast-enhanced portions of the examinations, each patient received 150 mL of either iothalamate meglumine (Conray-60; Mallinckrodt Medical, St Louis, Mo) or ioversol (Optiray 320; Mallinckrodt Medical) intravenously by means of a mechanical power injector (model OP 100; Medrad, Pittsburgh, Pa). Contrast material was injected at rates that ranged from 2.5 to 5.0 mL/sec, with a monophasic rate of injection. The slower injection rates were often used in patients with poor intravenous access or who had central venous catheters that were not rated for injection rates greater than 3 mL/sec.
After initiating infusion of contrast material, we used scanning delays of 20 seconds (for 4.0–5.0 mL/sec injections) or 28 seconds (for 2.5–3.5 mL/sec injections) before starting to obtain the arterial phase images. Imaging in the portal venous phase was performed after a scanning delay of 60 seconds (4.0–5.0 mL/sec rate) or 70 seconds (2.5–3.5 mL/sec rate) from initiation of infusion of contrast material. This variation in the length of scanning delays is necessitated by the known relationship among peak hepatic parenchymal enhancement, time to equilibrium, and the rate of infusion of contrast material (11–13).
Image Analysis
Images were evaluated retrospectively and independently by three readers other than the study coordinator, who were experienced abdominal radiologists and who were blinded to the diagnosis, any patient information, and any correlative imaging studies. They were shown one image each obtained at nonenhanced, arterial phase, and portal venous phase CT in each patient in whom one hepatic lesion was demonstrated. On the images, black film masks excluded all but the minimum area, including the hepatic lesion, hepatic veins, portal vein, inferior vena cava, and aorta, to minimize reviewer bias.
Each reader evaluated the nonenhanced, arterial phase, and portal venous phase CT images of each lesion separately, with each reading separated by 1 week. Then the combination of arterial and portal venous phase images was evaluated by each radiologist. Finally, the combination of nonenhanced, arterial, and portal venous phase images was interpreted. Each of these readings was separated by 1 week.
Each lesion was characterized by its appearance at CT. The attenuation of the lesion was judged subjectively as being similar or not similar to that of blood vessels on nonenhanced images. (Region-of-interest calculations were not made or recorded.) On arterial and portal venous phase images, the degree and pattern of enhancement were evaluated for any enhanced portion of the lesion. The degree of enhancement was judged relative to that of the aorta on the arterial phase images (similar to that of the aorta, substantial but less than that of the aorta, minimal, or none).
On the portal venous phase images, the degree of enhancement was judged relative to that of the blood pool and liver (similar to that of the blood pool, greater than that of liver parenchyma but less than that of the blood pool, similar to that of the liver parenchyma, or less than that of liver parenchyma).
The pattern of enhancement was also evaluated on arterial and portal venous phase images (globular or nodular, diffuse homogeneous, diffuse heterogeneous, ring, or no enhancement). A change in the degree or pattern of enhancement between the arterial and portal venous phases was also evaluated and characterized as expansion, washout, or none. Expansion was defined by an increasing area of enhancement within the lesion. Washout was defined as a change from hyperattenuation relative to the liver at arterial phase CT to iso- or hypoattenuation relative to the liver at portal venous phase CT.
Finally, each reader characterized each lesion as a hemangioma or a malignant lesion by using his own assessment of its CT appearance. Each reader used a five-point scale to assign a confidence level to his evaluation (1, definitely malignant; 2, probably malignant; 3, equivocal; 4, probably hemangioma; 5, definitely hemangioma). The same evaluation was conducted by using arterial phase images alone; portal venous phase images alone; a combination of arterial and portal venous phase images; and a combination of nonenhanced, arterial, and portal venous phase images.
To assess interobserver variability in the evaluation of the CT characteristics of the lesions and assignment of a confidence level to the lesion status, the unweighted 
value was calculated (14). The level of agreement was defined as follows: values of less than 0 indicated no agreement, values of 0–0.40 indicated poor agreement, values of 0.41–0.75 represented good agreement, and values of 0.76–1.00 represented excellent agreement.
Receiver operating characteristic curves were calculated to assist in comparing the results of readings of the arterial phase images alone, of combinations of arterial phase and portal venous phase images, and of combinations of nonenhanced, arterial phase, and portal venous phase images with each other (15). True-positive cases were defined as hemangiomas that were correctly diagnosed. False-positive cases were defined as malignant lesions that were incorrectly diagnosed as hemangiomas. The diagnostic capability was determined by calculating the area under each reader-specific receiver operating characteristic curve (Az).
Results were expressed as means ± 1 SD. The Az values were then compared with each other by using a method proposed by Obuchowski and Rockette (15) for the review of the following: arterial phase images alone; portal venous phase images alone; a combination of arterial phase and portal venous phase images; and a combination of nonenhanced, arterial phase, and portal venous phase images. This method was used to compare the Az for different modalities in a multireader study while considering the correlations resulting because all readers reviewed the same cases.
Confidence-level ratings of the images were also used to calculate the sensitivity and specificity for each observer in the diagnosis of hemangioma with nonenhanced and two-phase helical CT. Ratings of 1 or 2 indicated a reading of a malignant lesion; ratings of 4 or 5 indicated a reading of a hemangioma. Ratings of 3 (equivocal) were considered to indicate incorrect readings.
Sensitivity, specificity, and accuracy were calculated as follows: Sensitivity equals (the number of correct diagnoses of hemangioma divided by the number of proved hemangiomas) multiplied by 100. Specificity equals (the number of correct diagnoses of malignancy divided by the number of proved malignancies) multiplied by 100. Accuracy equals (the number of correct diagnoses of hemangiomas plus the number of correct diagnoses of malignant lesions) divided by the number of proved malignancies and hemangiomas) multiplied by 100. The sensitivity and specificity for reviewing arterial phase images alone, portal venous phase images alone, and a combination of arterial phase and portal venous phase images were compared with each other for the group of three readers by using the test for a marginal homogeneity (16).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
values for interobserver variability in judging the appearance (CT characteristics) of the lesions were 0.48 for the attenuation at nonenhanced CT, 0.63 for the pattern and 0.47 for the degree of enhancement at arterial phase CT, 0.51 for the pattern and 0.49 for the degree of enhancement at portal venous phase CT, and 0.48 for the change of enhancement between arterial and portal venous phase CT.
The unweighted values in the evaluation of the appearance of the lesions indicated good agreement. Weighted values for the interobserver variability in confidence ratings were 0.7 for reviewing arterial phase CT images alone; 0.72 for reviewing portal venous phase CT images; 0.73 for reviewing arterial and portal venous phase CT images together; and 0.66 for reviewing nonenhanced, arterial phase, and portal venous phase CT images together. The weighted values in the assignment of a confidence level of diagnosis indicated good to excellent agreement between readers.
The CT characteristics of the lesions evaluated by the readers are summarized in Tables 1–4. At nonenhanced CT, the attenuation of 38%–49% of the hemangiomas was judged to be similar to that of the blood pool, while the attenuation of only 14%–20% of the malignant lesions was judged to be similar to that of the blood pool (Table 1). However, the attenuation of 27%–38% of the hemangiomas was judged not to be similar to that of the blood pool (Table 1).
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Van Leeuwen et al (7) reported on characterization of focal liver lesions with triphasic (arterial, portal, and equilibrium phase) helical CT, and they mainly evaluated the degree but not the pattern of enhancement. They reported that 34 (83%) of 41 of the hemangiomas that were smaller than 3 cm in diameter showed isoattenuation compared with the arterial system in all three phases of enhanced scanning.
In our results, only 19%–32% of the small hemangiomas showed isoattenuating enhancement relative to the aorta at arterial phase helical CT, and only 43%–54% showed isoattenuating enhancement relative to the blood pool at portal venous phase CT. The explanation for the discrepancy between our results and those of Van Leeuwen et al (7) is not certain but may be due to the fact that our readers were blinded to all information and were able to evaluate only a single hepatic lesion in each case. Our results showed that lesion enhancement similar to that of the aorta at arterial phase CT is extremely specific but not sensitive for differentiating small hemangiomas from small hypervascular malignant liver tumors.
It has been reported (20–22) that globular enhancement, isoattenuating to the aorta or blood pool at single-pass CT, is a sensitive and specific sign for hemangiomas. In the results of Leslie et al (22), globular enhancement was observed in 88%, and enhancement to the aorta was observed in 52%–72% of hemangiomas. However, in our investigation, these helpful CT findings were much less frequent. The study of Leslie et al (22) included larger hemangiomas (the size range of the hemangiomas was 2–18 cm [mean, 3.8 cm]), while our study included only lesions that were smaller than 3 cm. These facts indicate that small hemangiomas less frequently show globular and isoattenuating enhancement, relative to the aorta at arterial phase or portal venous phase helical CT; these are typical enhancement patterns for hemangiomas.
The effect of partial volume averaging may also play a role in obscuring these characteristic findings for small hemangiomas. The advent of multi–detector row helical CT may allow more accurate characterization of small hepatic masses, because this technology allows routine acquisition of thin sections (eg, 2.5 mm) through the liver during the optimal phases of contrast enhancement (24,25).
Variations in enhancement patterns and partial volume averaging may also constitute problems in the MR diagnosis of small hemangiomas. Semelka et al (26) found uniform enhancement at 1 second in almost half of hemangiomas smaller than 1.5 cm in diameter.
Hanafusa et al (9) also observed that small hemangiomas are likely to demonstrate homogeneous enhancement, which they observed in 15 (38%) of 40 hemangiomas smaller than 3 cm in diameter during the first phase of a two-phase dynamic incremental (nonhelical) CT examination.
In our study, homogeneous enhancement was seen in 11%–14% of hemangiomas and 24%–37% of malignant lesions during the arterial phase. We were surprised to find some hemangiomas and hypervascular malignancies that showed no apparent enhancement on arterial phase images. We believe there is a logical technical explanation for both observations, with important clinical implications. By using a nonhelical scanner, most of the subjects of Hanafusa et al (9) demonstrated homogeneous and/or progressive enhancement of hemangiomas. By using a faster helical scanner, a higher percentage of our patients demonstrated no enhancement (3%–11%) or globular enhancement (62%–68%) of hemangiomas because we were able to obtain a series of thin sections through the lesions before they had enhanced completely. On the basis of our experience in this and other investigations of helical CT evaluation of liver tumors, we believe that our scanning protocols required modification. The scanning delays that we used (20 seconds for an injection rate of 4.0 or 5.0 mL/sec; 28 seconds for rates of 2.5–3.5 mL/sec) were probably insufficient for optimal identification and characterization of hepatic hypervascular tumors.
Current investigations demonstrate that for most patients hypervascular tumors are best demonstrated on CT images obtained by using a scanning delay of approximately 35 seconds. Optimal timing is best determined with a test bolus of contrast material and bolus timing, with the acquisition of arterial phase images occurring during early portal venous opacification but before hepatic venous opacification (27,28).
Progressive enhancement starting peripherally has long been identified as a characteristic feature of hemangiomas (9,18,19), but prior researchers have not studied small lesions with helical dual-phase CT imaging. While 46%–59% of our hemangiomas were judged as demonstrating progressive expansion, 32%–38% demonstrated no visible change in enhancement. We can only speculate that some of these hemangiomas were thrombosed or had such reduced blood flow that the 40-second interval that we allowed between the arterial and portal venous phases was insufficient to allow obvious enhancement.
At enhanced CT, the attenuation of hemangiomas relative to circulating blood or liver parenchyma has been cited as a diagnostic criterion (10,18,29). We do not consider the liver to be a useful reference, since its attenuation relative to that of blood or hemangiomas is highly dependent on numerous variables, including the presence of excessive fat, glycogen, or iron in the liver; anemia; and other factors. We anticipated that small hemangiomas should be nearly isoattenuating to the blood pool on nonenhanced scans, and we confirmed this in most cases, in which 38%–49% of hemangiomas were isoattenuating and an additional 19%–27% were not visible (Table 1). However, 27%–38% were judged not similar to vascular attenuation.
We may have improved our perception of isoattenuation if we had relied on objective region-of-interest determinations rather than on subjective visual assessment. It is also likely that partial volume averaging artificially increased the attenuation of small hemangiomas in some cases. Nevertheless, isoattenuation relative to blood vessels on nonenhanced images was not an accurate or useful criterion in distinguishing small hemangiomas in this study.
The receiver operating characteristic analysis in our study demonstrates that the use of the combination of arterial and portal venous phase images improves the specificity, if not the sensitivity, in distinguishing small hemangiomas from malignancies. By using the combined phases, the correct diagnosis of a malignancy (specificity) was achieved in 95% of cases, while the correct diagnosis of hemangioma (sensitivity) was achieved in only 47%–53% of cases.
When we detect globular enhancement—with enhanced areas being similar to those of the aorta at the arterial phase and to those of the blood pool at the portal venous phase—and expansion of enhancement, hemangioma can be diagnosed with great confidence. Conversely, a lesion that enhances homogeneously but less than the aorta and subsequently shows washout to become hypoattenuating to the liver and the aorta is much more likely to represent a malignant lesion than a hemangioma.
The results of our investigation may represent a worst-case scenario because we challenged our CT readers to distinguish hemangioma from malignancy while withholding all clinical patient information (such as the presence of cirrhotic hepatic morphology, other hepatic lesions, and interval enlargement of hepatic lesions) and even blinding them to other information that would be available to radiologists interpreting CT scans. In spite of these handicaps, multiphase hepatic CT allowed confident diagnosis of small hypervascular malignancies. Confident diagnosis of small hemangiomas remains difficult and may require improvements in our CT techniques or diagnostic criteria or confirmatory findings of additional studies, such as MR imaging or biopsy.
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, T.K., R.L.B.; study concepts, T.K., R.L.B.; study design, T.K., R.L.B., M.P.F.; literature research, T.K.; clinical studies, all authors; data acquisition, T.K.; data analysis/interpretation, T.K., R.L.B., M.P.F.; statistical analysis, T.K.; manuscript preparation and definition of intellectual content, T.K., M.P.F., R.L.B.; manuscript editing and revision/review, M.P.F.; manuscript final version approval, all authors.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  H. Onishi, T. Murakami, T. Kim, M. Hori, R. Iannaccone, M. Kuwabara, H. Abe, S. Nakata, K. Osuga, K. Tomoda, et al. Hepatic Metastases: Detection with Multi-Detector Row CT, SPIO-enhanced MR Imaging, and Both Techniques Combined Radiology, April 1, 2006; 239(1): 131 - 138. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. H. Byun, T. K. Kim, C. W. Lee, J. K. Lee, A. Y. Kim, P. N. Kim, H. K. Ha, and M.-G. Lee Arterioportal Shunt: Prevalence in Small Hemangiomas versus That in Hepatocellular Carcinomas 3 cm or Smaller at Two-Phase Helical CT Radiology, August 1, 2004; 232(2): 354 - 360. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. A. Teefey, C. C. Hildeboldt, F. Dehdashti, B. A. Siegel, M. G. Peters, J. P. Heiken, J. J. Brown, E. G. McFarland, W. D. Middleton, D. M. Balfe, et al. Detection of Primary Hepatic Malignancy in Liver Transplant Candidates: Prospective Comparison of CT, MR Imaging, US, and PET Radiology, February 1, 2003; 226(2): 533 - 542. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H.-J. Jang, T. K. Kim, H. K. Lim, S. J. Park, J. S. Sim, H. Y. Kim, and J.-H. Lee Hepatic Hemangioma: Atypical Appearances on CT, MR Imaging, and Sonography Am. J. Roentgenol., January 1, 2003; 180(1): 135 - 141. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
