| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Vascular and Interventional Radiology |
1 From the Department of Radiology, University of Texas Health Science Center at San Antonio, 7703 Floyd Curl Dr, San Antonio, TX 78284-7800. Received October 29, 1998; revision requested December 18; final revision received July 13, 1999; accepted July 30. Address reprint requests to S.C. (e-mail: chopra@uthscsa.edu).
 |
Abstract |
|---|
 TOP Abstract IntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
MATERIALS AND METHODS: Thirty-seven patients underwent 50 helical CT angiographic examinations and, within 2 weeks of each examination, portography with measurement of the portosystemic pressure gradient. Helical CT angiograms were independently interpreted by three radiologists who were blinded to the results of portography. Results of helical CT angiography and portography were compared. Sensitivity and specificity of helical CT angiography were separately calculated for the demonstration of morphologic abnormalities and the determination of their hemodynamic significance.
RESULTS: Of the 50 portograms, 31 (62%) demonstrated morphologic TIPS abnormalities, 24 (77%) with and seven (23%) without elevated portosystemic pressure gradients. Helical CT angiograms correctly demonstrated 30 (97%) of the 31 morphologic abnormalities and allowed correct diagnosis of 22 (92%) of the 24 hemodynamically significant abnormalities. Nineteen (38%) portograms were normal; helical CT angiograms correctly demonstrated the absence of abnormality in 17 (90%) of these cases. Sensitivity and specificity of helical CT angiography for all morphologic abnormalities were 97% and 89%, respectively, and, for hemodynamically significant abnormalities, 92% and 77%.
CONCLUSION: Helical CT angiography holds promise as a screening modality for the detection of TIPS stenoses or occlusions.
Index terms: Computed tomography (CT), angiography • Computed tomography (CT), comparative studies, 95.124, 95.12916 • Computed tomography (CT), helical, 957.12912, 957.12915, 957.12916 • Hypertension, portal, 957.711 • Portography, 95.124 • Shunts, portosystemic, 95.453, 957.1268 • Stents and prostheses, 957.1268
|
Introduction |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
We believe that, because of the prevailing difficulty in the performance of Doppler US of TIPS and uncertainty in the interpretation of findings, the development of a noninvasive test that is easier to perform and interpret is desirable. The use of helical computed tomographic (CT) angiography in the imaging of TIPS has been described (5). To our knowledge, there is no article that describes a detailed prospective investigation of the sensitivity, specificity, and accuracy of this technique. Therefore, we undertook a prospective, blinded investigation to determine the sensitivity, specificity, and accuracy of helical CT angiography in the detection of shunt abnormalities in patients with TIPS. A secondary objective was to determine whether the best results were obtained with one of the display methods (transverse source, multiplanar reformation [MPR], or opacity-assigned) or with a combination of methods.
|
MATERIALS AND METHODS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
After we confirmed the absence of risk factors for allergic reactions to iodinated contrast material, we reviewed the patients' recent serum creatinine levels, and any patient with a serum creatinine level higher than 2 mg/dL (177 µmol/L) was excluded. Thirty-nine patients met the inclusion criteria. Two patients were excluded because of poor venous access.
Twenty-seven patients underwent one helical CT angiographic examination each, seven underwent two studies each, and three underwent three studies each. Of the 10 patients who underwent helical CT angiography more than once, examinations were performed at intervals of 6–19 months in seven and at intervals of 8–17 days in three. These three patients underwent helical CT angiography before portography and after intervention to correct a shunt abnormality. Thus, a total of 50 helical CT angiographic examinations were performed in 37 patients (30 men, seven women; age range, 34–76 years; mean age, 52 years).
During the initial part of the study, our department protocol included routine portography for the follow-up of TIPS patency. However, during the study, the protocol changed, and US became the primary screening modality. Thereafter, all portograms were obtained on the basis of abnormal sonograms. Also, US was routinely performed at 24 hours after shunt creation or revision. Thus, the first 16 helical CT angiographic examinations were performed without a preceding sonogram; the remaining 34 had a preceding sonogram.
Helical CT Angiography
Helical CT angiography was performed with a Picker PQ or PQ5000 helical CT scanner (Picker International, Cleveland, Ohio). Although the initial 10 helical CT angiographic examinations were performed in patients with an empty stomach, the remaining 40 were performed in patients who ate a light breakfast because they were reluctant to fast for a research study.
A scout image of the upper abdomen was obtained. The range of planned data acquisition was defined from 2 cm cephalad to the hepatic venous end of the stent to 2 cm caudal to the portal venous end of the stent. The cephalad position was adjusted as necessary to include the confluence of the hepatic veins with the inferior vena cava.
A 150-mL bolus of low-osmolality contrast medium (Optiray 320 [ioversol 68%]; Mallinckrodt, St Louis, Mo) was intravenously injected at a rate of 5 mL/sec with a power injector (Medrad, Pittsburgh, Pa). By starting 50 seconds after the initiation of the injection, helical CT data were acquired in the volume in the defined range by using a section collimation of 3 mm, pitch of 1, 180–220 mA, and 120 kVp. We used a 50-second delay because of our past experience with this technique (5), and we did not perform timing trials with the generation of time-density curves to keep the procedure time short and the dose of contrast material reasonable. All scanning was performed during a single breath-hold, with the patient's breathing suspended in inspiration after hyperventilation.
Image Reformation and Hard-copy Printing
The acquired data were reprocessed to obtain 3-mm-thick overlapping transverse source images at 1-mm intervals. Fields of view ranged from 15 to 17 cm; were centered at the level of middle of the stent; and encompassed the portal vein, stent, hepatic vein, and inferior vena cava. The total number of transverse source images was 66–122, depending on the length of the stent.
The transverse source images were transferred to the Voxel Q workstation (Picker International) and were used to produce MPR images. Proprietary 4D Angio software (Picker International) that can assign different opacity values to pixels within the user-defined range of CT numbers was used to enhance the opacity of the column of contrast material and to minimize the opacity of the metallic stent to create images (opacity-assigned images) that showed the inside of the stent.
Image reformation was performed immediately after the examination by a single operator (S.C.) who was not one of the readers. The images obtained with each display method were printed on separate sheets of film. The transverse source images were printed in an abbreviated form, with every third image printed sequentially. The MPR images were created in the coronal and oblique coronal planes to demonstrate the portal vein, entire length of the stent, hepatic vein, and inferior vena cava. If all of these structures could not be included on the same image, sequential images were created to demonstrate these structures to the best possible advantage. In patients in whom the stent curved excessively, curved planar reformation images were obtained. The opacity-assigned images were created in the coronal plane and were viewed at various angles of obliquity and degrees of rotation. Images that best showed the portal vein, stent, and hepatic vein were printed on film.
All images were printed without patient identification. A code number was given to each image, and the key was available only to the principal investigator. The procedure time for each helical CT angiographic examination was recorded as the time at which the patient arrived at the CT suite to the end of image printing.
Image Analysis
The helical CT angiograms (transverse source, MPR, and opacity-assigned images) were interpreted by three readers (G.D.D., K.N.C., C.C.E.), all of whom were abdominal imaging radiologists, who were blinded to the results of portography and who did not take part in image manipulation and printing. The readers were not aware that some patients had undergone helical CT angiography more than once.
A secondary objective of this study was to determine whether the best results were obtained with one of the display methods or with a combination of methods. Therefore, with intervals of at least 2 weeks between interpretations to minimize case recall, each of the three readers made independent interpretations with the five display modes for each helical CT angiographic examination. The interpretations were made in the following order: transverse source image alone; MPR image alone; opacity-assigned image alone; transverse source and MPR images combined; and transverse source, MPR, and opacity-assigned images combined.
In every reading session, each reader evaluated the image quality, portal vein, stent, and hepatic vein. Image quality was considered to be optimal if the portal vein, stent, and length of the draining hepatic vein between the stent and the inferior vena cava were clearly visualized and opacified without major artifacts. During each session, the readers assessed each helical CT angiogram for the presence and hemodynamic significance of morphologic abnormalities.
Helical CT angiograms were interpreted as normal if the portal vein, stent, and draining hepatic vein showed no filling defects or luminal narrowing. Stent stenosis due to neointimal hyperplasia or thrombosis was diagnosed if there was a focal or diffuse filling defect that separated the column of contrast material within the stent from the stent wall. Hepatic venous stenosis was diagnosed if the diameter of the hepatic vein was narrowed compared with the internal diameter of the stent. Abnormalities were interpreted as significant if the residual lumen in the stenosed stent or hepatic vein was less than 50% of the internal diameter of the stent. The presence of each abnormality, assessment of significance, and assessment of image quality were recorded on standardized worksheets.
Portography
The final diagnosis was made at conventional portography, which was performed within 2–14 days (mean, 6 days) of helical CT angiography. The portograms were obtained by means of right jugular venipuncture with a standard angiographic technique. Injections of contrast material were administered with the catheter tip in the portal vein. Venous pressures were measured (by saline manometer) in the portal vein and hepatic vein or inferior vena cava. Portosystemic pressure gradients were calculated. The portograms were interpreted by radiologists (C.E.E., J.C.P.) who were specialized in angiography. The portograms were considered to be abnormal if they depicted filling defects within the stent and/or narrowing of the draining hepatic vein. The abnormalities were considered to be clinically significant if the portosystemic pressure gradient measured more than 15 cm of saline.
Data Analysis
Interpretations of the helical CT angiograms by all readers were separately tallied for each display mode. A cumulative impression was determined for each display mode from the unanimous impression when all three readers agreed and from the majority impression when they disagreed. The cumulative impression for each display mode was compared with that of its paired portogram.
For the diagnosis of morphologic abnormalities, helical CT angiograms that were interpreted as abnormal were considered to be true-positive if the portogram showed the same abnormality, irrespective of the portosystemic pressure gradient. For the diagnosis of significant abnormalities, helical CT angiograms that were interpreted as showing a significant abnormality were considered true-positive if the portosystemic pressure gradient was elevated at portography. The helical CT angiograms that were interpreted as normal or as showing an insignificant abnormality were considered to be true-negative for significant abnormality if the portosystemic pressure gradient was not elevated at portography.
The sensitivity, specificity, accuracy, and positive and negative predictive values for each display mode were separately calculated for the detection of all morphologic abnormalities and for the detection of significant TIPS abnormalities. Interreader agreement in the diagnosis of morphologic abnormalities and in the determination of physiologic importance was separately evaluated for each of the display methods and for the two combinations by using the 
statistic. The sign test with Bonferroni correction was applied to obtain the significance of the difference in the accuracy of the display methods and the two combinations; this test is considered to be more reliable than the McNemar test for relatively small samples. With the sign test, a P value of .005 or less indicated a significant difference. The Fisher exact test was used to determine the significance of the difference in false-positive rates at helical CT angiography for the diagnoses of hepatic venous stenosis and stent stenosis.
|
RESULTS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Portography
Overall, 31 (62%) of the 50 portograms demonstrated morphologic abnormalities, and 19 (38%) were normal. The abnormalities demonstrated were hepatic venous stenosis on 21 (68%) portograms, focal stent stenosis on five (16%), diffuse stent stenosis on four (13%), and a combination of diffuse stent and hepatic venous stenosis on one (3%). Of the 31 abnormal portograms, portosystemic pressure gradients were elevated in 24 (77%) and normal in seven (23%). All 19 morphologically normal portograms had normal portosystemic pressure gradients.
Detection of All Morphologic Abnormalities at Helical CT Angiography
For each of the helical CT angiographic display methods and for the two combinations, the number of true- and false-positive and true- and false-negative findings in the detection of all morphologic abnormalities is shown in Table 1. The sensitivity, specificity, accuracy, and positive and negative predictive values are shown in Table 2. The appearances of normal and abnormal TIPS at portography and with the various display methods at helical CT angiography are shown in Figures 1–3.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
The combined transverse and MPR mode had similar results, with a slightly lower specificity of 84% and an overall accuracy of 92%. Although the accuracy of the individual display methods (transverse source, MPR, and opacity-assigned) was lower than that of the combinations for all readers, no significant difference was found (sign test, P = .021).
Detection of Hemodynamically Significant Abnormalities at Helical CT Angiography
For each of the helical CT angiographic display methods and for the two combinations, the number of true- and false-positive and true- and false-negative readings in the detection of hemodynamically significant abnormalities is shown in Table 3. The sensitivity, specificity, accuracy, and positive and negative predictive values are shown in Table 4.
|
|
Of the two false-negative findings that were made with this combination, one was in a patient with significant focal stent stenosis near the portal venous end (described previously); the abnormality was missed by all readers at helical CT angiography. In the other false-negative finding, hepatic venous stenosis was correctly diagnosed at helical CT angiography, but it was incorrectly considered to be insignificant. Of the six false-positive findings, five were in patients who had stent stenosis that appeared to be morphologically severe at helical CT angiography, as well as portography, but the portosystemic pressure gradients were not elevated. The sixth finding was in a patient with a normal portogram, but the helical CT angiogram was interpreted as depicting significant hepatic venous stenosis. Thus, one (5%) of 21 and five (71%) of seven helical CT angiograms were false-positive for significant hepatic venous stenosis and significant stent stenosis, respectively.
The false-positive rate for significant stent stenosis was significantly higher than that of significant hepatic venous stenosis (P < .001, Fisher exact test). Combined interpretation of the transverse source and MPR images had similar results, but with somewhat lower specificity and overall accuracy of 73% and 82%, respectively. Although the accuracy of the individual display methods was lower than that of the combinations for all readers, no significant difference was found (sign test, P = .032).
Interreader Agreement
Interreader agreement was highest with the transverse source and MPR combination in the detection of all morphologic abnormalities and for hemodynamically significant morphologic abnormalities, with values of 0.75 and 0.84, respectively. As shown in Table 5, the values for interreader agreement for the separate display methods and for the combination of transverse source, MPR, and opacity-assigned images were less than that of the combination of transverse source and MPR images.
|
Procedure Times
The total time for data acquisition and reprocessing to obtain the transverse source image was 8–15 minutes (mean, 12 minutes). The total postprocessing time to obtain MPR and opacity-assigned images was 5–15 minutes (mean, 10 minutes). The total procedure time was 15–26 minutes (mean, 22 minutes).
|
DISCUSSION |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Helical CT angiography was less accurate for the detection of significant abnormalities than for the detection of all abnormalities mainly because of a lower specificity that resulted from a higher false-positive rate with severe stent stenoses. This finding probably represents the inadequacy of the criterion of significance used in our study. To our knowledge, there are no published criteria for the determination of the significance of the abnormalities depicted at helical CT angiography. Therefore, we used the angiographic criterion for significant abnormalities of a reduction of more than 50% in the stent lumen or the diameter of the hepatic vein.
In our experience, the visible degree of stent stenosis at helical CT angiography does not always correlate well with the portosystemic pressure gradient. A possible explanation is that artifactual thickening of the stent on helical CT angiograms exaggerates the apparent severity of lesions within the stent by making the diameter of the stent lumen appear narrower than it actually is. However, even when the severity of the morphologic lesions that produced stent stenosis was comparable on portograms and helical CT angiograms, it did not translate to an elevated portosystemic pressure gradient. The development and testing of other criteria of significance (eg, recurrence of varices or ascites) may improve the results of helical CT angiography in the future.
All data obtained at helical CT angiography were depicted on a transverse source image. However, the large number of transverse source images and the curved course of the stent make the use of a three-dimensional display method desirable. Among the CT angiographic examinations, helical CT angiography of TIPS has the unique requirement of the demonstration of abnormalities in the lumen of the high-attenuating metallic stent. Therefore, the maximum intensity projection and shaded-surface display methods that use the highest-attenuating pixels in a volume of data to produce angiographic images are not appropriate for displaying helical CT angiograms of TIPS.
In our study, we used the MPR display method because it is the simplest and most widely available. In addition, we used commercially available opacity-assigned software with the idea that by rendering the stent walls transparent, we might increase our accuracy in the interpretation of helical CT angiograms of TIPS. However, because the software is not universally available, we believed it was important to determine whether its use improved diagnostic accuracy. Therefore, we compared the accuracy of the interpretation of helical CT angiograms of TIPS in the following five different display modes: transverse source alone; MPR alone; opacity-assigned alone; transverse source and MPR combination; and transverse source, MPR, and opacity-assigned combination. The difference in the accuracy of the various display modes did not reach statistical significance. However, in our opinion, this finding was due to a small sample size.
Overall, the results obtained from the interpretation with the combined displays were better than the results of the different display methods alone. Of the two combined displays, the transverse source and MPR combination was almost as accurate as the transverse source, MPR, and opacity-assigned combination. Furthermore, the degree of interreader agreement was highest with the transverse source and MPR combination. In our opinion, the use of MPR images with transverse source images improves accuracy. No significant advantage in diagnostic accuracy was gained by adding the opacity-assigned display. The special software is not essential to achieve optimum results with helical CT angiography of TIPS. Therefore, helical CT angiography of TIPS can be satisfactorily performed in any radiology department that has a helical CT scanner capable of MPR.
Currently, Doppler US is the technique that is most widely used for routine imaging of TIPS. It is inexpensive, noninvasive, and fairly accurate when performed by experienced clinicians. Doppler US has a widely accepted sensitivity of 92% and specificity of 72% for the detection of TIPS abnormalities (4,7). Therefore, any new modality for noninvasive imaging of TIPS will be compared with Doppler US.
However, despite its advantages, Doppler US of TIPS has some limitations. In our experience, the poor window provided by the cirrhotic liver, deep location of the stent, difficulty in obtaining an adequate angle of insonation, and presence of ascites make Doppler US of TIPS difficult to perform, even by experienced sonographers. The uncertainty and multiplicity of the diagnostic criteria for abnormalities make interpretation of the findings difficult. This difficulty is reflected in the poor accuracy of Doppler US of TIPS that some investigators recently reported (8).
In addition, we have found that Doppler US of TIPS is demanding for the patient because it is time-consuming and because it involves repeated and prolonged breath-holding. On the other hand, helical CT angiography requires only a single breath hold, and is performed more quickly. It provides a set of easily reproducible images and shows abnormalities clearly. Although a direct comparison cannot be made because of the different patient populations, the sensitivity, specificity, and accuracy of helical CT angiography determined in our study appear to approach those of Doppler US that have been reported in the literature.
Helical CT angiography itself has certain limitations. It uses ionizing radiation and requires a relatively high dose of intravascular contrast medium that is intravenously injected at a higher rate in a single 30-second breath hold. However, similar amounts of ionizing radiation, amounts of contrast medium, and rates of intravenous injection are routinely used at CT examination before TIPS creation, before transplantation, and for hepatocellular carcinoma screening in patients with cirrhosis. Therefore, regarding CT for other indications, patients with previous reactions to contrast material, renal insufficiency, inaccessible peripheral veins, or difficulty with breath-holding are not appropriate candidates for helical CT angiography. In our study, two (5%) of the 39 patients who agreed to take part could not undergo helical CT angiography for these reasons. Nevertheless, we postulate that, in general, these limitations of helical CT angiography are outweighed by the ease with which it is performed and with which the findings are interpreted.
There are a few limitations of this study. First, findings from helical CT angiography was interpreted on the hard-copy images rather than at the CT workstation. With most computer-generated images, manipulation of the images at the workstation by the reader increases the confidence of interpretation. It is possible that the results in this study could have been better if our readers had interpreted the images on the workstation. However, interpretation at the workstation can be tedious and is often time-consuming.
Our aim was to assess the accuracy of helical CT angiography with the simplest form of image presentation (ie, hard-copy images) to achieve maximum time efficiency. Our results show that a high accuracy can be obtained from the interpretation of hard-copy helical CT angiograms. For this study, image manipulation was performed by one of the radiologists (S.C.) who was not involved in the interpretation of the images. In our opinion, this task will be performed, in practice, by a technologist who is trained in three-dimensional image manipulation. However, we believe that, in the future, the ongoing expansion of interactive image display technology will allow quicker, direct, and probably more accurate interpretation of helical CT angiographic findings of TIPS on a workstation.
Second, our patient population was not truly representative of a screening population because of a bias that arose from the selection of some patients on the basis of an abnormal sonogram. Therefore, our results do not necessarily represent the performance of this test as a screening modality.
Third, in this study, we did not directly compare helical CT angiography with US. Because of the lack of existing data on the accuracy of helical CT angiography of TIPS, we believed that it was premature to conduct a comparison study of the two modalities without proof of the relative merit of helical CT angiography. The results of helical CT angiography of TIPS are promising enough to warrant further study.
In conclusion, we found that helical CT angiography has a high sensitivity and a reasonable specificity for the detection of TIPS abnormalities. Helical CT angiography holds promise as a screening modality for the detection of TIPS abnormalities. However, further studies are necessary to compare helical CT angiography with Doppler US of TIPS and to determine its potential to supplant US for the detection of TIPS abnormalities.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Abbreviations: MPR = multiplanar reformation TIPS = transjugular intrahepatic portosystemic shunt
Author contributions: Guarantor of integrity of entire study, S.C.; study concepts, G.D.D.; study design, S.C., G.D.D., A.A.G.; definition of intellectual content, S.C., G.D.D.; literature research, S.C.; clinical studies, S.C., A.A.G., C.E.E., J.C.P.; data acquisition, G.D.D., K.N.C., C.C.E., H.R.; data analysis, S.C.; manuscript preparation and editing, S.C.; manuscript review, K.N.C., S.C., G.D.D.
|
References |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Related Article
This article has been cited by other articles:
|
|
 |
|
  P. Otal, T. Smayra, C. Bureau, J. M. Peron, V. Chabbert, P. Chemla, F. Joffre, J. P. Vinel, and H. Rousseau Preliminary Results of a New Expanded-Polytetrafluoroethylene--Covered Stent-Graft for Transjugular Intrahepatic Portosystemic Shunt Procedures Am. J. Roentgenol., January 1, 2002; 178(1): 141 - 147. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Matsumoto, M. Kitamoto, M. Imamura, T. Nakanishi, C. Ono, K. Ito, and G. Kajiyama Three-Dimensional Portography Using Multislice Helical CT Is Clinically Useful for Management of Gastric Fundic Varices Am. J. Roentgenol., April 1, 2001; 176(4): 899 - 905. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
