| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1
Bracco, Medical and Regulatory Affairs, Via Egidio Folli 50, 20134 Milan,
Italy.
2
Department of Radiology, Scientific Institute S. Raffaele, University
Hospital, Via Olgettina 60, 20132 Milan, Italy.
3
Department of Radiology, Hospital Universitario Dr. Peset, Valencia,
Spain.
4
Il Servizio di Radiologia, Ente Ospedaliero «Spedali Civili»,
Piazzale Ospedale 1, 25100 Brescia, Italy.
5
Instituto di Radiologia, Policlinico «A. Gemelli», Largo A.
Gemelli 8, 00168 Roma, Italy.
6
MR Imaging Centre, Princess Grace Hospital, Ave. Pasteur, 98000 Monaco,
Principaute de Monaco.
7
Ludwig-Maximilians-Universität-München,
Institut für Radiologische Klinik, Klinikum
Marchioninistr. 15, 81377 München,
Germany.
8
Westfälische
Wilhelms-Universität
Münster, Institut
für Klinische Radiologie,
Röntgendiagnostik, A.-Schweitzer-Str. 33, 48129
Münster, Germany.
9
Institut für
Röntgendiagnostik, Medizinische
Fakultät der
Humboldt-Universität,
Charité, Schumannstr. 20/21, 10098 Berlin,
Germany.
Received December 6, 1999;
accepted after revision March 24, 2000.
Address correspondence to A. Spinazzi.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Eighty-six patients were imaged before gadobenate dimeglumine administration, immediately after the 2 mL/sec bolus administration of a 0.05 mmol/kg dose (dynamic imaging), and at 60-120 min after the IV infusion at 10 mL/min of a further 0.05 nmol/kg dose (delayed imaging). The accuracy for lesion characterization was assessed for a total of 107 lesions. Sensitivity for lesion detection was assessed for a total of 149 lesions detected on either intra-operative sonography, iodized oil CT, CT during arterial portography, or follow-up contrast-enhanced CT as the gold standard.
RESULTS. The accuracy in differentiating benign from malignant liver lesions increased from 75% and 82% (the findings of two observers) on unenhanced images alone, to 89% and 80% on dynamic images alone (p < 0.001, p = 0.8), and to 90.7% when combining the unenhanced and dynamic image sets (p < 0.001, p = 0.023). Delayed images did not further improve accuracy (90% and 91%; p = 0.002, p < 0.05). A similar trend was apparent in terms of accuracy for specific diagnosis: values ranged from 49% and 62% on unenhanced images alone, to 76% and 70% on combined unenhanced and dynamic images (p < 0.001, p = 0.06), and to 75% and 70% on inclusion of delayed images (p < 0.001, p = 0.12). The sensitivity for lesion detection increased from 77% and 81% on unenhanced images alone, to 87% and 85% on combined unenhanced and dynamic images (p = 0.001, p = 0.267), and to 92% and 89% when all images were considered (p < 0.001, p = 0.01).
CONCLUSION. Contrast-enhanced dynamic MR imaging with gadobenate dimeglumine significantly increases sensitivity and accuracy over unenhanced imaging for the characterization of focal hepatic lesions, and delayed MR imaging contributes to the improved detection of lesions.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Gadobenate dimeglumine (MultiHance; Bracco, Milano, Italy) is a gadoliniumbased contrast agent that differs from conventional gadolinium agents in showing an almost twofold higher relaxivity in blood [5] and in the fact that 3-5% of the injected dose is taken up into functioning hepatocytes and excreted in the bile [6]. Although the overall pharmacokinetic profile of gadobenate dimeglumine shows no appreciable difference from the profiles of other agents [6,7,8], the fraction taken up by hepatocytes has been shown to produce a marked and long-lasting enhancement of the normal liver parenchyma [9, 10] that results in significantly increased sensitivity for liver lesion detection, particularly of small (<1 cm) metastases, when T1-weighted MR images are acquired between 40 and 120 min after administration [11]. Recently, a phase III clinical study showed the usefulness of gadobenate dimeglumine at 0.05 mmol/kg of body weight for hepatic lesion characterization during the dynamic phase of contrast enhancement [12].
The primary objective of our study was to assess the accuracy with which a 0.05 mmol/kg dose of gadobenate dimeglumine on dynamic T1-weighted MR imaging can differentiate benign from malignant focal liver lesions and can facilitate the correct specific diagnosis of detected, histologically proven focal lesions. Our secondary objective was to determine the contribution of delayed (60-120 min after injection) MR imaging to both the characterization and detection of focal lesions.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Patients
Eighty-six patients (50 men and 36 women; mean age, 59.6 years; range,
27-79 years) were enrolled between August 23, 1994 and August 14, 1995.
Patients were enrolled if they were older than 18 years, had been referred for
MR imaging for diagnosis or follow-up of suspected or known focal liver
lesions, and had given written informed consent. Patients were excluded if
they were pregnant or nursing, had a history of hypersensitivity to any metals
or chelates of gadolinium, had undergone treatment with any other IV contrast
material in the 48 hr preceding the MR examination, had congestive heart
failure, or were otherwise contraindicated for MR imaging. For 78 patients,
definite proof of at least one focal liver lesion from either biopsy or
surgery was available. Overall, 46 patients had a solitary lesion, 15 patients
had two lesions each, 10 patients had three lesions, one patient had four
lesions, two patients had five lesions, one patient had six lesions, one
patient had seven lesions, and two patients had eight lesions. For patients
with multiple lesions who underwent biopsy of a single lesion, the
histologically proven lesion was included in the analysis of lesion
characterization. Eight patients with no liver lesions at surgery but who had
undergone intraoperative sonography of the liver during surgery for pancreatic
cancer were also included in the study to ensure a more accurate and reliable
assessment.
Histopathologic specimens were obtained between 2 and 15 days after the MR examination by tumor resection in 41 patients and by sonographically guided biopsy in 37 patients. Overall, 149 lesions were detected, with 107 lesions assessed histologically. The malignant lesions included 38 examples of hepatocellular carcinoma (HCC), of which 21 were present in cirrhotic livers; 11 were examples of cholangiocellular carcinoma; 42 were metastases (37 colorectal, two breast, one stomach, one pancreatic, and one of unknown origin); one non-Hodgkin's lymphoma; and one malignant fibrous histiocytoma. For one lesion that was defined as definitely malignant, the histology was unclear. The benign lesions comprised 10 examples of hemangioma, one of focal nodular hyperplasia, one of hepatocellular adenoma, and one of hemangioendothelioma. The benignity of these lesions was determined in each case after surgical resection.
Contrast Material
All patients were administered 0.2 mL/kg (0.1 mmol/kg) of a 0.5-mol/L
solution of gadobenate dimeglumine as a biphasic IV injection of half the dose
as a 2 mL/sec bolus (administered manually), followed 10 min later (after the
acquisition of dynamic phase images) by the other half as a 10 mL/min
infusion. The split-dose approach was adopted because the relaxivity of
gadobenate dimeglumine in blood is roughly twice that of conventional
gadolinium chelates [5],
suggesting that only half the standard dose of 0.1 mmol/kg would be needed for
effective dynamic phase imaging in which contrast is based solely on the
exploitation of differences in blood flow between lesions and normal
parenchyma. For delayed phase imaging, a dose of 0.1 mmol/kg was considered
appropriate because quantitative data revealed better liver parenchyma signal
intensity enhancement with this dose than with a dose of 0.05 mmol/kg
[10].
Imaging Procedures
All imaging centers were equipped with superconducting clinical imagers
operating at 0.5 T (Vectra; General Electric Medical Systems, Paris, France:
one center, 21 patients), 1.0 T (Magnetom Expert; Siemens, Erlangen, Germany:
one center, 13 patients), or 1.5 T (Magnetom, Siemens: three centers; Gyroscan
S15, Philips, Eindhoven, The Netherlands: one center; 52 patients overall).
The liver was imaged in the axial plane at all centers, and in each case the
section thickness (10 mm) and interslice gap (10-20%) was maintained on images
both before and after the injection of contrast material. Each patient was
examined with the same MR imaging protocol both before and at 60-120 min
(delayed scans) after the administration of the full 0.1 mmol/kg dose of
gadobenate dimeglumine. The window for the delayed acquisitions was selected
on the basis of earlier studies that showed that both the signal intensity in
the normal liver parenchyma and the lesion-liver contrast-to-noise ratio
peaked and remained relatively constant at between 40 and 120 min after
injection [6,
10].
The MR imaging protocol required acquisition of conventional T2-weighted or T2-weighted turbo spinecho images, T1-weighted spin-echo images, and breath-hold T1-weighted gradient-echo images. The imaging parameters differed among centers because of the different MR equipment available but were nevertheless kept constant throughout each study. All examinations were performed using a standard body coil. The following parameters were used: for T1-weighted spin-echo images, a TR range/TE range of 300-600/12-17, four acquisitions, a matrix of 128-224 x 256, and a field of view of 300-500 x 300-500; for conventional T2-weighted spin-echo images, 2000-3000/90-130, two acquisitions, a 128-192 x 256 matrix, and a 300-500 x 300-500 field of view; for T2-weighted turbo spin-echo sequences, TR/effective TE of 6000/112, an echo train length of four or eight, four acqusitions, a 210 x 256 matrix, and a 380 x 380 field of view; for breath-hold T1-weighted gradient-echo sequences, TR range/TE range of 45-150/4-13, a 45-75° flip angle, one or two acquisitions, a 128-256 x 256 matrix, and a 300-500 x 300-500 field of view. Predominantly in-phase images were acquired, with breath-hold times generally ranging between 12 and 20 sec. No fat-saturation techniques were used with any sequence.
The dynamic examination was performed using breath-hold T1-weighted gradient-echo sequences. Images were obtained at 30-45 sec (arterial phase), 70-90 sec (portal venous phase), 2-4 min (equilibrium phase), and 5-8 min (delayed phase) after the IV bolus injection of 0.05 mmol/kg of gadobenate dimeglumine, with the patient positioned in the bore of the magnet. Each dynamic sequence yielded a multislice package of two to five images at the level of the focal lesions as determined on contrast-enhanced CT and unenhanced MR imaging. Dynamic scanning of the eight patients with pancreatic cancer who had no liver lesions on intraoperative sonography was performed at the level of the pancreas and included the lower part of the liver.
A reference examination was performed for 56 patients within 1 month of the MR imaging examination. Specifically, intraoperative sonography was performed in 41 patients referred for surgery using a 5.0- to 7.5- MHz high-resolution intraoperative probe, CT during arterial portography was performed in two patients with colorectal cancer and liver metastases, and iodized oil-enhanced CT was performed in 15 patients referred for chemoembolization of an HCC. In cases in which multiple gold standard procedures were performed (n = 2), intraoperative sonography was considered first, followed by iodized oil-enhanced CT, and, finally, CT during arterial portography.
In patients for whom a reference examination was not available (n = 30), follow-up contrast-enhanced CT or MR imaging for at least 6 months was considered appropriate to serve as reference for the assessment of accuracy in lesion detection. For patients undergoing follow-up contrast-enhanced CT, a 7- to 10-mm thick section collimation was used with dynamic scanning performed after the administration of a 100- to 180-mL bolus of an iodinated contrast agent at 270-370 mg I/mL. Follow-up examinations for these patients occurred at 6-12 months (mean, 8 months). In all cases, the study reference was assumed to have given the definitive number and location of lesions per patient.
Image Selection
Selection of adequate MR images for each patient was made by investigators
other than the offsite blinded observers on the basis of the complete data
available on lesion histology using the dynamic scans as a reference for
anatomic liver coverage. The dynamic scans did not cover the entire liver but
only two to five liver slices encompassing the lesions of which the
hemodynamics were to be studied and characterized. A similar slice-selective
approach has been adopted previously by other authors for the purpose of image
analysis [13]. Although no
specific guidelines were adopted, the lesion to be studied was usually
selected because it was the only lesion evident on contrast-enhanced CT or
unenhanced MR imaging, or, when multiple lesions were apparent on unenhanced
images, because it was the most amenable in terms of size or location for
biopsy. All the histologically proven lesions were included in the analysis of
accuracy for lesion characterization and were present in the slices acquired
during dynamic imaging.
Using the dynamic scans as the reference for anatomic liver coverage, the corresponding MR images for both unenhanced and delayed contrast-enhanced scans were identified. Only these MR image slices were used for the blinded assessments. The remaining unenhanced and delayed contrast-enhanced MR images were obscured and were never displayed.
Image Analysis
Two independent experienced radiologists not affiliated with the enrollment
centers and unaware of all information on the patient population (both
clinical and radiologic) performed image assessment. Both unmatched and
matched assessment was performed. Four groups of interpreting sessions were
conducted by each observer. First, the unenhanced (T1-weighted spin-echo,
T1-weighted gradient-echo, and T2-weighted spin-echo sequences) and dynamic
image sets were interpreted separately in random order with a single
randomization schedule (unmatched assessment). Second, the image sets were
interpreted jointly, first unenhanced and dynamic image sets together and then
unenhanced, dynamic, and delayed image sets (matched assessments). The joint
approach to image interpretation was adopted to assess the diagnostic
performance achieved by combining the unenhanced images with the
contrast-enhanced images (as occurs in routine clinical practice).
Image assessment focused on subjective image quality (1 = poor, 2 =
sufficient for diagnosis, 3 = good, 4 = excellent), lesion detection, lesion
location based on an eight-segment partition
[14,
15], lesion size (class 1,
5 mm in the greatest dimension; class 2, from 6 mm to 1 cm; class 3, 1-2
cm; class 4, 2-5 cm; class 5, > 5 cm), the presence and pattern of lesion
enhancement, specific diagnosis, and confidence level for diagnosis (on a
5-point scale from 0 = not confident to 4 = definitely confident). Seven
patterns of lesion enhancement were considered for the dynamic scans: 1,
peripheral nodular pattern of hyperintensity with progressive partial
filling-in; 2, peripheral nodular pattern of hyperintensity with progressive
and complete filling-in; 3, peripheral hyperintensity with no subsequent
filling-in; 4, peripheral thin rim, thick rind, or reticulated pattern of
hyperintensity; 5, homogeneous hyperintensity with no preceding peripheral
enhancement; 6, diffuse heterogeneous hyperintensity; and 7, homogeneous
hypointensity relative to a hyperintense normal liver parenchyma. For the
delayed scans only four patterns of lesion enhancement were considered: 1,
central enhancement; 2, peripheral rim enhancement; 3, entirely diffuse
enhancement; and 4, entirely hypointense to the normal liver parenchyma.
For the characterization of the focal liver lesions detected in each interpreting session, the observers were asked first to decide whether a lesion was benign or malignant and then to choose one diagnosis from a list of possible lesion diagnoses. Specifically, the list of benign lesions comprised cyst, hemangioma, focal nodular hyperplasia, hepatocellular adenoma, regenerating nodule, and other; and the list of malignant lesions comprised HCC, cholangiocellular carcinoma, metastasis, and other. Each observer could also state, if necessary, that he was unable to give a specific diagnosis.
All reference procedures were assessed by unblinded local investigators. The segmental anatomy of the liver according to Couinaud [14] and Bismuth [15] was used to prepare liver maps for all observers (both blinded and unblinded) for localizing the lesions detected on MR imaging and reference procedures in order to best guide the matching between MR imaging and the reference procedure findings.
A lesion was considered benign when it fulfilled the diagnostic criteria for angioma (smooth or lobulated margins with homogeneous hypointensity on T1-weighted images, homogeneous hyperintensity on T2-weighted images, and peripheral nodular pattern of hyperintensity with progressive partial or complete filling-in during the contrast-enhanced study); cyst (smooth margins, homogeneous hypointensity on T1-weighted images, homogeneous hyperintensity on T2-weighted images, and no enhancement during the contrast-enhanced study); or focal nodular hyperplasia (smooth margins, isointensity or mild hyperintensity to the normal liver parenchyma on T2-weighted images, isointensity or mild hypointensity to the liver parenchyma on T1-weighted images, homogeneous hyperintensity with no preceding peripheral enhancement on dynamic images, evidence of a central scar that is hypointense in early acquisitions becoming hyperintense on delayed acquisitions).
Lesions that could not be classified as benign on the basis of these criteria and lesions that showed irregular margins, pseudocapsule, inhomogeneous enhancement, or portal invasion were judged to be malignant or indeterminate in nature.
The specific histologic diagnosis for each lesion was formulated on the basis of previously published diagnostic criteria [16,17,18,19].
Statistical Analysis
The sensitivity, specificity, and diagnostic accuracy of the MR image sets
were calculated separately for the nature of the lesion (i.e., malignant or
benign) and for the specific diagnosis of the focal lesion (i.e., metastasis,
HCC, angioma, focal nodular hyperplasia). The reference standard was the
histologic diagnosis for each lesion. Definitions were as described by Hamm et
al. [20] and are given as
footnotes in Tables 1 and
2.
|
|
The sensitivity of the MR image sets for the detection of focal liver lesions was also calculated. Reference standards for detection were taken from intraoperative sonography, iodized oil CT, CT during arterial portography, or contrast-enhanced CT findings and from follow-up CT or MR imaging in cases in which no reference standard was available. The sensitivity for lesion detection was defined as the number of true-positive lesions divided by the total of true-positive and false-negative lesions. Because no analysis was conducted on a per patient or per segment basis, there were no true-negative findings in the study; hence, the specificity for lesion detection could not be calculated.
Inferential statistics using the McNemar test with one degree of freedom [21] were used to determine the significance of changes in the primary efficacy variables (sensitivity, specificity, and diagnostic accuracy for lesion characterization, and sensitivity for lesion detection) from unenhanced MR images to dynamic scans, unenhanced plus dynamic scans, and all scans together. Interobserver agreement in assessing the accuracy of differentiating benign from malignant lesions and of making specific diagnoses on the basis of the MR imaging results was assessed on the basis of Cohen's kappa coefficient [22].
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The ability of the two observers to distinguish benign from malignant liver lesions using gadobenate dimeglumine-enhanced MR images is shown in Table 1. The accuracy in differentiating benign from malignant lesions increased from 74.8% and 82.2% on unenhanced images alone for observers 1 and 2, respectively, to 90.7% for both observers when a matched evaluation of the unenhanced and dynamic images was performed (p < 0.05). Evaluation of all the study images together (unenhanced, dynamic, and delayed acquisitions) added no further diagnostic information (89.6% and 90.6%) beyond that obtained during the matched assessment of unenhanced and dynamic phase images. The observer agreement in differentiating benign from malignant lesions rose from a kappa value of 0.472 for the evaluation of unenhanced images alone to a kappa value of 0.571 with the dynamic studies alone and to a kappa value of 0.717 when the unenhanced and dynamic image sets were assessed in the matched evaluation. Inclusion of the delayed images reduced the agreement to a kappa value of 0.450.
The results in terms of specific lesion diagnosis are shown in
Table 2. The accuracy achieved
by evaluating the unenhanced images only was limited (48.6% and 61.7% for
observers 1 and 2, respectively) but increased with the dynamic study to 73.8%
and 65.4%, respectively. The matched evaluation of the unenhanced and dynamic
images together allowed a further increase of accuracy (75.7% and 70.1%,
respectively). Although the increases in accuracy were statistically highly
significant over unenhanced values for observer 1 (p < 0.001),
only the matched pairs assessment approached statistical significance for
observer 2 (p < 0.06). The introduction of the delayed (60-120
min) images provided no additional information toward the correct specific
diagnosis of the detected lesions. The observer agreement for specific lesion
diagnosis rose from a kappa value of 0.516 for the assessment of unenhanced
images alone to a kappa value of 0.642 for the dynamic images alone. No
further increase was observed (
= 0.636) when the unenhanced and
dynamic images were evaluated in matched fashion, whereas inclusion of the
delayed images resulted in a slightly reduced agreement ( = 0.574). In
terms of the observer's confidence in making a diagnosis, a score of 3 or 4
was given to only 50% of lesions when unenhanced images alone were displayed.
When dynamic images alone or combined unenhanced and dynamic phase images were
displayed, a confidence score of 3 or 4 was given to 75% or lesions in both
cases. Finally, when delayed images were included, a confidence score of 3 or
4 was given to 85% of all lesions.
With respect to the enhancement patterns of the lesions during the dynamic phase of contrast enhancement, homogeneous or heterogeneous hypointensity to the liver with or without rim enhancement (pattern 3 = peripheral hyperintensity with no subsequent filling-in; pattern 4 = peripheral thin rim, thick rind, or reticulated pattern of hyperintensity; or pattern 7 = homogeneous or heterogeneous hypointensity relative to the hyperintense liver) was identified in 32 (91.4%) of the 35 correctly diagnosed metastases for observer 1 and in 34 (91.9%) of the 37 correctly diagnosed metastases for observer 2 (Fig. 1A,1B,1C,1D,1E).
|
|
|
|
|
For the HCCs in this lesion population, transient increased enhancement (pattern 5 = homogeneous hyperintensity with no preceding peripheral enhancement or pattern 6 = diffuse heterogeneous hyperintensity) was observed for 23 (74.2%) of the 31 correctly diagnosed HCC lesions for observer 1 and for 24 (85.7%) of the 28 correctly diagnosed HCC lesions for observer 2 (Fig. 2A,2B,2C,2D,2E). Both observers misclassified six hypovascular HCC lesions as metastases because of an atypical hypointensity during the arterial phase of the dynamic acquisitions. Five of these six patients had underlying hepatic cirrhosis.
|
|
|
|
|
The 11 cholangiocarcinomas were inhomogeneously hypointense to the liver in the arterial phase, but most of them (7/11) showed an inhomogeneous hyperintense central component during the equilibrium phase (2-4 min) of the dynamic acquisitions. This hyperintensity tended to be homogeneous in the smaller lesions but was limited to only the inner part of the lesion in the larger tumors. For the remaining four cholangiocarcinomas, the enhancement patterns were described as heterogeneously hypointense with more delayed heterogeneous hyperintensity.
Finally, the enhancement patterns during the dynamic studies of nine of the 10 hemangiomas were classified by both observers as peripheral nodular hyperintensity with progressive partial or complete filling-in (Fig. 3A,3B,3C,3D,3E). Only one hemangioma showed homogeneous hyperintensity with no preceding peripheral enhancement. This lesion was diagnosed on the basis of a well-delineated, rounded form and a homogenous hyperintensity during the arterial phase that persisted into each of the subsequent phases of the dynamic series.
|
|
|
|
|
The only focal nodular hyperplasia present in this lesion population showed a fast and diffuse hyperintensity with a central area of hypointensity that became hyperintense in the 5-min contrast-enhanced acquisition.
The enhancement patterns observed on the delayed (60-120 min) acquisitions were not characteristic of any kind of lesion. Metastases showed central enhancement (pattern 1) or hypointensity to the liver (pattern 4) in 93.8% (30/32) and 77.1% (27/35) of the lesions for observers 1 and 2, respectively. Higher variability was seen for HCC lesions. Observer 1 noted hypointensity to the liver for 16 (51.6%) of 31 correctly diagnosed HCC lesions and entirely diffuse enhancement (pattern 3) for eight (25.8%) of 31 nodules, whereas observer 2 described nine (33.3%) of 27 correctly diagnosed HCC lesions as having no enhancement and eight (29.6%) of 27 lesions as having central enhancement. The hemangiomas showed either hypointensity to the surrounding liver (30%), entirely diffuse enhancement (50%), or central enhancement (20%). Overall, the availability of the delayed acquisitions did not improve the accuracy of lesion characterization.
Lesion Detection
A total of 149 focal liver lesions were identified using the combination of
reference procedures and follow-up information. Six (4%) of the 149 lesions
were smaller than 5 mm, 18 lesions (12.1%) were from 6 mm to 1 cm, 40 lesions
(26.8%) were 1-2 cm, 52 lesions (35%) were 2-5 cm, and 33 lesions (22.1%) were
5-14 cm.
Of the 149 lesions, observer 1 identified 114 lesions on unenhanced images alone, 119 on dynamic images alone, 129 on unenhanced plus dynamic images, and 136 on unenhanced plus dynamic plus delayed images. Observer 2 detected 121 lesions on unenhanced images alone, 120 on dynamic images alone, 126 on unenhanced plus dynamic images, and 131 on unenhanced plus dynamic plus delayed images. Assessment of the unenhanced images alone resulted in sensitivities for lesion detection of 76.5% and 81.2% for observers 1 and 2, respectively. Although the sensitivities altered little when the contrast-enhanced dynamic phase images were assessed alone (79.9% and 80.5%, respectively), they increased to 86.6% and 84.6%, respectively, when the dynamic phase images were assessed in combination with the unenhanced image set, and to 91.9% and 88.5%, respectively, when all the image sets, including the 60- to 120-min delayed set, were assessed together. Specifically, the inclusion of the delayed phase images led to the additional identification of 5% and 3.5% more lesions, respectively, compared with those detected on the combined unenhanced plus dynamic phase images. Overall, the increase in sensitivity over unenhanced values alone was statistically significant (p < 0.01) for both observers when all images were considered together.
The lesions that were not detected by either observer consisted mainly of lesions smaller than 1 cm. Overall, 75% of the lesions undetected by either observer were smaller than 1 cm (50% between 5 and 10 mm and 25% between 1 and 5 mm), and the remaining 25% were between 10 and 15 mm. For those lesions with histologic confirmation (n = 107), observer 1 missed four HCC nodules and seven small metastases when unenhanced images were assessed alone. For the HCC nodules, the number missed decreased to two when unenhanced and dynamic phase images were assessed together and when delayed images were included in the assessment. For the metastases, the number missed decreased to four when the dynamic images were assessed in combination with the unenhanced images, and to just one when all images were displayed together. For observer 2, only one HCC nodule was missed in each of the assessments. For the metastases, however, six were missed when the unenhanced images were considered alone but only two when all scans were displayed together.
A separate analysis of only smaller lesions (64 lesions < 2 cm in greatest dimension, size classes 1-3) revealed increases in the sensitivity for lesion detection from 57.8% and 62.5% (observers 1 and 2, respectively) when unenhanced images were assessed alone to 71.9% and 75.0% when unenhanced images were assessed in combination with dynamic phase images and up to 84.1% and 79.4% when the unenhanced and dynamic phase images were assessed in combination with the delayed images. Statistically significant increases over unenhanced values were noted by both observers both for the combined assessment of unenhanced and dynamic images (p < 0.05) and, in particular, for the combined assessment of all scans including delayed images (p < 0.01). In this latter assessment, the inclusion of delayed images led to the additional identification of 11% and 3% more lesions (observers 1 and 2, respectively) than those detected on the combined unenhanced plus dynamic phase images.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Gadobenate dimeglumine is a novel gadolinium-based contrast agent for MR imaging. This agent combines the properties of a conventional nonspecific gadolinium-based agent with those of a liver-targeted agent [29]. Hence, gadobenate dimeglumine is effective not only in the dynamic phase of contrast enhancement after bolus administration [12], but also in a more delayed phase when uptake into functioning hepatocytes of 3-5% of the injected dose results in a marked and long-lasting enhancement of the signal intensity of normal liver parenchyma and a corresponding increased sensitivity of MR imaging for the detection of focal lesions [6, 11, 12, 29].
In a recent report, a 0.05 mmol/kg bolus dose of gadobenate dimeglumine was shown to yield additional information on lesion characterization for up to 43% of patients for whom assessment was made [12]. A criticism of this earlier study, however, was the lack of histologic correlation for the MR imaging findings. In our study, histologic proof of pathologic findings was available for 107 of the 149 lesions detected by reference standards. With histologic proof as reference, the two observers in our study obtained an accuracy of 90.7% for the differentiation of benign from malignant lesions when assessing unenhanced and dynamic phase image sets together in the absence of any further clinical data. In a comparable study with the nonspecific agent gadopentetate dimeglumine (Magnevist; Schering, Berlin, Germany), an accuracy of 79.2% was reported for the combined assessment of unenhanced and dynamic phase image sets, which increased to 87.4% with the inclusion of additional clinical data into the assessment [20]. Significantly, the dose used in our study for the dynamic phase acquisitions was half that used in the study on gadopentetate dimeglumine. Possibly the effectiveness of a dose of only 0.05 mmol/kg of gadobenate dimeglumine for accurate liver lesion characterization during the dynamic phase of contrast enhancement is related to the fact that this agent shows elevated T1-relaxivity in blood because of a weak affinity for serum albumin [5].
In terms of the accuracy for specific lesion diagnosis, the values for the combined assessment of unenhanced and dynamic phase image sets (75.7% and 70.1% for observers 1 and 2, respectively) are similarly comparable to values observed elsewhere for nonspecific gadolinium chelates at twice the dose [20, 30]. Interestingly, however, the improvement in specific diagnosis on contrast-enhanced images was much greater for observer 1 than for observer 2. This difference may have been a reflection of the experience of the two observers in using gadolinium-based contrast agents in routine clinical practice. In this regard, observer 1 routinely uses gadolinium contrast agents and dynamic phase imaging for all liver examinations, whereas observer 2 uses contrast enhancement only for doubtful cases on unenhanced imaging. This may explain why observer 1 was more accurate when assessing contrast-enhanced images than unenhanced images, whereas observer 2 was comparatively more accurate in assessing unenhanced images.
As for image evaluation, neither of the two observers had any experience in interpreting gadobenate dimeglumine—enhanced images before beginning the assessment. Nevertheless, each observer recognized enhancement patterns typical of hemangioma, HCC, cholangiocarcinoma, and metastases in a high percentage of cases, indicating that the patterns observed were similar to those known to occur with other gadolinium chelates. Also, Cohen's kappa coefficient—a measure of interobserver agreement in making a given diagnosis—increased for the differentiation of benign from malignant lesions from a kappa value of 0.472 for the evaluation of unenhanced images alone to a kappa value of 0.717 for the matched pairs evaluation of unenhanced and dynamic phase images. Because evaluation of contrast-enhanced dynamic phase images alone resulted in a kappa value of just 0.571, these findings emphasize the need to perform combined evaluations of all unenhanced and enhanced images together if more correct diagnoses are to be reached with higher reproducibility.
Of particular interest were the enhancement patterns observed for seven of the 11 cholangiocarcinomas in the study. Unlike the patterns observed after administration of nonspecific gadolinium chelates in which a relatively delayed (10 min) hyperintensity is noted [31], in the present study a much more precocious central hyperintensity was observed by both observers, with enhancement already evident in the last two acquisitions of the dynamic series (3-8 min). Such observations may be of great interest given the similarity of the histologic features of peripheral cholangiocarcinoma and metastatic adenocarcinoma and the difficulty of differentiating between the two lesion types on MR imaging [32]. In our study, 60% of metastases failed to show any enhancement before the final acquisitions of the dynamic sequence. Although it is possible the earlier hyperintensity of cholangiocarcinoma with gadobenate dimeglumine is the result of the superior T1-relaxivity of this molecule compared with other gadolinium agents during the distribution phase in the interstitial spaces of the tumor [5, 29], this is by no means certain, and further work will be necessary to clarify the situation. Overall, no obvious differences (e.g., no desmoplasia or fibrosis) were seen in terms of histology between the 11 cholangiocarcinomas, and no differences on the delayed images that in all cases showed either peripheral rim enhancement or heterogeneous hypo- or hyperintensity.
Interestingly, the delayed (60-120 min) images in the present study provided little or no additional information toward the accurate characterization of lesions beyond that available on the unenhanced and dynamic phase images; indeed, few changes in specific diagnosis were made (3/107 and 2/107 for observers 1 and 2, respectively) after inclusion of the delayed images. However, separate assessment of delayed images was not performed, and it is not known to what extent emphasis was placed on these images when they were included with the unenhanced and dynamic phase images in the final matched assessment. Given each observer's inexperience in evaluating gadobenate dimeglumine—enhanced images, it is possible that only minimum consideration was given to the delayed images when assessing the accuracy for lesion characterization. However, it is equally possible that the delayed images helped improve the observer's confidence in making certain diagnoses. In this regard, a diagnostic confidence score of 3 or 4 was given to 10% more lesions after addition of the delayed images to the unenhanced and dynamic image sets. That delayed gadobenate dimeglumine—enhanced images provide additional information for the characterization of HCC has been shown in two recent studies [33, 34].
As we have mentioned, a principal limitation of our study was that separate assessment of delayed images was not performed. Nevertheless, overall sensitivities of 86.6% and 84.6% (observers 1 and 2, respectively) were observed for the combined assessment of all unenhanced and dynamic phase images, and increased sensitivities of 91.9% and 88.5% were observed when the delayed images were added to the assessment. Although such values for the present mixed-lesion population compare well with previously reported sensitivities of up to 72% and 89% for delayed gadobenate dimeglumine—enhanced MR imaging of primary hepatic malignancies and metastases, respectively [11], the inclusion of delayed images in our study permitted the additional detection of only 5% and 3.5% (observers 1 and 2, respectively) more lesions than those detected in the matched assessment of unenhanced and dynamic phase images. In terms of lesions smaller than 2 cm, only 11% and 3% (observers 1 and 2, respectively) more were detected after inclusion of the delayed images. These results may be explained by the fact that the assessment methodology was designed primarily to determine the value of gadobenate dimeglumine for accurate liver lesion characterization rather than liver lesion detection. Hence, blinded image evaluation and subsequent comparison of MR images with gold standard findings was performed only at the level of hepatic lesions previously identified on unenhanced MR imaging or contrast-enhanced CT. A possible result, therefore, was that a number of lesions were missed through restricting the search for additional lesions to a limited number of slices. If the analysis of lesion detection had focused on the entire liver, as is possible in 2000 given faster imaging sequences during breath-holding, it is possible that improved results would have been obtained.
Finally, relatively thick slices (10 mm) were used for all image acquisitions in our study, and most lesions that remained undetected by both observers in all assessments were smaller than 1 cm. Given that delayed MR imaging with gadobenate dimeglumine has previously been shown to be particularly advantageous for the detection of subcentimeter lesions [11], it is possible that the sensitivity for lesion detection would have been improved with the use of thinner slices.
In conclusion, our study confirms that gadobenate dimeglumine behaves as a nonspecific gadolinium chelate in the first minutes after administration and as a liver-targeted agent in a later delayed phase. On the basis of the results obtained, a gadobenate dimeglumine dose of 0.05 mmol/kg, administered as a single IV bolus and followed by MR imaging during the dynamic phase of contrast enhancement, can be recommended for both the accurate differentiation of benign from malignant liver lesions and for specific lesion diagnosis. In patients for whom additional lesion detection is warranted—for example, patients with primary carcinoma of the lung, breast, colon, or pancreas who may be suspected of harboring hepatic metastases—additional MR images can be acquired at 60-120 min after injection to assess both tumor burden and location. In this regard, delayed MR imaging with gadobenate dimeglumine has also been shown to be effective for both the detection [11] and the further characterization of HCC [33, 34].
Finally, as to the dose to be administered, although the dose was split in our study, this need not be the case in routine clinical practice. Although preliminary studies showed a greater enhancement of liver parenchyma signal intensity on delayed MR images with a 0.1 mmol/kg dose than with a 0.05 mmol/kg dose [10], no statistically significant differences between these doses in terms of lesion detection have been noted [35].
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  S. H. Choi, J. M. Lee, N. C. Yu, K.-S. Suh, J.-J. Jang, S. H. Kim, and B. I. Choi Hepatocellular Carcinoma in Liver Transplantation Candidates: Detection with Gadobenate Dimeglumine-Enhanced MRI Am. J. Roentgenol., August 1, 2008; 191(2): 529 - 536. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. Leen, P. Ceccotti, C. Kalogeropoulou, W. J. Angerson, S. J. Moug, and P. G. Horgan Prospective multicenter trial evaluating a novel method of characterizing focal liver lesions using contrast-enhanced sonography. Am. J. Roentgenol., June 1, 2006; 186(6): 1551 - 1559. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
