| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 All authors: Department of Diagnostic Radiology, Graduate School of Medical Sciences, Kumamoto University, 1-1-1 Honjyo, Kumamoto 860-8556, Japan.
Received March 9, 2007;
accepted after revision July 15, 2007.
WEB This is a Web exclusive article.
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. This prospective study included 40 patients: 22 had hypervascular hepatocellular carcinoma (HCC), and 18 were without liver tumors. We obtained four-phase contrast-enhanced scans using a 16-MDCT unit. The section thickness and interval were 2 and 0.5 mm, respectively. Scanning for the first through fourth scans was started 10, 35, 70, and 180 seconds after the inception of contrast injection, respectively. Scanning for the first and second phase was within a single breath-hold. We subtracted the first-phase images from the second-phase images using software developed in-house. We used receiver operating characteristic (ROC) analysis with a continuous rating scale from 1 to 100 to compare observer performance in the detection of focal enhancement on second-phase images. Eight radiologists participated in the observer performance test, and their performances with unenhanced and contrast-enhanced original images were compared with their performances using contrast-enhanced subtracted images.
RESULTS. For the eight observers, the mean area under the best-fit ROC curve (Az) values without and with the subtracted images were 0.86 ± 0.05 (SD) and 0.91 ± 0.03, respectively. The difference was significant (p < 0.01, two-tailed paired Student's t test).
CONCLUSION. The display of subtracted images significantly improved the diagnostic performance of radiologists in the detection of focal enhancement during the hepatic arterial phase (p < 0.01).
Keywords: cirrhosis • hemodynamics • hepatocellular carcinoma • liver disease • MDCT • oncologic imaging • subtraction MDCT
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
MDCT features high-volume coverage speed and is now widely used. Because the entire liver can be scanned within several seconds, two sequential acquisitions are possible during a single breath-hold [22-27]. Spielmann et al. [28], who reported the technical feasibility of performing single breath-hold dynamic subtraction CT of the liver with MDCT, found that the mean lesion contrast was 2.5 times greater on subtracted scans than on HAP scans. However, they did not compare the detectability of enhanced lesions in the liver on scans obtained by the conventional technique with that on scans obtained by the new imaging technique.
The purpose of our study was to evaluate the effect of single breath-hold 3D pixel shift dynamic subtraction MDCT of the liver on the performance of radiologists charged with detecting focal enhancement during the HAP.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Patients and Tumors
Between February and May 2004, we enrolled 51 patients who met our
inclusion criteria. These criteria were, first, type B, type C, or alcoholic
hepatitis; second, confirmed HCC that had been untreated during the 3 months
preceding our study or suspicion of space-occupying hepatic lesions based on
sonographic evidence or elevated tumor markers (
-fetoprotein) or
protein induced by vitamin K absence or antagonist-II (PIVKA-II); third,
absence of both renal failure (serum creatinine < 1.5 mg/dL) and
contraindication for iodinated contrast material. Eleven patients were
excluded: two had difficulty with the breath-hold, five had multiple tumors
involving the entire liver, and four had hypovascular HCCs. We excluded
patients with multiple tumors involving the entire liver because of the
possibility that their presence changed the hepatic hemodynamics and rendered
tumor enhancement during the HAP unclear.
The definitive diagnosis of hypovascular HCC was based on histopathologic findings after hepatic surgery (n = 3) and needle biopsy (n = 1) and on the detection of hypovascularity on CT during hepatic arteriography. The final study population of 40 patients consisted of 29 men and 11 women who ranged in age from 26 to 84 years (mean, 62.6 years). For the men, the age range was 29-84 years (mean, 62.4 years), and for the women, it was 26-80 years (mean, 63.2 years). There was no significant age difference between the men and women (p = 0.89, two-tailed Student's t test).
In the 22 patients with HCCs, the definitive diagnosis of hypervascular HCC
was based on histopathologic evidence after hepatic surgery (n = 11
patients); needle biopsy (n = 7); or substantially increased levels
of -fetoprotein or PIVKA-II, with follow-up CT showing no change or an
increase in the tumor size within 6 months (n = 4). Furthermore, two
board-certified radiologists with 20 and 8 years of experience with liver CT,
respectively, who did not participate in the observer performance study,
consensually confirmed the presence of hypervascular tumors on CT during
hepatic arteriography. At our institution, we routinely subject all patients
with HCCs to CT arteriography or CT portography before starting treatments.
There were 39 HCC nodules in the 22 patients; the mean size of the HCC nodules
was 16.1 ± 8.1 (SD) mm (range, 7-43 mm).
CT and Contrast Injection Protocol
All patients were scanned with a 16-MDCT scanner (IDT16, Philips Medical
Systems). The parameters were 0.75-second rotation time, 16 x 1.5 mm
detector collimation, 2.0-mm image thickness, 0.5-mm image interval, 0.9
helical pitch (beam pitch), 32.2 mm/s table movement, 40-cm scanning field of
view, 120 kV, and 300 mAs. Image reconstruction was in a 25- to 35-cm display
field of view depending on the patient's physique.
Helical CT scanning was started at the top of the liver in a cephalocaudal direction, and four-phase contrast-enhanced helical scans of the entire liver were obtained. Patients were instructed to hold their breath with tidal inspiration during scanning. First- and second-phase contrast-enhanced scanning was started 10 and 35 seconds after the inception of contrast injection, respectively. The total acquisition time for first- and second-phase scanning was approximately 30 seconds; scans were obtained during a single breath-hold. The contrast material did not arrive at the proper hepatic artery during the first phase in any of the patients. We consider the second phase to correspond with the ordinary HAP. We used first- and second-phase images as unenhanced and contrast-enhanced images, respectively, for the generation of subtracted images.
We did not perform late HAP scanning (
40-50 s after the inception of
contrast injection), a time at which hypervascular HCCs were enhanced
maximally [23,
24,
26], because the interval
between first-phase (10 s after the inception of contrast injection) and late
HAP scanning was too long and patients may not be able to hold their breath
for that length of time. We began third- and fourth-phase contrast-enhanced CT
70 and 180 seconds after the start of contrast injection; these scans
corresponded with the portal venous phase and equilibrium phase,
respectively.
A power injector (Dual Shot, Nemoto Kyorindo Co.) was used to administer 100 mL of iopamidol (Iopamiron, Nihon Schering) with an iodine concentration of 370 mg I/mL via a 20-gauge IV catheter inserted into an antecubital vein. The injection rate was 4.0 mL/s in all patients, and contrast delivery was followed by flushing with 30 mL of physiologic saline at the same injection rate.
Generation of Contrast-Enhanced Subtracted Images
One board-certified radiologist with 8 years of experience with liver CT
performed image registration and subtraction of unenhanced from
contrast-enhanced images using software he developed.
First, CT image data were transferred to a PC (EdiCube S270P, Epson) with a 1.5-GHz processor (Pentium M, Intel). Second, 3 x 3 median and 3 x 3 mean filters were applied to the unenhanced and contrast-enhanced images to reduce the noise on the subtracted images. Although the use of the filters may blur the edge of enhanced lesions somewhat, it improves the signal-to-noise ratio. Because the primary purpose of the contrast-enhanced subtracted images was improved identification of the tumor stain, we applied these filters to the subtracted images. Third, the same radiologist manually performed 3D pixel shift registration using physical locations similar to those of the internal markers. Then the attenuation values for paired images from unenhanced and contrast-enhanced data sets that coincided with respect to their physical location in the patient's body were subtracted on a voxel-by-voxel basis. Finally, subtracted axial images were output in DICOM format as another series of CT studies.
The mean calculation time for the generation of subtracted images in each case was approximately 10 minutes.
Observer Performance Test
Using a sequential test method
[29], we applied ROC analysis
to evaluate whether subtracted CT images improved the detection of focal
enhancement during the second phase (i.e., HAP). The observers were four
board-certified radiologists with 8-23 years of experience (mean, 12.8 years)
and four attending radiologists with 3-5 years of experience with liver CT
(mean, 4.0 years). All observers specialize in body imaging and read liver CT
images regularly.
To expedite the observer performance study, a radiologist selected 10 contiguous sections, including those that contained the largest tumor on unenhanced, contrast-enhanced original, and contrast-enhanced subtracted images in each patient with HCC. The mean size of the selected HCCs was 16.1 ± 8.1 (SD) mm (range, 7-43 mm). The 10 selected section levels on all images were in conformance. Similarly, in the 18 patients without HCC, he arbitrarily selected 10 contiguous sections on unenhanced, contrast-enhanced original, and contrast-enhanced subtracted images.
To indicate their judgment with respect to the detection of focal enhancement, the observers marked their confidence level on a continuous rating scale [30, 31]. Using a mouse, they logged their judgment on a horizontal bar displayed on the screen; "definitely present" and "definitely absent" were indicated at the right and left ends of the bar, respectively. The unenhanced and contrast-enhanced original images were presented first. After each observer indicated the initial level of confidence, contrast-enhanced subtracted images were displayed on the monitor. Each observer then had the opportunity to change the previously indicated confidence level. The instructions were to use the rating scale consistently and uniformly. To ensure their ability to operate the observer interface, before participating in the observer performance test all observers were trained on three HCC cases not included in the subsequent test. They were not given specific criteria for judging the presence of focal enhancement; rather, they were instructed to apply their previous experience.
|
Quantitative Analysis
In the 22 patients with HCC, we measured tumor attenuation on unenhanced,
contrast-enhanced original, and subtracted images using the same tumor as in
the observer performance test. An attempt was made to maintain a region of
interest (ROI) of approximately 0.5 cm2; the ROI areas ranged from
0.3 to 0.5 cm2.
We also measured hepatic attenuation in three separate areas (left lobe and anterior and posterior segments of the right lobe) on unenhanced, contrast-enhanced original, and subtracted images. The attenuation values for each image series were averaged. An attempt was made to maintain a constant ROI of approximately 2 cm2; the range of the ROI areas was 0.8-2.0 cm2. Visible blood vessels, bile ducts, and artifacts were carefully excluded from ROI measurements in the hepatic parenchyma.
The conspicuity of a hepatic tumor can be expressed by the attenuation difference between it and the hepatic parenchyma, the so-called tumor-to-liver contrast [15]. We defined tumor-to-liver contrast as the value obtained by subtracting parenchymal attenuation from hepatic tumor attenuation. We determined the tumor-to-liver contrast for the same tumor that we selected for the observer performance test on unenhanced, contrast-enhanced original, and subtracted images. We attempted to maintain an ROI area of approximately 0.5 cm2 (range, 0.3-0.5 cm2).
To obtain the parenchymal attenuation value used for tumor-to-liver contrast calculations, we measured the normal hepatic parenchyma at least 1 cm from the edge of the tumor to nullify the risk of encountering fibrosis. An attempt was made to maintain a constant ROI of approximately 2 cm2; the range of the ROI areas was 0.8-2.0 cm2. All attenuation values were determined by the board-certified radiologist who also measured the length of misregistration on contrast-enhanced subtracted images at the peak of the right diaphragm in each patient.
We determined the detectability of HCCs on contrast-enhanced original images. We considered a tumor detected when more than four observers assigned a confidence level of greater than 50% to their judgment of the presence of focal enhancement in the observer performance study. Similarly, we considered a tumor undetected when more than four observers assigned a confidence level of 50% or less to their judgment of the presence of focal enhancement.
Statistical Analysis
ROC analysis was used to compare the observers' performances in detecting
focal enhancement during the HAP on original and subtracted images. A binormal
ROC curve was fit to each reader's confidence rating data derived under two
reading conditions for quasimaximum likelihood estimation
[30]. We used a computer
program (ROCKIT, Charles E. Metz, University of Chicago) to obtain binormal
ROC curves from the ordinal-scale rating data
[30]. The area under the
best-fit ROC curve (Az) plotted in the unit square was
calculated for each fitted curve. We applied the two-tailed paired Student's
t test to determine the statistical significance of the difference
between the ROC curves obtained with original and subtracted images and the
contrast-enhanced original and subtracted images. The statistical significance
of the difference in tumor-to-liver contrast on unenhanced and
contrast-enhanced original or subtracted CT images was tested using the paired
Student's t test.
The paired Student's t test was performed with a statistical software package (version 14.0, SPSS), and p values of less than 0.05 were considered to indicate a statistically significant difference.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
|
Of the 18 patients without liver tumors, one was incorrectly diagnosed by
three radiologists. In this patient, the false-positive diagnosis was
attributable to a blood flow anomaly that was retrospectively identified by
two board-certified radiologists. They confirmed the presence of hypervascular
lesions on CT during hepatic arteriography and CT during arterial portography;
follow-up CT scans obtained 3 months later showed no change in the lesion size
and no increase in tumor marker levels (i.e., -fetoprotein or
PIVKA-II). Two other patients were incorrectly diagnosed by two radiologists,
and six were incorrectly diagnosed by one. In these eight patients, the
reasons for the incorrect diagnoses could not be determined.
|
|
|
|
|
More than five of the eight observers judged 13 (59%) of the 22 patients with HCC as having detectable tumors; in four of the 13 patients, the tumors were not detected on contrast-enhanced original images. All of the undetected tumors were judged to be visible on contrast-enhanced subtracted images (Figs. 2A, 2B, 2C, 2D and 3A, 3B, 3C, 3D).
|
|
|
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Our quantitative data suggest that observers found it difficult to identify the undetected tumors on contrast-enhanced original images because their enhancement resulted in decreased tumor-to-liver contrast on contrast-enhanced original images. Thus, the tumor-to-liver contrast on contrast-enhanced images does not correspond with the degree of enhancement and only contributes to lesion conspicuity [15]. The unenhanced CT study helps to determine whether the hepatic tumor is hypervascular, and contrast-enhanced subtraction CT facilitates the identification of hypervascular tumors during the HAP.
Honda et al. [10] reported that 19 of 49 HCCs (38.8%) that manifested hypervascularity on selective celiac arteriographs were iso- or hypoattenuated on early-phase images of dynamic incremental hepatic CT. The lack of correspondence between the tumor-to-liver contrast on hepatic CT and lesion enhancement on HAP images may explain this discrepancy between CT and angiographic findings.
Spielmann et al. [28] subtracted unenhanced from HAP CT scans. They used 5-mm-thick images and 2-mm intervals and found that in 40% of their patients the subtracted images had more than 5 mm of misregistration and that the accuracy of subtraction by their method was unsatisfactory. We used 2-mm-thick images and 0.5-mm intervals and performed 3D subtraction of unenhanced from HAP CT scans pixel by pixel with software developed in-house. The mean misregistration on our contrast-enhanced subtracted images was 1.9 mm, and the quality of the extracted images was excellent in all patients. The recently introduced 40- and 64-MDCT scanners can scan the entire liver in less than 3 seconds with a spatial resolution of less than 1.0 mm and can yield subtracted images of excellent quality.
In this study, we manually performed 3D pixel shift registration between different-phase images. We are in the process of developing fully automated 3D nonlinear registration methods to generate subtracted images. Although the cirrhotic liver is solid, the normal liver is considerably soft and elastic; therefore, different degrees of breath-hold transform this organ to a certain degree. Nonlinear registration facilitates accurate subtraction among different breath-holds, and long breath-holds, as used in our study, may not be necessary. In addition, fully automated registration may drastically reduce the time required for radiologists to detect focal hepatic enhancements. We are planning to apply computer-aided diagnosis (CAD) to the detection and 3D quantification of early enhancement in the liver.
We must point out that our technique improves only the identification of lesion stain that occurs during the HAP. We excluded patients with hepatic blood flow anomalies from our study population to simplify the interpretation of observer performance study results. In routine clinical studies, false-positive results attributable to blood flow anomalies may be encountered.
There are some potential limitations in our study. First, we subtracted the first-from the second-phase scans. Although the contrast material did not arrive at the proper hepatic artery during the first phase in any of our patients, strictly considered, the first-phase scans were not unenhanced scans. In the first phase, HCC and the hepatic parenchyma might be slightly enhanced. Second, to expedite the observer performance study, in patients with HCC we selected 10 contiguous sections that included the largest tumor; this may have introduced selection bias. Third, only 40 cases were presented to the eight participants in the observer performance test to allow them to perform their interpretation in a relatively short time. A larger number of cases and observers may be needed to validate our statistical results.
In conclusion, single breath-hold dynamic subtraction MDCT was useful for the detection of focal hepatic enhancement in patients with HCC. We are in the process of performing detailed analyses of the shape and internal texture of focal hepatic enhancements on contrast-enhanced subtracted images to determine whether this method advances differentiation of various liver tumors.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
1.5 cm) malignant hepatic neoplasms.
AJR 1995; 164:879
-884
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
