| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, University of Trieste, Cattinara Hospital, Strada di
Fiume 447, Trieste 34149, Italy.
2 Department of Radiology, Hospital of Mestre-Venice, Venice, Italy.
Received March 24, 2005;
accepted after revision August 12, 2005.
Address correspondence to E. Quaia
(equaia{at}yahoo.com).
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionAPPENDIX 1: Imaging Criteria...References |
|---|
SUBJECTS AND METHODS. This series comprised 166 liver tumors (1-5 cm in diameter) in 166 patients (99 men, 67 women; mean age ± SD, 58 ± 11 years) scanned at low transmit power (mechanical index: 0.1-0.14) after sulfur hexafluoride-filled microbubble injection. Digital cine clips recorded at the arterial phase (10-40 sec after contrast injection) and late phase (100-300 sec) were analyzed to characterize liver tumors as benign or malignant. Visual analysis was performed by three independent blinded reviewers who evaluated enhancement patterns at the arterial phase and subjective tumor conspicuity at the late phase. Quantitative analysis of videotape intensity (VI: gray-scale levels, 0-255) was performed to calculate objective tumor conspicuity at the late phase: (VItumor - VIliver) / VIliver.
RESULTS. Characteristic enhancement patterns were observed in malignant tumors (peripheral rimlike) and benign tumors (peripheral nodular or central and spoke-wheel-shaped). Malignant (n = 95) versus benign (n = 71) tumors differed for subjective (median value: -1 vs 1, respectively) and objective conspicuity at the late phase (-0.6 vs 0.15, respectively; p = 0.001, Mann-Whitney U test) due to persistent microbubble uptake in benign tumors. Diagnostic performance of visual (odds ratio: reviewer 1 = 4.28, reviewer 2 = 10.18, reviewer 3 = 9.56) and quantitative (odds ratio: 89.33) analyses differed significantly in the characterization of liver tumors (p = 0.01, chi-square test).
CONCLUSION. Quantitative analysis revealed higher diagnostic performance than visual analysis to characterize liver tumors insonated at low transmit power after microbubble contrast agent injection.
Keywords: contrast media • harmonic sonography • insonation • liver • liver disease • sonography
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionAPPENDIX 1: Imaging Criteria...References |
|---|
Microbubbles present a small diameter (
3 µ), a shell of
biologically inert material (galactose [e.g., Levovist, SH U 508A, Schering];
albumin [e.g., Optison, Amersham Health]; or phospholipids [e.g., Definity,
Bristol-Myers Squibb]), and a different filling gas (air [e.g., Levovist];
perfluorocarbon [e.g., Definity]; or sulfur hexafluoride [SonoVue,
Bracco-Italy]). High-transmit-power insonation (mechanical index = 0.9-1.2)
produces bubbles destruction with emission of a wideband harmonic signal and
presents several drawbacks such as the transiency of harmonic signals, the
lack of suppression of background stationary tissues, and the strong presence
of artifacts [6,
7]. Conversely,
low-transmit-power insonation (mechanical index = 0.08-0.2) produces selective
resonance of microbubbles with persistent harmonics emission and effective
background suppression [7]. The
vascularity of solid liver tumors is depicted at the arterial phase (10-40 sec
from contrast injection) with enhancement patterns differing according to the
tumor histotype [3-6,
8-12],
while the malignant or benign nature of the tumor is defined at the late phase
(100-300 sec from contrast injection) by evidence of persistent microbubble
uptake in benign tumors and washout in malignant tumors
[13-15].
Visual analysis is the first-line method to assess liver tumors after contrast injection, although it is limited by wide interobserver variability [5, 8, 10, 16] and depends on observer experience, with a consequent low reproducibility of results. Quantitative analysis is the second-line method, which provides more objective, reliable, and reproducible results [8, 16, 17]. Quantitation of sonography videotape intensity was previously shown to effectively characterize liver tumors after high-transmit-power insonation [13, 17], even though visual analysis and quantitative analysis were not compared. Moreover, because low-transmit-power insonation produces fewer artifacts than high transmit power [18], it is expected to allow an even more reliable quantitative analysis of sonography videotape intensity.
The aim of this study was to compare diagnostic performance of visual analysis with quantitative analysis for the characterization of liver tumors insonated at low transmit power after microbubble contrast agent injection.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionAPPENDIX 1: Imaging Criteria...References |
|---|
|
Contrast-Enhanced Sonography
From 1 to 28 days after identification, each liver tumor was examined on
contrast-enhanced sonography by the supervising radiologist, who had 7 years
of experience in contrast-enhanced sonography of the liver at the time of the
study. Approval was obtained by the ethics review board of our hospital, and
informed consent was obtained from all patients at the time of scanning after
the nature of the procedure had been fully explained.
In this study, only one liver tumor was scanned in each patient. If more than one liver tumor was identified, the tumor considered indeterminate at baseline and with the best possible acoustic window for scanning was selected. A preliminary baseline grayscale sonography scan was obtained, and liver tumors were located by liver segment according to Bismuth [19] and Couinaud [20] classification systems. The tumor diameter was measured in the transverse and longitudinal planes, and the largest diameter in centimeters was registered (Table 1).
Each tumor was then examined after the injection of an IV bolus (2.4 mL in
2 sec) of sulfur hexafluoride-filled microbubbles (BR 1, SonoVue, Bracco)
followed by a 2-mL flush of 0.9% NaCl. Sonography was performed while the
patient was breathing normally or during breath-holding, depending on which
yielded the best visualization of the tumor. For consistency in imaging
technique and analysis, all sonography examinations were performed using the
same system (HDI 5000 with C5-2 convex array probe, Advanced Technology
Laboratories-Philips). The technical parameters were as follows: pulse
inversion mode as the contrast-specific sonography technique, central transmit
frequency of 3.5-3.7 MHz, low transmit power (mechanical index: 0.1-0.14),
dynamic range of 65 dB, temporal resolution between frames of 75-100 msec
(10-13 frames for seconds), echo-signal gain below noise visibility, signal
persistence turned off, and one focus below the level of the tumor.
Uncompressed audio video interleaved (AVI) digital cine clips were recorded by a digital video camera (PC 105E, Sony) connected to the sonography equipment during the arterial (10-40 sec from contrast injection), portal (45-95 sec), and late (100-300 sec) phases and were successively stored in a PC (Intel, Pentium 4). Digital cine clips showed 10-20 sec during the arterial and portal phases and 60-80 sec during the late phase.
|
|
The diagnostic criteria reported in Figures 1A and 1B were obtained from previous studies [3-6]. Each reviewer was presented with digital cine clips registered at the arterial and late phases and was asked to diagnose each tumor as benign or malignant on the basis of the enhancement patterns at the arterial phase and subjective tumor conspicuity at the late phase. In the assessment of tumor conspicuity, each reviewer compared the gray-scale intensity of the most enhancing portion of the tumor with the adjacent liver parenchyma according to a 5-grade system (Figs. 2A, 2B, 2C, 2D and 2E). For those enhancement patterns (e.g., diffuse homogeneous, diffuse heterogeneous, or absent) that are common for both malignant and benign tumors, tumor conspicuity at the late phase was considered determinant for differential diagnosis (hypovascular = malignant; isoto hypervascular = benign).
|
|
|
|
|
The TIFF images were analyzed using software (Photoshop [release 7.0], Adobe Systems). In each image, two circular (range, 4,100-40,110 pixels; mean, 21,430 pixels) manually defined regions of interests (ROIs) were drawn in consensus by the two observers: The first encompassed the major portion of the tumor (Fig. 3A) and the second, a portion of the adjacent liver parenchyma with homogeneous appearance (Fig. 3B). Both ROIs were drawn at approximately the same depth, avoiding blood vessels, artifacts, and the echogenic walls of portal vessels.
|
|
|
 |
which we developed by also considering previous studies
[13]. Objective tumor
conspicuity was < 0 in tumors that were hypovascular compared with adjacent
liver parenchyma and 
0 in those that were isoor hypervascular. The
predefined value of 0 was selected as the cutoff to differentiate benign from
malignant tumors. This cutoff value was based on previous studies
[3-5,
8-15]
in which malignant tumors were described as mostly hypovascular compared with
adjacent liver, whereas benign tumors revealed persistent microbubble uptake
and appeared isoto hypervascular.
|
|
|
|
|
|
|
value < 0.20), moderate ( 0.20 and < 0.40),
fair ( 0.40 and < 0.60), good ( 0.60 and < 0.80), or very good
( 0.8-1). After the Shapiro-Wilk test failed to show evidence of normal distribution, the nonparametric Mann-Whitney U test was used to test the differences between malignant and benign tumors in subjective and objective conspicuity median values. The difference in diagnostic performance between visual and quantitative analysis was assessed through the chi-square test with Yates correction. A p value of less than 0.05 was considered to indicate significant difference.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionAPPENDIX 1: Imaging Criteria...References |
|---|
Visual Analysis
Enhancement patterns at the arterial phase— Hepatocellular
carcinomas revealed diffuse homogeneous (n = 22) or heterogeneous
(n = 27) contrast enhancement (Figs.
4A and
4B). Metastases revealed
absent (n = 12), diffuse (n = 30), or rimlike (n =
2) enhancement. Peripheral cholangiocarcinomas displayed persistent absent
enhancement with a hypovascular appearance.
Most liver hemangiomas (n = 32) revealed nodular peripheral enhancement with progressive centripetal fill-in (Figs. 5A and 5B). Diffuse contrast enhancement was identified in three liver hemangiomas < 3 cm that were proven to be hypervascular at histology, and absent contrast enhancement was observed in two liver hemangiomas < 3 cm that were proven to be thrombosed at histology. Most macroregenerative nodules (n = 14) revealed absent contrast enhancement, whereas diffuse contrast enhancement was observed in the remaining macroregenerative nodules with a dysplastic pattern (n = 5). Focal nodular hyperplasias revealed diffuse homogeneous contrast enhancement, preceded (n = 7) or not (n = 2) by central-spoke-wheel-shaped enhancement lasting for 2-5 sec. Hepatocellular adenomas revealed persistent diffuse heterogeneous contrast enhancement.
Subjective tumor conspicuity at the late phase—Distribution
of subjective tumor conspicuity values at the late phase for each reviewer is
shown in Figures 6A,
6B and
6C. Malignant tumors appeared
prevalently hypovascular (median subjective conspicuity = -1 for all
reviewers), whereas benign tumors appeared prevalently hypervascular (median
subjective conspicuity = 1 for all reviewers). The difference in subjective
conspicuity grading was significant (p = 0.001) with good
interobserver agreement ( = 0.73, reviewer 1 vs reviewer 2; 0.76,
reviewer 2 vs 3; 0.77, reviewer 1 vs 3).
Quantitative Analysis: Objective Tumor Conspicuity at the Late Phase
The distribution of objective tumor conspicuity values at the late phase
for the different histotypes is shown in
Table 2, and the distribution
of objective tumor conspicuity values for benign and malignant tumors is shown
in Figure 7.
|
|
According to the tumor nature, objective conspicuity was < 0 in 80 (84%)
of 95 malignant tumors and 0 in 67 (94%) of 71 benign tumors, and the
difference in the median values was significant (p = 0.001).
Objective conspicuity was 0 in 15 (16%) of 95 malignant tumors, while it
was < 0 in four (6%) of 71 benign tumors.
According to the tumor histotypes, 15 (31%) of 49 well-differentiated
(n = 10) or moderately-poorly differentiated (n = 5)
hepatocellular carcinomas revealed objective conspicuity of 0, while two
(40%) of five dysplastic nodules with high (n = 1) or low (n
= 1) grade and two (5%) of 37 hemangiomas with a thrombotic pattern revealed
objective conspicuity of < 0. Intrahepatic cholangiocarcinomas revealed
conspicuity of < 0, similarly to the other malignant histotypes.
Comparison of Visual Versus Quantitative Analysis in Diagnostic Performance
Results of diagnostic performance of visual analysis and quantitative
analysis are shown in Table 3.
The difference in diagnostic performance between the different reviewers was
not significant (p 0.05), and the interobserver agreement was very
good ( = 0.86, reviewer 1 vs reviewer 2; 0.81, reviewer 2 vs 3; 0.88,
reviewer 1 vs 3).
|
The diagnostic performance of visual analysis and quantitative analysis differed significantly (p = 0.01) because quantitative analysis improved the characterization of the tumors that presented equivocal subjective conspicuity at the late phase for the reviewers (e.g., -1 for reviewer 1 and 0 for reviewers 2 and 3) and consequently were incorrectly characterized as benign or malignant at visual analysis. Quantitative analysis changed the initial diagnosis proposed after visual analysis in 34 cases (16 hepatocellular carcinomas, two metastases, two hemangiomas, 12 macroregenerative nodules, and two hepatocellular adenomas) for reviewer 1, in 20 cases (12 hepatocellular carcinomas, one hemangioma, five macroregenerative nodules, and two hepatocellular adenomas) for reviewer 2, and in 21 cases (15 hepatocellular carcinomas, four macroregenerative nodules, and two hepatocellular adenomas) for reviewer 3.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionAPPENDIX 1: Imaging Criteria...References |
|---|
3 µ) that are filled by air or gas with low solubility in the blood and
covered by a shell of biologically inert material (albumin or
phospholipids). SonoVue is a sulfur hexafluoride-filled microbubble contrast agent consisting of bubbles encapsulated by a flexible shell of phospholipids. According to the procedure suggested by the pharmaceutical company (Bracco), a white suspension of microbubble is obtained from 25 mg of lyophilisate powder by adding 5 mL of physiologic saline to the powder, followed by hand agitation [7, 21, 22]. The obtained microbubbles present a density of 2 x 108 microbubbles per milliliter and a mean diameter of 3 µ (90% of microbubbles < 8 µ) and are stable in the vial for a few hours (< 6 hr). Although not yet allowed for use in the United States, sulfur hexafluoride-filled microbubbles are licensed in most European countries for abdominal and vascular imaging and present a strong nonlinear harmonic response when insonated by low transmit power [7, 21, 22].
High transmit power is used after air-filled microbubble (e.g., Levovist) injection because the galactose shell of air-filled microbubbles presents very low harmonic behavior [7] and the only way to produce harmonics is to cause extensive bubble destruction with the emission of a wideband signal. On the other hand, low transmit power is used to insonate perfluorocarbon or sulfur hexafluoride-filled microbubbles. Low-transmit-power insonation causes the selective resonance of microbubbles with harmonics emission and the effective suppression of the background signal from stationary tissues [18]. Low-transmit-power insonation allows continuous scanning with the possibility to assess the tumor contrast enhancement in real time, even though a lower signal is produced in comparison with high-transmit-power insonation [18].
After contrast administration, visual analysis is the simplest method by which to assess liver tumors, despite the disadvantages of interobserver variability and low reproducibility of results [5, 16]. Quantitative analysis of sonography videotape intensity is more complex, although it is expected to provide more objective, reliable, and reproducible results [13, 17].
In this study, visual analysis and quantitative analysis were compared in characterizing liver tumors. As previously shown [3-6, 8-15], different enhancement patterns were observed in malignant tumors (rimlike in metastases) and benign tumors (peripheral nodular in hemangiomas and spoke-wheel-shaped in focal nodular hyperplasia) after microbubble contrast agent injection. Malignant and benign liver tumors also revealed common enhancement patterns (diffuse heterogeneous in both hepatocellular carcinomas and hepatocellular adenomas; diffuse homogeneous in both hepatocellular carcinomas and fibrous nodular hyperplasias; absent enhancement in both metastases and avascular thrombotic liver hemangiomas). For these reasons, enhancement patterns are not sufficient for characterizing liver tumors, and the analysis of the appearance of the liver tumors at the late phase is fundamental because malignant tumors prevalently showed microbubbles washout, whereas benign tumors prevalently showed a persistent microbubble uptake [13-15].
Diagnostic performance of visual analysis revealed very good interobserver agreement, even though observers present with different levels of experience with contrast-enhanced sonography. This result is explained by the characteristic contrast enhancement patterns observed in most examined tumors [3-6]. The residual interobserver variability in diagnostic performance was determined by the larger interobserver variability in the assessment of subjective tumor conspicuity at the late phase.
Quantitative analysis allowed improvement of the overall diagnostic performance compared with visual analysis due to more accurate assessment of tumor conspicuity at the late phase. This is probably because visual analysis is often penalized by the fact that the eyes of observers usually focus on a particular portion of the lesion, whereas quantitative analysis allows comparison of the echogenicity of the tumor and of the adjacent liver in a more global and reproducible manner. Quantitative analysis also allows the polarity of conspicuity—from hypovascular to isovascular or from isovascular to hypovascular—to be changed in many tumors with a consequent correction of the diagnosis proposed after visual analysis. This is also emphasized by the lower overlap between conspicuity data of malignant and of benign tumors in quantitative analysis compared with visual analysis.
In this study, a mathematic formula was proposed to calculate objective contrast-enhanced tumor conspicuity from gray-scale images obtained using the pulse inversion mode. This formula is original, even though a similar formula was developed and used in previous studies [13] in which tumor conspicuity was quantified after insonation by stimulated acoustic emission, which is a color Doppler contrast-specific mode.
Quantitative analysis of sonography videotape intensity depends on the several stages of postprocessing performed in sonography equipment for image videotape presentation, including log compression, that modify the original features of the signal. Numerous other factors may affect the gray-scale appearance of tumors and normal livers, such as tumor depth and signal absorption; the pre- and postprocessing settings on the sonography scanner, such as echo-signal gain, grayscale mapping, transmit power, and frame rate [23]; the microbubbles injection rate, volume, and concentration; and scanning delay time [24, 25]. These limitations are always present in the quantitative analysis of sonography videotape intensity [13, 17] and in this study were reduced by using one sonography unit and normalizing the gray-scale intensity measured in liver tumors to that measured in the adjacent liver.
Objective tumor conspicuity of 0 at the late phase—equal intensity of the tumor and liver after contrast injection—was selected as the cutoff to differentiate malignant from benign liver tumors because microbubbles washout was previously described as prevalent in malignant lesions, whereas persistent microbubbles uptake has been described in benign lesions [3-5, 8-15]. Our study confirmed those results, even though sulfur hexafluoride-filled microbubbles do not present a late liver-specific phase as air-filled microbubbles do, probably because of the different shell composition with a lower affinity for liver sinusoids [26]. The persistent bubble uptake in benign lesions observed in the present study was probably determined by the similar vascular network in terms of vessels structure and flow velocity in tumors and adjacent liver—for example, in focal nodular hyperplasia—or by the persistent microbubbles pooling in vessels with slow flow—for example, in hemangiomas. The different vascular architecture compared with adjacent liver probably determined microbubbles washout at the late phase in malignant tumors.
A clear overlap between benign and malignant tumors persists even after
quantitative analysis because some liver tumors are prevalent in patients with
chronic liver disease or cirrhosis (or both). In the present study, 30% of the
hepatocellular carcinomas revealed a conspicuity of 0 at the late phase,
which is similar to benign tumors. This finding confirms those of previous
studies [5,
11] in which the percentage of
isoto hypervascular hepatocellular carcinomas at the late phase was reported
to range from 33% to 38%, and it was determined by the variable grade of
microbubbles uptake in different hepatocellular carcinomas, which correlates
with the grade of tumor differentiation
[27,
28], or by the absence of
liver-specific uptake for sulfur hexafluoride-filled microbubbles
[6,
7,
12].
On the other hand, some low- or high-grade dysplastic macroregenerative nodules revealed an objective conspicuity of < 0 as for malignant tumors. The atypical conspicuity of these tumors at the late phase resulted in an incorrect diagnosis even after quantitative analysis. According to these findings, any tumor identified in a patient with chronic liver disease or cirrhosis should be analyzed with high suspicion. Each tumor with a hypovascular appearance at the late phase after microbubble contrast agent injection must be considered malignant in both the normal and the cirrhotic liver [5, 29, 30]. Moreover, besides tumor conspicuity, tumor appearance at the arterial phase is fundamental in patients with chronic liver disease because malignant tumors—namely, hepatocellular carcinomas—appear prevalently hypervascular, whereas benign tumors—namely, macroregenerative nodules and hemangiomas—appear prevalently hypoor isovascular [29, 30]. However, distinction between dysplastic macroregenerative nodules and hypovascular well-differentiated hepatocellular carcinomas remains difficult both by imaging techniques and at histology [31].
Besides tumors in patients with chronic liver disease, thrombotic liver hemangiomas showed an atypical appearance at the late phase with a conspicuity of < 0 and a hypovascular appearance. This finding was determined by the absence of persistent microbubbles uptake in thrombotic avascular tumors. Intrahepatic cholangiocarcinoma appeared hypovascular to the adjacent liver at the late phase, similar to the other malignant tumors. This tumor may retain contrast material on delayed contrast-enhanced CT or MR images because contrast material in the interstitial spaces of the tumor diffuses slowly [32]. The appearance of cholangiocarcinoma on contrast-enhanced sonography was due to the peculiar properties of microbubble contrast agents, which do not leak in the fibrous stroma [7] and determine the persistent hypovascular appearance of tumor.
The first limitation of this study was the fact that a single liver tumor was assessed on contrast-enhanced sonography because the transducer must remain still during scanning. This protocol was selected for consistency in the comparison of visual analysis and quantitative analysis because the assessment of more tumors in the same patients could make this comparison difficult. In practice, additional microbubble injections would be necessary to characterize adjunctive liver tumors, and this factor also must be considered a disadvantage in comparison with multiphasic CT and MRI, which allow simultaneous characterization of multiple liver tumors of similar or different nature in the same patient.
The second limitation was the lack of histologic correlation in slightly more than one third of the tumors. All liver tumors that were not examined at histology were characterized by strictly established diagnostic criteria consisting of typical contrast enhancement patterns observed on multiphase contrast-enhanced CT or MRI. Lesions with an atypical appearance or incompletely characterized on CT or MRI were all biopsied or were surgically removed for histologic analysis.
In conclusion, quantitative analysis revealed higher diagnostic performance than visual analysis for the characterization of liver tumors insonated at low transmit power after microbubble contrast agent injection.
APPENDIX 1: Imaging Criteria for the Diagnosis of 69 Tumors That Were Not Biopsied for Clinical Reasons or Because of the High Probability of Hemangioma
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionAPPENDIX 1: Imaging Criteria...References |
|---|
Metastasis
Hemangioma
Focal nodular hyperplasia
Intrahepatic cholangiocarcinomas, macroregenerative nodules, and hepatocellular adenomas
• Histology result
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionAPPENDIX 1: Imaging Criteria...References |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  S. R. Wilson, H.-J. Jang, T. K. Kim, and P. N. Burns Diagnosis of Focal Liver Masses on Ultrasonography: Comparison of Unenhanced and Contrast-Enhanced Scans J. Ultrasound Med., June 1, 2007; 26(6): 775 - 787. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||
