Abstract

OBJECTIVE. The purpose of this study was to evaluate the potential value of MR elastography (MRE) in the characterization of solid liver tumors.
MATERIALS AND METHODS. Forty-four liver tumors (14 metastatic lesions, 12 hepatocellular carcinomas, nine hemangiomas, five cholangiocarcinomas, three cases of focal nodular hyperplasia, and one hepatic adenoma) were evaluated with MRE. MRE was performed with a 1.5-T system with a modified phase-contrast gradient-echo sequence to collect axial wave images sensitized along the through-plane motion direction. The tumors were identified on T2- and T1-weighted and gadolinium-enhanced T1-weighted images, and the MRE images were obtained through the tumor. A stiffness map (elastogram) was generated in an automated process consisting of an inversion algorithm. The mean shear stiffness of the tumor was calculated with a manually specified region of interest over the tumor in the stiffness map. The stiffness value of tumor-free hepatic parenchyma was calculated. Statistical analysis was performed on the stiffness values for differentiation of normal liver, fibrotic liver, benign tumors, and malignant tumors.
RESULTS. Malignant liver tumors had significantly greater mean shear stiffness than benign tumors (10.1 kPa vs 2.7 kPa, p < 0.001), fibrotic liver (10.1 kPa vs 5.9 kPa, p < 0.001), and normal liver (10.1 kPa vs 2.3 kPa, p < 0.001). Fibrotic livers had stiffness values overlapping both the benign and the malignant tumors. A cutoff value of 5 kPa accurately differentiated malignant tumors from benign tumors and normal liver parenchyma in this preliminary investigation.
CONCLUSION. MR elastography is a promising noninvasive technique for assessing solid liver tumors. Use of MRE may lead to new quantitative tissue characterization parameters for differentiating benign and malignant liver tumors.

Introduction

Tumors frequently are detected through physical palpation as hard masses located within softer surrounding tissue [1]. Palpation is used to assess the tendency of tissue to resist deformation, a physical property of tissue known as elasticity, which varies over a much wider range than other physical properties, such as x-ray absorption and MR relaxation time [2, 3]. It is generally agreed that no other physical parameter of tissue changes with physiologic and pathologic processes to as great an extent as does elasticity [4]. Researchers using sonography [5] and MRI [6] have developed imaging techniques for noninvasive assessment of the mechanical properties of tissues. These techniques, known collectively as elastography, are used to measure internal displacement or strain in tissue that results from application of a static, quasi static, or dynamic stress on that tissue.
Sonography-based elastography, in which static stress is used as a probe, has been found useful in the differential diagnosis of breast, thyroid, and prostate cancers [7-9] and in detection of cervical lymph node metastasis [10]. However, the elastic modulus of tissue cannot be quantitatively measured with the technique.
MR elastography (MRE) is a phase contrast-based MRI technique for direct visualization and quantitative measurement of propagating mechanical shear waves in biologic tissue [6, 11, 12]. The technique is used to obtain spatial maps and measurements of shear wave displacement patterns. The wave images are processed to generate maps known as elastograms, which show local quantitative values of the shear modulus of tissues [4].
MRE has been found useful for noninvasive assessment of hepatic fibrosis [13, 14]. Studies have shown that MRE can be used to differentiate normal liver from fibrotic liver with a high degree of accuracy and that the technique is promising for evaluating the stage of hepatic fibrosis [15]. In other applications, MRE has been found to have promise for differentiating benign breast lesions from malignant tumors [16, 17]. Aside from one publication describing MRE of brain tumors [18], there have been few reports of the use of MRE to evaluate masses in other parts of the body.
Inspired by the successful application of MRE to noninvasive evaluation of fibrosis of the liver in our clinical practice, we sought to evaluate the utility of MRE in characterizing liver tumors. Noninvasive characterization of focal liver lesions is typically based on the appearance of lesions on unenhanced T1- and T2-weighted images and on the dynamic enhancement patterns of lesions after bolus injection of a gadolinium-based contrast agent. MRE can be performed after contrastenhanced MRI, so immediate elastographic analysis of enhanced lesions detected on MRI is feasible. The main aim of our study was to determine the feasibility of MRE of liver tumors in vivo and to characterize the tumors. Our hypothesis was that MRE can depict the elastic properties of solid liver tumors and that malignant liver tumors have high shear stiffness compared with normal liver tissue.

Materials and Methods

This HIPAA-compliant study was approved by the institutional review board of our institution, which waived the requirement for informed consent for retrospective data analysis. The MRE sequence was performed as a clinical application as part of routine conventional MRI of the liver. Patients gave informed consent for all MRI studies.

Patients

Between January 2007 and July 2007, 29 patients (16 men, 13 women; mean age, 56.8 years; range, 18-78 years) with solid liver masses detected with contrast-enhanced MRI were recruited for the MRE study. The indications for MRI were follow-up of chronic liver disease, evaluation of lesions suspected on sonography or CT, and preoperative staging or follow-up of a known primary malignant tumor elsewhere than in the liver.
In patients with multiple tumors, we excluded tumors smaller than 10 mm to avoid partial volume effects with surrounding liver parenchyma and targeted the largest tumors wherever possible. Eight metastatic lesions in two patients were excluded from the study because the MRE images did not include the metastatic lesions. Two hepato cellular carcinomas in two patients previously treated with chemoembolization also were excluded from the study. The final study group of 44 liver tumors comprised 12 hepato cellular carcinomas (HCCs), 14 metastatic lesions, nine hemangiomas, five cholangio carcinomas, three cases of focal nodular hyperplasia, and one hepatic adenoma. The size of the liver masses ranged from 1.4 to 11 cm in maximal diameter (mean diameter, 3.8 cm).
The final diagnoses of primary liver tumors were established with biopsy or surgery in four cases of HCC, five of cholangiocarcinoma, one of focal nodular hyperplasia, and one of hepatic adenoma. The diagnoses of eight HCCs were based on the widely accepted imaging criteria of the European Association for the Study of the Liver and the American Association for the Study of Liver Diseases [19, 20]. These criteria are presence of a nodule larger than 2 cm in diameter and showing characteristic arterial phase hypervascularity on two imaging madalities, including CT or MRI. When, however, the tumor nodule demonstrates arterial phase hypervascularity and venous phase washout, only a single imaging technique is required for diagnosis. The diagnoses were made by experienced attending radiologists and gastroenterologists. Hemangiomas were diagnosed on the basis of their typical appearance on unenhanced and contrast-enhanced T1- and T2-weighted MRI and supportive CT or sonography and the presence of lesion stability on serial images and with clinical observation (mean period, 22.6 months; range, 6-38 months).
For all five patients with metastasis, the diagnosis was based on the surgical and histopathologic findings. The primary lesions were colonic cancer in two patients and intestinal neuroendocrine carcinoma in two patients. The primary site was unknown in one patient with type 1 neurofibromatosis and multiple neuroendocrine carcinoma metastases to the liver. Histologic proof was available for seven metastatic nodules (five neuroendocrine carcinomas, two colonic carcinomas). Two colonic carcinoma metastatic nodules were confirmed with uptake on FDG PET, surgical palpation, and intraoperative sonography. The five cases of metastasis of neuroendocrine carcinoma were confirmed with avid somatostatin uptake on octreotide scans.
Sixteen patients had a history of diffuse disease of the liver parenchyma or cirrhosis of the liver. The causes of hepatic fibrosis were hepatitis C in three cases, hepatitis B in three cases, primary sclerosing cholangitis in one case, nonalcoholic steatohepatitis in two cases, cryptogenic in four cases, alcohol abuse in two cases, and primary biliary cirrhosis in one case. Histologic confirmation of fibrosis or cirrhosis was available for seven patients (Metavir score 1 in one case, 2 in one case, 3 in two cases, 4 in three cases). In the other patients with diffuse liver disease, the diagnosis of hepatic fibrosis was based on presence of a risk factor, visualization of nodularity, and/or cirrhotic morphology on imaging (nine patients) and signs of portal hypertension, including splenomegaly (seven patients), portosystemic collateral vessels (two patients), increased serum level of liver enzymes (eight patients), aspartate amino transferase-to-platelet ratio index greater than 1 (six patients), and presence of esophageal varices on endoscopy (one patient).
Thirteen patients had no history of chronic liver disease and no known risk factors, and the liver parenchyma did not exhibit nodularity or a cirrhotic configuration on MRI. These patients also had normal serum levels of liver enzymes and an aspartate aminotransferase-to-platelet ratio index less than 1. The liver parenchyma in these patients was considered normal. Diffuse mild to moderate fatty change was found in 12 patients and geographic fatty change in one patient. No peritumoral fatty change was found in the study group. Correlation of fatty change with stiffness values was not performed in this study as previous studies have already showed that fatty change in liver does not affect stiffness measurement.

Imaging Technique

MRI was performed with a 1.5-T system (Signa, GE Healthcare) with a phased-array torso coil. The standard liver imaging protocol included the following sequences: coronal single-shot fast spinecho T2-weighted sequence, respiration-triggered fast spin-echo T2-weighted sequence or axial breath-hold fast recovery fast spin-echo T2-weighted sequence, axial dual-echo in- and out-of phase spoiled gradient-echo sequence, axial dynamic 3D fat-saturated spoiled gradient-echo sequence (liver acquisition with volume acceleration) before and after administration of a contrast agent, and delayed 2D axial fast spoiled gradient-echo sequence. Gadodiamide (Omniscan, GE Healthcare) 0.1 mmol/kg or gadobenate dimeglumine (MultiHance, Bracco) 0.05 mmol/kg was injected IV at a rate of 2-3 mL/s with an automated injector and was followed by a 30-mL saline flush. Arterial, portal venous, and delayed phase images were obtained for all patients. A 2-mL test bolus was administered to determine scan delay after contrast injection to optimize the arterial phase acquisition. All sequences were performed with a breath-hold at end-inspiration.
Fig. 1 77-year-old man with cholangiocarcinoma.
A, Gadolinium-enhanced T1-weighted MR image shows enhanced tumor (arrow) involving left hepatic duct and invading surrounding left lobe of liver.
B, Elastogram shows mean stiffness of tumor (arrow) is 15.5 kPa and that of liver parenchyma is 4.3 kPa.
Fig. 2 78-year-old man with cryptogenic cirrhosis and biopsy-proven hepatocellular carcinoma.
A and B, Dynamic gadolinium-enhanced T1-weighted MR images show enhanced tumor (arrow) in right lobe of liver during arterial phase (A) with washout during portal venous phase (B). Arrowhead = fat-containing region within tumor.
C and D, Dynamic gadolinium-enhanced T1-weighted MR images show enhanced tumor (arrow) in right lobe of liver during arterial phase (A) with washout during portal venous phase (B). Arrowhead = fat-containing region within tumor.

MR Elastography

MRE was performed at the end of the examination after the standard MRI protocol. A 19-cm-diameter 1.5-cm-thick cylindric passive driver developed within our institution was placed against the right chest wall over the liver with the center of the driver at the level of the xiphisternum. The passive driver was held in place with an abdominal binder. Continuous acoustic vibration at 60 Hz transmitted from an active driver to the passive driver through a flexible vinyl tube was used to produce propagating shear waves in the liver. A test vibration was first applied on the patient to familiarize the patient with the vibration [11]. The MRE sequence was performed with either a body coil or a torso-array coil. The choice of coil was based on patient size and need to accommodate the passive driver. The passive driver was easily introduced between the patient's chest or abdominal wall and the phased-array coil without an effect on image quality. Some large patients, for whom MRI or MRE could not be performed with a phasedarray coil, underwent MRE with a body coil alone.
The propagating shear waves were imaged with a modified phase contrast, gradient-echo sequence (MRE sequence) for collection of axial wave images sensitized along the through-plane direction of motion. The sequence parameters were TR/TE, 100/25.6; bandwidth, ±31.25 kHz; flip angle, 30°; field of view, 32-42 cm; matrix size, 256 × 96; slice thickness, 6-10 mm; gap, 2 mm. There was no special limitation on the field of view of the MRE sequence. Four to eight MRE slices were obtained for each patient. The total acquisition time was split into four periods of suspended respiration of 16 seconds for acquisition of wave images at four phase offsets. To obtain a consistent position of the liver for each phase offset, patients were asked to hold their breath at the end of expiration. The tumors were identified on T2- and contrast-enhanced T1-weighted MR images, and the MRE slice was targeted to the tumor. The slices that had the largest cross section of tumor were selected. All tumors within the slice were analyzed. The slice thicknesses were 6-10 mm modified according to the size of the focal lesion studied. The smallest tumor was 14 mm and the largest 110 mm. We excluded tumors smaller than 10 mm to avoid the risk of a partial volume effect with the surrounding liver parenchyma.
Fig. 3 55-year-old woman with metastatic colon cancer in right lobe of liver and diffuse fatty liver.
A, T2-weighted image shows single hyperintense lesion (arrow) in periphery of right lobe of liver.
B, Wave image shows prolongation of shear wave through tumor (arrow) compared with surrounding normal liver parenchyma.
C, Elastogram shows tumor (arrow) as hot spot with stiffness value of 6.2 kPa, suggestive of malignant tumor. Finding was confirmed at surgery to be metastasis from colon cancer.

Generation of MR Elastograms

MR elastograms of the liver were obtained by processing the acquired images of propagating shear waves with a previously described local frequency estimation inversion algorithm [4, 21]. This algorithm combines local estimates of instantaneous spatial frequency over several scales to give robust estimates of shear stiffness. A gaussian bandpass filter was applied to the original wave data to remove low-frequency wave information caused by longitudinal waves and bulk motion and to eliminate high-frequency noise. The cutoff frequencies of the bandpass filter were carefully chosen to be far away from the dominant spatial frequencies observed in the liver data. The high-end spatial frequency cutoff value is 0.95 cm-1, which corresponds to a stiffness value of approximately 0.4 kPa. The low-end cutoff value is 0.125 cm-1, which corresponds to stiffness values greater than 23 kPa. Before applying the local frequency estimation inversion algorithm, we used eight motion direction filters [22] evenly spaced between 0° and 360° and combined in a weighted least-squares method to improve the performance of the algorithm. This step was taken because complex interference of shear waves from all directions can produce areas with low shear displacement amplitude. All of these processing steps were applied automatically, that is, without human intervention, to produce quantitative maps in which tissue shear stiffness was measured in kilopascals.
Mean shear stiffness of the tumor was calculated with a manually specified region of interest (ROI). The ROI was placed by one reader, who was not aware of the final diagnosis of the solid tumors. The reader was experienced in interpreting MRE images and elastograms. The ROIs were oval or circular and covered most of the tumor in the magnitude image obtained with the MRE sequence and then copied to the stiffness map, which gave the stiffness values in kilopascals. For data analysis, lesions were assessed individually and compared across the entire group of lesions rather than grouped by patient.
The stiffness values of tumor-free hepatic parenchyma were calculated by placing multiple ROIs (at least three) in the parenchyma 2-3 cm away from the tumors. Every attempt was made to use the values in the image that showed mostly tumor-free liver parenchyma, preferably a slice that did not show any tumor. The ROIs were circular and 1-3 cm in diameter and were placed in the region of the parenchyma excluding vessels. The mean value from multiple ROIs was calculated.

Statistical Analysis

Statistical analysis was performed with commercially available software (SAS version 9.01, SAS Institute). One-way analysis of variance was performed for comparison of four groups of tissues: normal, fibrotic, benign, and malignant. Pairwise comparison was done for the different groups and tumor types; the Waller-Duncan method for correcting multiple comparisons was used. The analysis model was focused on identifying commonalities in shear stiffness by tumor type. The overall level of statistical significance was set at an alpha value of 0.05.

Results

MRE was technically successful for all 44 lesions studied. All liver tumors were well illuminated on the wave images. The patients tolerated the examination well, and no adverse effects were reported. Malignant liver tumors (10.1 kPa; 95% CI, 8.7-11.4) had significantly greater mean shear stiffness than benign tumors (2.7 kPa; 95% CI, 2.4-3.0) (p < 0.001). Malignant tumors also had significantly greater shear stiffness than normal liver parenchyma (2.3 kPa; 95% CI, 2.1-2.4) (p < 0.001) and fibrotic liver parenchyma (5.9 kPa; 95% CI, 4.5-7.2) (p < 0.001) (Table 1). Fibro tic livers were significantly stiffer than benign tumors (p < 0.001) and normal liver tissue (p < 0.001). The mean stiffness of benign tumors was not significantly different from that of normal liver parenchyma (p = 0.13).
TABLE 1: Shear Stiffness Values in Liver Parenchyma and Liver Tumors
Shear Stiffness (kPa)
GroupnMean ± SDRange
Malignant tumors3110.1 ± 3.66.2-19.6
Benign tumors132.7 ± 0.41.6-3.2
Fibrotic liver165.9 ± 2.53.1-12.2
Normal liver
13
2.3 ± 0.3
1.8-2.8
Cholangiocarcinoma (Fig. 1) and HCC (Fig. 2) had significantly greater stiffness than fibrotic liver, benign tumors, and normal liver parenchyma. Metastatic tumors (Fig. 3) were not significantly stiffer than fibrotic liver but were stiffer than all benign tumors and normal liver parenchyma. Cholangiocarcinoma (16.2 kPa) had significantly greater mean stiffness than HCC (10.3 kPa) (p < 0.001) and metastasis (7.6 kPa) (p < 0.001) (Table 2). HCC was significantly stiffer than metastatic lesions (p < 0.001).
TABLE 2: Shear Stiffness Values of Individual Tumor Types
Shear Stiffness (kPa)
TumornMean ± SDRange
Focal nodular hyperplasia32.7 ± 0.22.4-2.9
Hemangioma92.7 ± 0.51.6-3.2
Hepatic adenoma13.1 
Cholangiocarcinoma516.2 ± 3.410.8-19.6
Hepatocellular carcinoma1210.3 ± 2.07.6-14.2
Metastasis
14
7.6 ± 1.7
6.2-12.2
Fig. 4 39-year-old man with hepatic adenoma.
A, T2-weighted MR image (A) shows hyperintense 8-cm adenoma (arrow) in right lobe of liver.
B, Gadolinium-enhanced 3D spoiled gradient-recalled echo MR image shows intense arterial phase enhancement (arrow). Washout was evident in portal venous phase (not shown).
C, Axial MR elastographic wave image shows good illumination of tumor (circle). Waves in tumor have slightly longer wavelength than those in surrounding normal liver parenchyma.
D, Elastogram with region of interest corresponding to tumor shows shear stiffness value of tumor is 3.1 kPa and of surrounding liver is 2.4 kPa.
Among the benign tumors, hepatic adenoma (Fig. 4) had the greatest stiffness, but the value was not significantly different from that of hemangioma (Fig. 5), focal nodular hyperplasia, normal liver parenchyma, and fibrotic liver. Focal nodular hyperplasia similarly had stiffness values overlapping those of the other benign tumors and normal liver. Hemangioma, however, had significantly less shear stiffness (2.7 kPa; 95% CI, 2.3-3.1) than fibrotic liver (p < 0.001). The stiffness of hemangioma was not significantly different from that of normal liver.
A cutoff value of 5 kPa was accurate (100%) for differentiation of malignant tumors from benign tumors and normal liver parenchyma (Fig. 6). The stiffness values of fibrotic liver overlapped those of malignant tumors over a wide range. There was minimal overlap of the stiffness values of fibrotic liver and those of benign tumors. A linear correlation was found between size of the lesion and shear stiffness, but this trend was not significant (R2 = 0.20).

Discussion

Our preliminary study results show that MRE is feasible for imaging and characterizing solid liver tumors. Malignant tumors had greater stiffness values than benign tumors and normal liver parenchyma. Our results suggest that a threshold value of approximately 5.0 kPa may be useful for differentiating benign focal masses from malignant tumors. In the absence of a focal mass, however, stiffness values in liver parenchyma may be greater than 5.0 kPa owing to hepatic fibrosis and cirrhosis [15].
The shear stiffness values of normal liver parenchyma and cirrhotic liver in our study are similar to those reported earlier [13-15]. These results show that MRE is a robust technique for estimation of liver stiffness. Cholangiocarcinoma had the most stiffness among the tumors. This finding was not surprising because these tumors are known to be scirrhous and to have a fibrous or desmoplastic stroma [23]. Hepatocellular carcinoma was significantly less stiff than cholangiocarcinoma. This difference might have occurred because some of the HCCs in our study group had a fat component, which can result in a lower stiffness value for HCC. The metastasis group in our study had less stiffness than the cholangiocarcinoma or HCC group. The metastatic lesions were from colonic and neuroendocrine tumors, and these histologic types may have lower stiffness values. Although there appears to be a trend in stiffness values among malignant tumors, studies with larger numbers of tumor types are needed to establish the stiffness values.
The exact cause of high shear stiffness in malignant tumors is not known. Tumors proliferate within a mechanically restricted microenvironment, and tissue is a mechanical elastic solid [24]. Tumor rigidity probably reflects an elevation in interstitial tissue pressure and solid stress due to altered vasculature and tumor expansion [25], an increase in the elastic modulus of transformed cells mediated by an altered cytoarchitecture [26], and matrix stiffening linked to fibrosis [24]. The results of this study are in alignment with the subjective experience of surgeons, who routinely palpate liver to identify malignant lesions as structures that tend to be stiffer than normal liver and benign lesions. The results support the hypothesis that quantitative assessment of shear stiffness is helpful in characterizing liver masses.
Benign tumors were significantly less stiff than all of the malignant tumors and the fibrotic liver in our study. Nonmalignant liver masses are increasingly being recognized with the widespread use of sonography, CT, and MRI. Most of these lesions are detected incidentally in patients who do not have symptoms. Our results indicate the potential use of MRE for characterization of incidentally found mass lesions and for follow-up. We speculate that another potential application may be assessment of tumors after treatment to ascertain response. Because it is sensitive to changes in tumor stiffness, MRE may be useful for detecting changes after specific antitumor therapy.
Fig. 5 51-year-old woman with hemangioma in liver.
A and B, T1-weighted images during arterial (A) and delayed (B) phases show enhancement pattern typical of hemangioma (arrow).
C, Stiffness map shows stiffness of hemangioma (circle) is 3.2 kPa, whereas that of surrounding liver parenchyma is 2.3 kPa.
Fig. 6 —Graph shows box plots of shear stiffness of tissues. Cutoff value of 5 kPa separates malignant tumors from benign tumors and normal liver. Stiffness values of fibrotic liver overlap those of benign and malignant tumors.
Our study had limitations. First, the sample size was relatively small, and benign liver tumors were not well represented. Second, histologic proof was not available for each tumor nodule. This limitation was unavoidable because it is not common practice at our institution or others to obtain histologic proof when imaging criteria for HCC are met. In cases of metastasis, some of the nodules did not have histologic confirmation but were regarded as metastatic because they had similar features on PET and octreotide scans. Furthermore, some of the metastatic nodules were confirmed with surgical palpation and intraoperative sonography. Some of the cases of chronic liver diseases did not have histologic proof but were diagnosed with laboratory and imaging criteria. We did not attempt to differentiate causes of liver fibrosis because the numbers of cases were small in each etiologic group. The results provide motivation for conducting a study with a larger number of lesions with histologic proof to confirm the preliminary findings.
The third limitation is with regard to the MRE technique used in this study. Planar wave imaging was performed with 2D wave inversion. The inversion process does not take into account propagation of waves at an angle relative to the plane of section. Particularly for small structures, this problem can yield stiffness values that may be incorrectly low owing to partial volume effects and edge effects in the inversion algorithm. This problem occurs whenever a lesion is smaller than the wavelength of the shear wave used. Therefore, one may have to use a higher frequency and smaller-wavelength acoustic waves to accurately estimate the stiffness of smaller structures. High-frequency waves, how ever, are more attenuated than lower-frequency waves in the liver.
Even with the limitations, the technology shows promise. The best approach in the future may be to use 3D acquisition and inversion of wave data to address the problems. To reduce acquisition time, 3D MRE can be performed with a reduced number of phase offsets if necessary. These techniques are currently under development.
MRE is a feasible technique for quantitative evaluation of the mechanical properties of liver masses, offering new parameters for tissue characterization with MRI. The technique can be readily combined with conventional MRI of the abdomen. MRE shows promise for improving characterization of liver tumors.

Acknowledgments

The authors thank Stephen S. Cha, Division of Biostatistics, Mayo Clinic, Rochester, for assistance in statistical analysis.

Footnotes

Address correspondence to R. L. Ehman.
Supported by National Institutes of Health (NIH) grant EB001981.

References

1.
Paszek MJ, Zahir N, Johnson KR, et al. Tensional homeostasis and the malignant phenotype. Cancer Cell 2005; 8:241-254
2.
Duck FA. Physical properties of tissue: a comprehensive reference book. San Diego, CA: Academic Press, 1990:225-248
3.
Sarvazyan A, Goukassian D, Maevsky E, et al. Elasticity imaging as a new modality of medical imaging for cancer detection. In: Proceedings of an international workshop on interaction of ultrasound with biological media. Valenciennes, France;1994 : 68-81
4.
Manduca A, Oliphant TE, Dresner MA, et al. Magnetic resonance elastography: non-invasive mapping of tissue elasticity. Med Image Anal 2001; 5:237-254
5.
Ophir J, Cespedes I, Ponnekanti H, Yazdi Y, Li X. Elastography: a quantitative method for imaging the elasticity of biological tissues. Ultrason Imaging 1991; 13:111-134
6.
Muthupillai R, Lomas DJ, Rossman PJ, Greenleaf JF, Manduca A, Ehman RL. Magnetic resonance elastography by direct visualization of propagating acoustic strain waves. Science 1995; 269:1854-1857
7.
Garra BS, Cespedes EI, Ophir J, et al. Elastography of breast lesions: initial clinical results. Radiology 1997; 202:79-86
8.
Cochlin DL, Ganatra RH, Griffiths DF. Elastography in the detection of prostatic cancer. Clin Radiol 2002; 57:1014-1020
9.
Lyshchik A, Higashi T, Asato R, et al. Thyroid gland tumor diagnosis at US elastography. Radiology 2005; 237:202-211
10.
Lyshchik A, Higashi T, Asato R, et al. Cervical lymph node metastases: diagnosis at sonoelastography—initial experience. Radiology 2007; 243:258-267
11.
Muthupillai R, Ehman RL. Magnetic resonance elastography. Nat Med 1996; 2:601-603
12.
Ehman RL, Crues JV, Lenkinski RE, et al. Magnetic resonance. Radiology 1996; 198:920-926
13.
Rouviere O, Yin M, Dresner MA, et al. MR elastography of the liver: preliminary results. Radiology 2006; 240:440-448
14.
Huwart L, Peeters F, Sinkus R, et al. Liver fibrosis: non-invasive assessment with MR elastography. NMR Biomed 2006; 19:173-179
15.
Yin M, Talwalkar JA, Glaser KJ, et al. Assessment of hepatic fibrosis with magnetic resonance elastography. Clin Gastroenterol Hepatol 2007; 5:1207-1213
16.
McKnight AL, Kugel JL, Rossman PJ, Manduca A, Hartmann LC, Ehman RL. MR elastography of breast cancer: preliminary results. AJR 2002; 178:1411-1417
17.
Sinkus R, Tanter M, Xydeas T, Catheline S, Bercoff J, Fink M. Viscoelastic shear properties of in vivo breast lesions measured by MR elastography. Magn Reson Imaging 2005; 23:159-165
18.
Xu L, Lin Y, Han JC, Xi ZN, Shen H, Gao PY. Magnetic resonance elastography of brain tumors: preliminary results. Acta Radiol 2007; 48:327-330
19.
Bruix J, Sherman M, Llovet JM, et al. Clinical management of hepatocellular carcinoma: conclusions of the Barcelona-2000 EASL conference. J Hepatol 2001; 35:421-430
20.
Bruix J, Sherman M. Management of hepatocellular carcinoma. Hepatology 2005; 42:1208-1236
21.
Westin CF, Wigstrom L, Loock T, Sjoqvist L, Kikinis R, Knutsson H. Three-dimensional adaptive filtering in magnetic resonance angiography. J Magn Reson Imaging 2001; 14:63-71
22.
Manduca A, Lake DS, Kruse SA, Ehman RL. Spatio-temporal directional filtering for improved inversion of MR elastography images. Med Image Anal 2003; 7:465-473
23.
Goodman ZD. Neoplasms of the liver. Mod Pathol 2007; 20[suppl 1]: S49-S60
24.
Paszek MJ, Weaver VM. The tension mounts: mechanics meets morphogenesis and malignancy. J Mammary Gland Biol Neoplasia 2004; 9:325-342
25.
Padera TP, Stoll BR, Tooredman JB, Capen D, di Tomaso E, Jain RK. Pathology: cancer cells compress intratumour vessels. Nature 2004; 427:695
26.
Beil M, Micoulet A, von Wichert G, et al. Sphingosylphosphorylcholine regulates keratin network architecture and visco-elastic properties of human cancer cells. Nat Cell Biol 2003; 5:803-811

Information & Authors

Information

Published In

American Journal of Roentgenology
Pages: 1534 - 1540
PubMed: 18492904

History

Submitted: September 7, 2007
Accepted: December 13, 2007

Keywords

  1. benign tumors
  2. fibrotic liver
  3. liver tumors
  4. malignant tumors
  5. MR elastography

Authors

Affiliations

Sudhakar K. Venkatesh
Department of Radiology, Mayo Clinic and Foundation, 200 First St. SW, Opus Bldg., Rochester, MN 55905.
Department of Diagnostic Radiology, Yong Loo Lin School of Medicine, National University of Singapore, Singapore.
Meng Yin
Department of Radiology, Mayo Clinic and Foundation, 200 First St. SW, Opus Bldg., Rochester, MN 55905.
James F. Glockner
Department of Radiology, Mayo Clinic and Foundation, 200 First St. SW, Opus Bldg., Rochester, MN 55905.
Naoki Takahashi
Department of Radiology, Mayo Clinic and Foundation, 200 First St. SW, Opus Bldg., Rochester, MN 55905.
Philip A. Araoz
Department of Radiology, Mayo Clinic and Foundation, 200 First St. SW, Opus Bldg., Rochester, MN 55905.
Jayant A. Talwalkar
Department of Gastroenterology and Hepatology, Mayo Clinic, Rochester, MN.
Richard L. Ehman
Department of Radiology, Mayo Clinic and Foundation, 200 First St. SW, Opus Bldg., Rochester, MN 55905.

Metrics & Citations

Metrics

Citations

Export Citations

To download the citation to this article, select your reference manager software.

Articles citing this article

Media

Figures

Other

Tables

Share

Share

Copy the content Link

Share on social media