| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Medical Imaging, Institut Gustave Roussy, 39 rue Camille
Desmoulins, Villejuif 94805, France.
2 Department of Gastroenterology, Institut Gustave Roussy, Villejuif 94805,
France.
3 Department of Surgery, Institut Gustave Roussy, Villejuif 94805, France.
Received March 21, 2007;
accepted after revision September 18, 2007.
Presented at the 2005 annual meeting of the Radiological Society of North
America, Chicago, IL.
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. This prospective study included 116 patients. All patients underwent a preoperative conventional sonography examination followed by sonography after injection of contrast agent combined with the use of perfusion software (vascular recognition imaging or pulse subtraction imaging). Histopathologic analysis was the reference standard used to compare the diagnostic value of baseline sonography versus contrast-enhanced sonography.
RESULTS. Eighty-two patients underwent hepatic surgery, 31 did not because of disseminated lesions, and the remaining three patients did not meet inclusion criteria. Three hundred six surgically proven lesions were taken into account for comparison of the two techniques: 147 were detected on baseline sonography and 177 on contrast-enhanced sonography. Histopathologic analysis revealed 233 malignant and 73 benign lesions. Sensitivity and specificity were improved on contrast-enhanced sonography compared with baseline sonography for the detection of malignant lesions: 68.7% versus 58.8% and 67% versus 50.7%, respectively. Contrast-enhanced sonography detected 23 additional malignant lesions that had been seen as lacuna at the portal venous phase and characterized as 19 benign nodules, thus improving the performance of sonography in 13.7% of the cases.
CONCLUSION. Contrast injection improved the sensitivity and specificity of baseline sonography and should be performed in routine practice if hepatic surgery is being considered for the management of liver lesions.
Keywords: contrast-enhanced sonography • hepatic lesions • intraoperative sonography • liver cancer • liver disease
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The prospect of new indications for sonography is becoming a reality, especially in surgical oncology where the treatment strategy, and consequently patient survival, is dependent on imaging-based staging. In oncologic digestive surgery, the operative strategy is currently based on slice imaging data (MRI, CT, PET, or a combination of these techniques) to detect and localize lesions. Over the past several years, these examinations have been increasingly complemented in our institute by B-mode sonography to guide intraoperative examinations [3, 4]. In this study, we compared sonographic exploration—baseline sonography and contrast-enhanced sonography—for the detection of hepatic metastases before surgery in cancer patients. The analysis was based on the pathologic specimen.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Sonography
All sonographic explorations were performed using a sonography scanner
(Aplio, Toshiba Medical Systems) with a 2- to 6-MHz wideband probe. The
operators' experience in contrast-enhanced sonography ranged from 6 months to
6 years. The operators were not blinded to the results of the other imaging
examinations (CT, MRI, PET, or a combination of these studies), so they could
confirm to the surgeon whether the lesions described on slice imaging were
visible on abdominal sonography to guide the intraoperative sonographic
examination.
During the same examination, analysis was conducted first in B mode, followed by injection of an IV bolus of 2.4 mL of aqueous suspension of phospholipid-stabilized microbubbles filled with sulfur hexafluoride (SonoVue, Bracco), and was possibly repeated because detection and characterization had to be performed at the same time during this real-time exploration.
On baseline sonography, the liver was explored as exhaustively as possible—that is, segment by segment—to detect lesions. Several tumor characteristics were noted: size; echostructure, defined as hypo-, iso-, or hyperechoic, compared with the adjacent healthy parenchyma; location according to Couinaud's concept of liver segmentation; and distance from the large vessels (i.e., portal and suprahepatic veins, vena cava) to guide the surgical procedure.
On contrast-enhanced sonography, lesions were detected as focally enhanced spots at the arterial phase, as lacunae or enhancement defects at the portal and delayed phases, or both. The benign or malignant nature of nodules was assessed during the portal venous and delayed phases: At these times, a hypovascularized lesion (lacuna or hypoechoic area) compared with the adjacent parenchyma was considered probably malignant, and conversely, persistence of enhancement in a lesion was considered benign.
SonoVue, a second-generation sonography contrast medium, was used according to the manufacturer's instructions—that is, by adding 5 mL of saline to a vial containing the sterile lyophilized powder to form a microbubble suspension. The nonlinear vibration properties of these microbubbles enable the use of a low mechanical index (MI) (< 0.1), which preserves the microbubbles throughout their half-life in microvessels and then allows real-time exploration of liver vasculature.
The perfusion software used was vascular recognition imaging (VRI) or pulse subtraction imaging (PSI) modesdeveloped by Toshiba. The VRI and PSI modes are contrast medium imaging modes designed for low-acoustic-power sonography emission. They are based on the same principle: detecting second-harmonic signals yielded by microbubbles circulating in vessels.
VRI software enables color functional images and gray-scale anatomic images to be superimposed on the visualization screen. The flow rate of the contrast medium is detected in the Doppler mode: Signals from contrast medium moving at a very low speed are displayed in green and signals from contrast medium flowing with the blood are displayed in blue or red. Conversely, the PSI mode produces a black-and-white image by suppressing the parenchymal echo components and by enhancing the signals from the contrast medium.
Sonographic findings before and after injection of contrast medium were noted on a liver diagram showing Couinaud's segmentation of the liver.
Each examination was recorded in the DICOM mode on a videotape and, in addition, on our PACS system.
Other Explorations
Imaging data—All patients underwent at least one of the
slice imaging techniques (CT, MRI, PET, or a combination of these techniques)
for reference standard images no more than 4 weeks before the operation. These
examinations were performed at different institutions, but all involved liver
analysis with contrast medium injection.
Operating data—During surgery, the parenchyma of the whole liver was explored on the basis of imaging findings (i.e., slice imaging techniques and sonography) by an experienced surgeon inspecting and palpating the organ and using intraoperative sonography (Ecocee, Toshiba). Unremoved lesions, including lesions treated by radiofrequency ablation, were biopsied.
Analysis
The contrast-enhanced sonography results were compared with those of
baseline sonography under similar visualization conditions. Then, each patient
was used as his or her own control to analyze how detection was improved with
contrast-enhanced sonography.
The sonographic findings (number and size of nodules, segmental site of nodules, and probability that nodules were malignant or benign) in all patients were correlated retrospectively with the histologic results. The predictive value of the presence or absence of a hypoechoic area for malignancy on contrast-enhanced sonography and the sensitivity and specificity of sonography for malignancy before and after injection of contrast medium were then calculated.
The limitations of contrast-enhanced sonography are discussed by analyzing technical limitations and calculating the false-positive and false-negative results of this technique.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Three hundred six nodules were histologically proven. Using histologic findings as the reference standard, we observed that some had previously been detected on imaging and others had been detected only during surgery: Sonography detected 204 lesions (66.7%) (baseline sonography, 147; contrast-enhanced sonography, 177), 205 (67%) were detected by at least one of the other imaging methods (CT, MRI, or PET), and 278 (90.8%) were detected during the intraoperative analysis through inspection, palpation, and intra-operative sonography.
There were, on average, 3.9 lesions per patient (range, 0–15 lesions per patient; SD, 3.6 lesions per patient). The average size of the nodules was 19.7 mm (range, 3–160 mm; SD, 20.6 mm).
The final histologic diagnosis was 233 (76.1%) malignant lesions and 73 (23.9%) benign lesions. The majority of malignant lesions were metastases from colorectal adenocarcinoma (60.5%) and seven lesions (3%) were HCC (Tables 1 and 2).
|
|
Diagnostic Improvement with Contrast-Enhanced Sonography
Enhancement defects (hypoechoic areas) on portal and delayed phase
images—Malignant lesions were depicted as hypoechoic areas on
portal venous phase images in 65.7% of cases (153/233). One hundred thirty
cases had already been detected on baseline sonography and 23 were additional
unknown nodules (Tables 3 and
4).
|
|
Eight (34.8%) of these 23 additional malignant lesions were small (< 1 cm) (Figs. 1A, 1B, 1C, 1D, 2A, 2B, 3A, 3B, 3C). Among the 177 hypoechoic areas, 153 corresponded to malignant disease at the final histo-logic diagnosis. The positive predictive value of a hypoechoic area sign for malignancy was 86.4% and the negative predictive value was 38%.
|
|
|
|
|
|
|
|
|
Finally, by detecting more lesions and by characterizing nodules as benign, contrast-enhanced sonography afforded a diagnostic benefit in 13.7% of cases (42/306) (23 additional lesions and 19 nodules characterized as benign) and improved the sensitivity and specificity of baseline sonography from 58.8% to 68.7% and from 50.7% to 67%, respectively.
Limitations of Contrast-Enhanced Sonography
Some malignant lesions were not seen on sonography (i.e., neither at
baseline sonography nor on contrast-enhanced sonography). In addition, some
nodules depicted on baseline sonography proved to be false-positives or
false-negatives after contrast injection.
False-positives—Twenty-four (38.2%) proven benign lesions were detected as hypoechoic areas after injection of contrast medium at portal venous phase imaging. They corresponded to fibrosis or sequelae resulting from metastases sterilized by chemotherapy, focal steatosis or focal healthy islet, hemangioma, hamartoma, or FNH (Table 5 and Fig. 4A, 4B).
|
|
|
False-negatives—Eighty malignant lesions were false-negatives of sonography. Seventy-three (31.3%) of the 233 histologically proven malignant lesions were not detected by either sonographic technique, baseline sono graphy or contrast-enhanced sonography: 50 lesions (68.5%) were subcentimeter and 37 (50.7%) were not detected by other imaging techniques.
Seven malignant nodules were misinterpreted on contrast-enhanced sonography
because they were depicted on baseline sonography but were not visible or did
not sustain enhancement after injection. They corresponded to metastasis from
colorectal carcinoma or endocrine tumor: two were subcentimeter and
hyperechoic on baseline sonography, and contrast uptake in the other five was
not analyzable, even though the lesions ranged from 13 to 30 mm, because the
lesions were deep-seated in a steatotic liver (
12 cm below the skin
surface).
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Differentiation of malignant lesions from benign lesions—Contrast-enhanced sonography is the subject of several recent studies that showed its usefulness in the characterization of focal hepatic lesions [1, 2, 9–15]. Although the techniques and contrast media used differ among the studies, similar enhancement patterns were described [1, 2, 9–15]. These studies have shown that contrast-enhanced sonography can differentiate malignant from benign lesions; for example, the results of a multi-centric study of 142 patients by Bryant and al. [16] showed improved accuracy by 80–90% for contrast-enhanced sonography versus base line sonography. Most authors agree that the portal and delayed phases are important because benign lesions are predominantly hyper- or isoechoic relative to the adjacent parenchyma, whereas malignant lesions are predominantly hypoechoic during those phases [1, 9, 13, 16]. Furthermore, in their recent study of a series of 152 patients, Nicolau et al. [17] concluded that analysis of the three vascular phases is not more accurate than analysis of the late phase alone for differentiating malignant from benign focal lesions in patients without chronic liver disease.
Animal and clinical studies have already shown a significant difference in uptake quantification parameters between tumors and perineoplastic parenchyma [18, 19]. Histopathologic analysis showed that sonography microbubble-like structures were internalized in Kupffer cells by macrophages located in the liver sinusoids. This was assessed by correlating MR images using superparamagnetic iron oxide (SPIO), a liver-specific agent [20–23]. Thus, the detectability of small hepatic tumors was greatly increased due to the absence of Kupffer cells in tumor tissue. In addition, the low contrast uptake in malignant lesions during the portal venous to late phase can be explained by the vascular particularities of tumors that are mainly supplied by the hepatic artery.
Detection of more lesions—Data regarding characterization patterns are complementary to detection. The usefulness of contrast-enhanced sonography in detecting liver lesions has been evaluated in clinical studies and was established in guidelines by the European Federation of Societies for Ultrasound in Medicine and Biology (EFSUMB) [24–28].
In our study, we used binary analysis during the portal venous and delayed phases (i.e., hypoechoic area = a malignant lesion and, conversely, no hypoechoic area = no malignant lesion) because it allowed the most exhaustive analysis of the whole liver during short-duration real-time imaging, which corresponds to the lifetime of microbubbles in vessels. According to data in the recent literature, two features were observed during these phases: first, lesion-to-liver contrast was increased [1, 11, 12, 27, 29, 30]; and, second, focal malignant liver lesions more often corresponded to a hypoechoic area (hypoechoic area) [9, 11, 16].
Then, subcentimeter and poorly visible or invisible malignant lesions on B-mode imaging (e.g., isoechoic nodules or small subcentimeter lesions) may become detectable because they should appear as obvious defects. Intraoperative sonography studies confirmed these results and showed that contrast-enhanced intraoperative sonography improved the detection of small and isoechoic lesions compared with intraoperative B-mode sonography [31–34]; intraoperative sonography is reported to be more effective than abdominal sonography for the detection of superficial nodules. In our patients, 30 additional nodules were detected after contrast medium injection (76.7% were malignant at histologic analysis). The positive predictive value of the hypoechoic area sign for malignancy was 86.4%, which is consistent with the results of Brannigan et al. [11] (91%).
In patients with cancer treated with chemotherapy, the liver is often fatty and heterogeneous and doubtful nodulelike findings are usually depicted on baseline sonography. Injection of contrast medium may improve the specificity of sonography by ruling out some of these findings by blurring them during the portal venous phase. They represented 19 of our benign lesions (26%).
Thus, the injection of contrast medium in our patients improved the performance of conventional sonography in 13.7% of the cases. Sensitivity and specificity for the detection of individual malignant lesions was improved from 58.8% to 68.7% and from 50.7% to 67%, respectively, with contrast-enhanced sonography compared with baseline sonography. This increased conspicuity is in accordance with the results of a few recent studies reporting that contrast-enhanced sonography is superior to baseline sonography in lesion detection, but their values were often higher than ours (87–90% for sensitivity and 84–88% for specificity). We believe that this discrepancy is perhaps because of the baseline reference standard used—namely, the results of slice imaging techniques (MRI or CT) [1, 16, 25–27, 29, 30]—and not histologic findings, as in our study.
Technical Limitations
Correlating sonographic findings with the histologic specimen depended on
the surgeon's analysis. The risk of mistranslation exists in any study
comparing any imaging technique with surgical findings, and this risk is
probably increased when sonography is used because it is operator-dependent
and because reproducibility is poorer. Nevertheless, in this study, the use of
several lesion characteristics—location according to Couinaud's
segmentation of the liver and in relation to large vessels and size—and
noting results on a diagram probably minimized this risk without ruling it
out.
The morphotype and liver echostructure (steatosis, hemochromatosis,
fibrosis) are well-known limitations in sonography examinations. In VRI mode,
as with other sonography contrast-specific software, the detection of the
signal yielded by microbubbles is limited for deep-seated lesions, especially
in steatotic livers, because of strong ultrasound beam attenuation. In our
patients, five of seven malignant lesions previously detected on baseline
sonography were not analyzable after injection. They were located in steatotic
livers in segment VII ( 12 cm below the skin), and no conclusion could
be drawn objectively on contrast-enhanced sonography despite their size
(13–30 mm) and even though they were obviously malignant on baseline
sonography. The limitation here therefore seems to be more patient-dependent
than truly operator-dependent, and this limitation has always been classically
described in sonography. In practice, in patients readily evaluated with
sonography, intra- and interobserver variations are probably cancelled out
because the two phases of examination (before and after injection of contrast
medium) are conducted carefully so that the previous scans acquired in the B
mode can be correlated with those obtained after injection. This impression
obviously needs to be confirmed by intra- and interoperator comparisons, which
were not performed in this study.
Diagnostic Limitations
In some cases, the principle that a defect in the portal venous-to-late
phase is more likely to be a malignant lesion may be contradicted. Indeed,
some malignant lesions may not appear as a hypoechoic area and some benign
lesions may. These findings may be related to the size of the lesion, its
histologic features, or its sono-graphic appearance on baseline
sonography.
False-negatives and undetected lesions—The small size of lesions remains a limitation of sonography, although the injection of contrast agent improves their detection: Small lesions (< 1 cm) represented 34.8% of our additional malignant lesions, but on the other hand, they accounted for 68.5% of the undetected ones.
Moreover, misinterpretation of sonographic findings may be linked to the echostructure of the lesion. Hyperechoic malignant lesions (e.g., metastasis from colorectal carcinoma or endocrine tumors) would not appear as evident hypoechoic areas after injection because their brightness produces high-signal-intensity artifacts on low-MI imaging before injection. In the present study, two metastases from colorectal carcinoma previously detected as hyperechoic nodules on B-mode imaging were not depicted as hypoechoic areas after contrast injection. In addition, they did not appear during the arterial phase because they were poorly vascularized. In contrast, this limitation of sonography is inconsequential for the diagnosis of hypervascularized metastases from endocrine tumors because they usually exhibit strong focal enhancement during the arterial phase.
HCCs exhibit particular histologic features and enhancement behavior. The behavior of HCC after contrast injection is reported to be variable, especially during the portal and late phases, and seems to be associated with histopathologic grade [16, 35–37] and probably with liver cirrhosis. In particular, well-differentiated HCCs are reported to exhibit initial homogeneous enhancement that persists during the portal venous and late phases. Conversely, in a recent study by Hong et al. [9] concerning 87 cases of HCC, the association of the two criteria—enhancement during the arterial phase and a defect during the portal phase—yielded a sensitivity of 92% and a specificity of 86.6% for the diagnosis of HCC. In our study, all the lesions were poorly differentiated or undifferentiated at histologic analysis and exhibited a more or less heterogeneous defect during the portal phase, which allowed us to confirm their malignant nature.
False-positives—The false-positives on contrast-enhanced sonography were lesions detected as hypoechoic areas after contrast injection that proved to be benign at histologic analysis. Benign lesions seem to frequently give rise to errors, but in our study this risk was probably reduced because the study population was patients who had a known cancer with a strong predominance of metastatic lesions (233 malignant vs 73 benign nodules). The false-positives corresponded to benign lesions depicted as hypoechoic nodules on baseline sonography with insufficient enhancement during portal and delayed phases on contrast-enhanced sonography.
Hypoechoic benign lesions, such as islets of steatosis or, conversely, of normal liver in steatotic livers or fibrous scars from metastases, that were sterilized by chemotherapy may appear as hypoechoic areas during the portal-to-late phase (32.8% of the benign lesions in our study). The microvascular changes and the impedance gap between these hypoechoic areas and the adjacent parenchyma are probably both involved in making them appear as a defect on contrast-enhanced sonography.
Lesions associated with steatosis are frequent and do not usually pose diagnostic problems on baseline sonography because they are typically located in the liver hilum and are devoid of a mass effect. However, in some cases, they may remain hypoechoic on portal phase images; their distinct histologic features and perfusion particularities are also found on CT. In the case of focal fibrous lesions, such as sterilized metastases, an excessive diagnosis of malignancy does not penalize the technique and does not change the management of patients because these lesions are usually found at the intraoperative stage by the surgeon by palpating the liver, using intraoperative sonography, or both; the surgeon will always remove them to check whether they have been sterilized.
Some hemangiomas exhibit a singular feature: They may produce insufficient enhancement leading to false-positive results. Usually hemangiomas are well characterized on baseline sonography alone, but in our population of cancer patients with postchemotherapy changes in the liver, the diagnosis was not easy. Furthermore, in some cases, the differential diagnosis between hemangioma and malignant tumors (e.g., metastasis from endocrine and thyroid cancers or a hyperechoic metastasis from colorectal carcinoma) can be difficult.
Misinterpretation of hemangioma can be linked to the behavior of such lesions on contrast-enhanced sonography. Indeed, some atypical hemangiomas are now identified on contrast-enhanced sonography because of the absence of central filling due to a thrombotic area, making them appear as hypoechoic areas on late phase images [11, 38]. Some authors have proposed considering a decrease in lesion diameter after contrast injection as a way of differentiating hemangioma from malignant nodules. However, cases of such misinterpretation have been reported in other studies [1, 12], and one of our false-positive diagnoses (5.9%) of malignancy in a patient with colorectal carcinoma is a case in point.
In conclusion, preoperative imaging before hepatectomy should be as exhaustive as possible to allow surgeons to plan resection under the best conditions. Sonography should probably not replace slice imaging techniques such as CT and MRI, but the more frequent use of intraoperative sonography by the surgeon will lead to more widespread use of complementary preoperative abdominal sonographic examinations.
In this study, contrast-enhanced sonography improved detection of malignant liver lesions compared with baseline sonography by depicting more lesions and revealing undetermined lesions, with a benefit in 13.7% of the cases. As with other imaging techniques, contrast-enhanced sonography requires a learning curve. Contrast-enhanced sonography is a low-cost technique with no radiation exposure. It should be used in routine practice to improve sonographic detection of liver lesions before surgical resection.
Acknowledgments
The authors are grateful to Lorna Saint Ange for editing.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
