| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 All authors: Department of Radiology, Boston University School of Medicine, Boston Medical Center, 88 E. Newton St., Atrium 2, Boston, MA 02118.
Received July 27, 2000;
accepted after revision October 4, 2000.
Portions presented at the meeting of the International Society for Magnetic
Resonance in Medicine, Sydney, Australia, April 1998, and at the annual
meeting of the American Roentgen Ray Society, New Orleans, May 1999.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. A cross-sectional study of 227 abdominal MR imaging examinations revealed 85 lesions in 67 patients (29 men, 38 women; age range, 29-78 years; mean age, 51.4 years) who were being examined for primary malignancy (n = 42) or unknown lesion characterization (n = 25). All were referred for MR imaging after CT or sonography. Patient demographics (age, sex, history of malignancy), lesion size and morphology, quantitative T2 calculation, and pattern of enhancement on gadopentetate dimeglumine administration were evaluated for predictive ability.
RESULTS. Thirty-two liver lesions were malignant (eight colon cancer, five breast cancer, four cervical cancer, three renal cancer, three lung cancer, and nine miscellaneous cancers), 53 were benign (37 hemangiomas, 15 cysts, and one focal nodular hyperplasia). Calculated T2 relaxation times (mean ± standard deviation [SD]) were as follows: malignant tumors (91.72 ± 21.9 msec), hemangiomas (136.1 ± 26.3 msec), cysts (284.1 ± 38.2 msec) (p < 0.001). Logistic regression analysis indicated that lesion size and sex and age of patient were not significant independent predictors (p > 0.05). However, the combination of a history of malignancy, T2 value, and gadopentetate dimeglumine—enhancement pattern allowed generation of a prediction rule with an area under the receiver operating characteristic curve of 0.95. The patient's weight, lesion morphology, and cell type of the primary malignancy did not provide additional predictive information (p > 0.2).
CONCLUSION. We recommend using the combination of T2 quantification and patient history of malignancy before deciding to administer gadopentetate dimeglumine for optimal lesion characterization, especially for equivocal lesions with T2 values between 90 and 130 msec. These factors allowed the construction of a prediction rule for lesion characterization.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Prediction rules are gaining a greater hold in evidence-based medical evaluation of problems such as ankle and knee trauma [3, 4] to decide which patients merit radiographic examination. Hence, the goals of this study were to establish which factors (e.g., lesion morphology, contrast enhancement, and patient demographics) have a high predictive ability to differentiate solid, likely malignant lesions from benign disease, and to construct a prediction rule for hepatic lesions revealed on MR imaging. Such a prediction rule could ultimately be used to stratify lesions into biopsy, watchful waiting, or benign categories with high reliability and reproducibility.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
A cohort of 67 patients with a total of 85 focal hepatic lesions was identified on the basis of these criteria and constituted the cross-sectional study group examined. The patients were 29 men and 38 women (age range, 29-78 years; mean age, 51.4 years). Forty-two patients had a known diagnosis of a primary malignancy. Of the 85 hepatic lesions, 32 were confirmed as malignant and 53 as benign. The thirty-two malignant lesions consisted of two primary hepatocellular carcinomas and 30 metastases (known primary malignancy: colorectal carcinoma [n = 8], breast carcinoma [n = 5], cervical carcinoma [n = 4], renal carcinoma [n = 3], bronchogenic carcinoma [n = 3], pancreatic carcinoma [n = 3], ovarian carcinoma [n = 2], malignant melanoma [n = 1], and small-bowel adenocarcinoma [n = 1]). The 53 benign lesions consisted of 37 hemangiomas, 15 cysts, and one focal nodular hyperplasia. The diameter of each hepatic lesion was measured from the hard copy of the moderately T2-weighted MR image by a single radiologist using handheld calipers. Mean lesion size was 18.2 mm (range, 8-66 mm).
Verification of Lesion Type
Of the 32 malignant hepatic lesions, direct histopathologic proof was
obtained for 18 lesions at imaging-guided fine-needle aspiration or core
biopsy (n = 14), intraoperative biopsy (n = 2), or autopsy
(n = 2). For nine lesions, histopathologic proof of malignancy in
another similar-appearing lesion was obtained and an interval increase in
lesion size was shown on follow-up imaging (follow-up range, 9-16 months; mean
follow-up, 11 months). Five lesions identified in patients with known primary
malignancies were reported as metastases on the basis of an interval increase
in size or number on serial cross-sectional images obtained over a period of 1
year or less in the same manner as that described by McFarland et al.
[1].
Of the 53 benign lesions, the diagnosis of hemangiomas was confirmed in 37 in one of three ways: two were biopsied and confirmed histopathologically; two showed characteristic uptake on 99mTc-RBC scintigraphy [5]; and 33, with noncystic appearances on sonography or contrast-enhanced CT, had features typical of hemangioma and showed no change in size or number of lesions at serial cross-sectional imaging at 1 year or more of follow-up (follow-up range, 12-84 months; mean follow-up, 18 months) [6]. In one patient, one area of focal nodular hyperplasia was confirmed histopathologically. Fifteen hepatic cysts were confirmed on the basis of a typically cystic appearance on sonography and unenhanced or contrast-enhanced CT (attenuation value on unenhanced CT, <15 H; increase in attenuation after contrast medium injection, <10 H) without evidence of an increase in size and number on serial cross-sectional imaging over a period of 1 year or more (follow-up range, 12-21 months; mean follow-up, 14 months) [1]. In essence, a presumptive diagnosis of benign cause was based on follow-up imaging showing no growth during more than 12 months in lesions for which diagnostic confirmation could not be made with other means.
MR Imaging
All MR images were acquired on a superconductive 1.5-T MR unit (Gyroscan
ACS NT; Philips Medical Systems, Best, The Netherlands) using conventional
spin-echo sequences. To obtain both moderately and heavily T2-weighted images,
the following standard scanning parameters were used: TR/first-echo
TE/second-echo TE, 3600/50/160; excitations, 2. Matrix size was 256 x
128, with a field of view of 32 x 37 cm, yielding an effective in-plane
resolution of 1.25-1.44 x 2.5-9 mm. A body coil was used for
transmission and reception. Section thickness was 8 mm with a 1-mm gap.
Respiratory and flow compensation were routinely used. Average imaging time
for double-echo T2-weighted sequences was 15 min 30 sec.
In the patients who received gadopentetate dimeglumine (Magnevist, Berlex, Wayne, NJ), the dose was 0.1 mmol/kg (to a maximum of 20 mL). Axial dynamic breath-hold MR imaging was performed centered over the volume of interest using eight to 11, 8-mm-thick sections with a 2-mm gap. These images were acquired during breath-holding with a two-dimensional fast spoiled gradient-echo sequence (TR range/TE range (minimum), 68-150/3.9-4.6; flip angle, 75°; excitation, 1) during the bolus administration of gadopentetate dimeglumine. Gadopentetate dimeglumine was administered over 20 sec by hand, with the first scan initiated at the start of the bolus administration. Acquisition time was 20-40 sec for the complete set, with a predefined break of 10-15 sec between each acquisition. This sequence was repeated six times sequentially. In-plane resolution was 2.5 x 1.2 mm (matrix size, 128 x 256; field of view, 300 x 300 mm). No saturation pulses were applied.
Image Analysis
The signal intensity of each lesion was measured by one radiologist who was
unaware of the patient's diagnosis. The largest oval or circular region of
interest located in the confines of the lesion was selected, except in the
case of heterogeneous lesions. In the latter situation, central areas of
T2-weighted hyperintensity consistent with necrosis were excluded from the
region of interest so that signal-intensity measurements represented the
nonnecrotic portions of each lesion
[1]. Peritumoral edema, when
present, was not included in the region of interest. Signal intensity
measurements were recorded for each lesion on both moderately and heavily
T2-weighted images using identically sized regions of interest on the same
transverse section. Three signal-intensity measurements were recorded for each
lesion at the 50- and 160-msec TEs, and mean values were calculated.
The mean signal intensity measurements of each lesion on the first- and second-echo TEs of the T2-weighted sequences were entered into a computer spreadsheet (Excel, version 4.0; Microsoft, Bothel, WA). T2 relaxation times were calculated using the algorithm based on a standard T2 index as described by Mirowitz et al. [7], whereby the natural logarithm of signal intensity on a spin-echo image is linearly related to TE with a slope of -1/T2. For each lesion, signal-intensity measurements and corresponding T2 relaxation time calculations were repeated three times to determine the reproducibility of the quantitative results.
Qualitative Image Analysis
Sixty of the 67 patients received 0.1 mmol/kg of gadopentetate dimeglumine.
The enhancement pattern was rated as hemangiomalike or not, using the criteria
of Quillin et al. [8], which
are based on the work of Hamm et al.
[9] (early peripheral nodular
enhancement with puddling is defined as hemangiomalike), by independent
experienced radiologists unaware of the final diagnosis. The enhancement
pattern was scored using an ordinal scale. The morphology of all 85 lesions
was reviewed independently by one observer and classified as simple
(homogeneous on T2-weighted MR images, no peritumoral edema, sharp borders) or
complex using an ordinal scale. The specific lesion to be analyzed was
indicated for each patient after removal of all identifying information.
Statistical Analysis
True-positive cases were defined as malignant lesions that were correctly
assigned. True-negative cases were defined as benign lesions that were
correctly assigned. Factors analyzed included the age, sex, and weight of the
patient; history of known malignancy; cell type of known malignancy
(adenocarcinoma, squamous cell, or other); lesion size; morphology; and
gadopentetate dimeglumine-enhancement pattern classification.
All data collected were placed into a spreadsheet (Excel), and logistic regression analysis using STATA, version 5.0 (STATA, College Park, TX) was performed. Univariate analysis was used to determine the significance of each factor in predicting malignancy by odds ratio evaluation. The most significant variables at the cutoff of p less than 0.05 were collected, and collinearity, confounding, and effect modifications were assessed by forward selection, backward selection, and selective elimination [10]. Odds ratios were then calculated for the resulting best-fit model with a receiver operating characteristic curve, and the Hosmer-Lemeshow goodness-of-fit statistic was calculated [10]. Using variables significant at a p value of less than 0.05, a series of prediction rules was constructed on the basis of the logistic regression model technique [10].
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
No confounding or effect modifications were seen with these factors. Collinearity between morphology, calculated T2, and size of the lesion indicated that the primary independent variables for predicting malignancy of a given liver lesion were, with adjusted odds ratios, 0.94 (95% CI, 0.90-0.97) for calculated T2 (i.e., for each increase in T2 by 1 msec the odds of malignancy are decreased by 0.94) 0.042 (95% CI, 0.008-0.10) for hemangiomalike enhancement, and 4.7 (95% CI, 1-24) for known history of primary malignancy. Receiver operating characteristic curve analysis [11] showed that use of these factors allows accurate differentiation of benign from malignant lesions with an area under the curve (Az) of 0.95 (95% CI, 0.90-0.99) (Fig. 1).
|
Model Construction
Using the variables significant at a p value of less than 0.05, a
series of prediction rules was constructed on the basis of the logistic
regression model technique to establish the probability of a lesion being
malignant [10] using the data
from this study. For example, if only patient history of a known malignancy is
used, the odds of a given lesion being malignant are increased sevenfold.
Hence, if the prevalence of a malignant lesion before MR imaging is 20%, then
the a priori odds of a malignant lesion are 1:4; if there is a known history
of malignancy, these odds are increased to 7:4, or a probability of being
malignant of 64%. Thus, using the lesion's calculated T2 time, a probability
distribution for the probability of a hepatic lesion being malignant can be
constructed (Fig. 2)
(Az = 0.93 [95% CI, 0.88-0.98]). This distribution
function can be modified by knowledge of a known malignancy
(Fig. 3), by the character of
enhancement on gadolinium dimeglumine administration
(Fig. 4), or both
(Fig. 5)
(Az = 0.95 [95% CI, 0.90-0.99]). This model is based on a
baseline probability of malignancy of 0.29 before MR imaging for a patient
without a known malignancy, a solid lesion, a T2 of 90 msec, and no
hemangiomalike enhancement. The probability function is given by:
|
|
|
|
Probability of malignancy = 1 / [1 + Exp (-6.6769 + 0.0643 x T2 (in msec) + 3.16 x gadopentetate dimeglumine enhancement (0 = not hemangioma, 1 = hemangiomalike)-1.9458 x known (0 = no malignancy, 1 = prior malignancy)].
Therefore, for a lesion with a T2 of 110 msec and hemangiomalike enhancement in a patient without a known history of malignancy, the probability that the lesion is malignant is 0.028.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Visual analyses of MR signal characteristics are inherently subjective and are prone to uncertainty and inconsistency [2]. Lesion characterization using calculated T2 relaxation times derived from double-echo T2-weighted sequences has been shown to be more accurate than other quantitative methods (lesion-to-liver contrast-to-noise, lesion-liver, lesion-fat, and lesion-muscle signal intensity ratios) at 1.5-T field strengths [1, 12, 22]. Furthermore, such measurements have been shown to be superior to subjective evaluations of lesion brightness and morphology [1, 2, 23].
In our study, double-echo T2-weighted MR images were acquired using conventional spin-echo sequences alone. A recent study by Ito et al. [18] reported a 100% accuracy in differentiating small benign lesions (<3 cm) from small malignancies using qualitative analysis of double-echo T2-weighted images acquired with fast spin-echo techniques. The high accuracy reported by Ito et al. most likely relates to greater differences in the signal intensities of solid and nonsolid lesions on fast spin-echo images because of increased magnetization transfer contrast and, possibly, diffusion effects [24]. More recently, Bosmans et al. [25] successfully used double-echo HASTE techniques for qualitative hepatic lesion characterization. Although observer performance may be improved using fast spin-echo and HASTE sequences rather than conventional spin-echo sequences, to our knowledge no data exist regarding the performance of T2 calculations under these circumstances.
Until the accuracy of T2 calculations obtained using fast spin-echo sequences has been evaluated, double-echo conventional spin-echo T2-weighted sequences are required to obtain accurate measurements of T2 relaxation times, particularly because the ability to measure T2 relaxation times using conventional spin-echo sequences is independent of the imaging system used [1, 16, 19, 22, 26]. Should quantitative measurements of T2 relaxation times using fast spin-echo or HASTE sequences prove reliable, they would be preferable to conventional spin-echo sequences in terms of scanning time, motion artifact reduction, patient tolerance, and throughput [18, 21, 27]. An expected limitation of using fast spin-echo sequences is that the calculations may vary with different echo spacing and echo-train lengths. Because fast spin-echo sequences are implemented differently by different manufacturers, unlike with conventional spin-echo sequences, T2 calculations may vary with the imaging system used. Ultimately, this prediction model would require adjustment for any other T2 calculation method.
Some limitations are associated with the use of T2 measurements for lesion characterization. Variability may occur due to the manner in which quantitative measurements are made [28,29,30]. For larger lesions, heterogeneity of signal intensity in the lesion is averaged if the largest region of interest is selected. By excluding central areas of hyperintensity corresponding to necrosis, we attempted to measure only the signal intensities of the nonnecrotic portions of each tumor. Although this method is subjective, T2 calculations performed with this method have been shown to be highly reliable [2].
When attempting to differentiate cysts or hemangiomas from malignant hepatic lesions on MR imaging, observers combine a visual assessment of many variables, including lesion brightness, lesion morphology, and patterns of contrast medium enhancement. Quantitative measures of T2 provide objective evidence of lesion type and help reduce the uncertainty and inconsistency associated with visual inspection. Interestingly, 22 of 32 metastases and all benign lesions in our series had a homogenous appearance without necrosis or peritumoral edema on T2-weighted images. In these patients, evaluation of lesion morphology would not have contributed to characterization. Indeed, assessment of lesion morphology on double-echo T2-weighted images has not been shown to significantly improve diagnostic accuracy compared with evaluation of lesion brightness alone [23]. The collinearity of these variables is confirmed in our regression analysis that showed that size, morphology, and T2 calculation have similar trends, with T2 being the most reliable and strongest predictor of the three variables. The additional information gained by the patient's history and the enhancement character on gadopentetate dimeglumine administration can be used for equivocal images with T2 values between 90 and 130 msec to more accurately classify given hepatic lesions. Although there may be a selection bias because lesions seen on MR imaging are difficult to characterize on other imaging modalities, this should have no effect on variables that have strong predictive values. In addition, the paucity of primary hepatic tumors in our population may limit the application of this information in this setting.
Because prediction rules are gaining a greater hold in evidence-based evaluation of problems such as ankle and knee trauma to determine which patients merit radiographic examination [3], the extension into decision making by the radiologist is a natural next step. The ability to apply logistic regression analysis with odds ratio calculation allows the construction of prediction rules that permit one to estimate the probability of a lesion's being malignant. In particular, such a set of prediction rules could ultimately be used to stratify lesions into biopsy, watchful waiting, or benign categories with high reliability and reproducibility. Thus, instead of ambiguous terms such as "probably malignant," explicit probability estimates for a given lesion can be calculated and used in any future decision analysis situation.
In this era of cost containment, it is possible to construct decision trees that can establish which probability cutoff to use before proceeding with interventions [31]. However, application of these analyses to individual patients requires explicitly determining the probability of a diagnosis in any given patient. It is in this void that we attempt to use factor analysis to calculate a probability function for hepatic lesions characterized on MR imaging. Determination of which probability cutoff will prove to be optimal for characterization of a lesion as benign (cyst or hemangioma), watchful waiting, or malignant requires cost-efficacy analysis beyond the scope of this study. An optimal analysis would take into account that for watchful-waiting lesions, the follow-up interval should be inversely proportional to the probability of malignancy.
In conclusion, we recommend the use of T2 quantification and patient history to identify patients with equivocal lesions in whom gadolinium enhancement with T2 relaxation values between 90 and 130 msec will provide additional information for optimal lesion characterization. The use of these factors in our prediction rule for lesion characterization may help improve patient outcomes, and we hope that probability estimates will become routine when reporting hepatic lesion characteristics on MR imaging.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  S. Bixby, R. Tello, and E. Kuligowska Presence of a Yolk Sac on Transvaginal Sonography Is the Most Reliable Predictor of Single-Dose Methotrexate Treatment Failure in Ectopic Pregnancy J. Ultrasound Med., May 1, 2005; 24(5): 591 - 598. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B Erol, U. Yilmaz, R Tanrikulu, and A Yilmaz Determinants of MR relaxation rates in jaw cysts: implications for diagnostic values of the relaxation times Dentomaxillofac. Radiol., May 1, 2004; 33(3): 183 - 187. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. C. Carlos, H. M. Kim, H. K. Hussain, I. R. Francis, H. V. Nghiem, and A. M. Fendrick Developing a Prediction Rule to Assess Hepatic Malignancy in Patients with Cirrhosis Am. J. Roentgenol., April 1, 2003; 180(4): 893 - 900. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. C. Mueller, H. K. Hussain, R. C. Carlos, H. V. Nghiem, and I. R. Francis Effectiveness of MR Imaging in Characterizing Small Hepatic Lesions: Routine Versus Expert Interpretation Am. J. Roentgenol., March 1, 2003; 180(3): 673 - 680. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 B. Taouli, V. Vilgrain, E. Dumont, J.-L. Daire, B. Fan, and Y. Menu Evaluation of Liver Diffusion Isotropy and Characterization of Focal Hepatic Lesions with Two Single-Shot Echo-planar MR Imaging Sequences: Prospective Study in 66 Patients Radiology, January 1, 2003; 226(1): 71 - 78. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
