Published online before print December 19, 2007, 10.1148/radiol.2462061312
(Radiology 2007;246:472.)
A more recent version of this article appeared on December 1, 2007
© RSNA, 2007
CT Histogram Analysis: Differentiation of Angiomyolipoma without Visible Fat from Renal Cell Carcinoma at CT Imaging1
Ji Yeon Kim, MD,
Jeong Kon Kim, MD,
Namkug Kim, MS, and
Kyoung-Sik Cho, MD
1 From the Department of Radiology, Asan Medical Center, University of Ulsan, 388-1 Poongnap-dong, Songpa-gu, Seoul 138-736, Korea. Received July 30, 2006; revision requested October 3; revision received January 16, 2007; accepted February 22; final version accepted July 19.
Address correspondence to J.K.K. (e-mail: rialto{at}amc.seoul.kr).
 |
ABSTRACT
|
|---|
Purpose: To retrospectively evaluate the diagnostic performance of computed tomographic (CT) histogram analysis for differentiating angiomyolipoma (AML) without visible fat from renal cell carcinoma (RCC) at CT, by using pathologic analysis and clinical diagnosis as the reference standard.
Materials and Methods: This retrospective study was approved by the institutional review board; informed consent was waived. The authors reviewed the medical records of 144 patients with pathologic confirmation of RCC or AML (105 men, 39 women; mean age, 52 years). Analysis of unenhanced CT histograms was performed on 34 AMLs without visible fat at CT and 110 size-matched RCCs. The percentages of voxels and pixels were compared in the two groups according to the CT number categories. The diagnostic performance of CT histogram analysis in differentiating AML from RCC was determined by using receiver operating characteristic (ROC) analysis.
Results: The percentages of voxels and pixels with a CT number less than –30 HU (2.7% and 3.4% vs 0.1% and 0.0%), less than –20 HU (4.3% and 5.1% vs 0.2% and 0.1%), less than –10 HU (7.0% and 8.1% vs 0.6% and 0.4%), and less than 0 HU (12.0% and 13.9% vs 2.0% and 2.0%) were significantly greater in the AML group than in the RCC group (P < .01), respectively. The area under the ROC curve was as high as 0.706 when a pixel percentage with a CT number less than –10 HU was used as a differentiating parameter. Corresponding to the specificity of 100% for differentiating AML from RCC, the sensitivity was as high as 20% when a pixel percentage of 6% with a CT number less than –10 HU was used as a criterion.
Conclusion: CT histogram analysis may be useful for differentiating AML without visible fat from RCC at CT.
© RSNA, 2007
|
INTRODUCTION
|
|---|
Angiomyolipoma (AML) is a common benign neoplasm of the kidney; it consists of a variable amount of fat, smooth muscle, and blood vessels. Identification of the intratumoral fat is crucial for making a correct diagnosis. However, for some AMLs, intratumoral fat cannot be identified at computed tomography (CT) because of the extremely small amount or immaturity of the fat (1–7). These tumors can mimic renal cell carcinoma (RCC), leading to unnecessary surgery.
There have been studies to analyze the imaging findings of AML without visible fat at CT, attempting to differentiate it from RCC. Kim et al (2) reported some helpful contrast material–enhanced CT findings, including homogeneous enhancement and prolonged enhancement pattern. Another recent study by Kim et al (3) showed that chemical shift magnetic resonance (MR) imaging accurately depicted the presence of the intratumoral fat that was not detected at CT, improving the sensitivity in differentiating AML from other renal neoplasms to 96%.
In clinical practice, biphasic or triphasic CT, including unenhanced CT, is the first-line imaging method used to evaluate renal tumors. Many previous studies have shown the usefulness of CT attenuation measurement and histogram analysis in the diagnosis of various diseases (8–11). We hypothesized that if the presence of fat in a tumor could be proved by using a histogram analysis of unenhanced CT, the diagnostic performance at CT would improve. Thus, we conducted our study to retrospectively evaluate the diagnostic performance of CT histogram analysis for differentiating AML without visible fat from RCC at CT by using pathologic analysis and clinical diagnosis as the reference standards.
|
MATERIALS AND METHODS
|
|---|
Our retrospective study was approved by the Institutional Review Board of Asan Medical Center (Seoul, Korea), and informed consent was waived.
Patient Selection
AML without visible fat at CT.—A computerized search of the medical records at our institution from June 2001 to March 2005 generated a list of 58 patients who had AML proved by pathologic analysis. In these patients, two experienced radiologists (S.B.P., K.C., with 10 and 20 years experience in renal CT, respectively) independently reviewed the unenhanced scans with regard to the presence of fat attenuation in the tumor. Tumors in which neither radiologist could detect fat attenuation by visual inspection were found in 21 patients and were included in our study.
We also included 13 patients who had been clinically diagnosed as having AMLs without visible fat at CT and MR imaging during the same period by using the following criteria: (a) The same two radiologists independently confirmed the absence of visible fat on unenhanced CT scans; (b) there was no tumor growth for at least 24 months (mean, 33 months; range, 24–60 months) on follow-up CT scans; and (c) the signal intensity index on double-echo gradient-echo chemical shift MR images was 40% or higher. This value (signal intensity index of 40%), derived from past study data, corresponded to a fat fraction of 30% in a phantom with a similar T1 value in the adrenal gland (12). Furthermore, the signal intensity index of 40% was twice as great as a cutoff value of 20%, which yielded a specificity of 100% in the diagnosis of adenoma in patients with hyperattenuating (>10 HU) adrenal masses in a previous study (13).
Finally, a total of 34 patients (21 with pathologic proof and 13 with clinical diagnostic proof of AML) without visible fat at CT were included in our study (mean age ± standard deviation, 50 years ± 12; range, 31–77 years; 15 women, 19 men). The mean maximum tumor diameter in the AML group was 2.3 cm (range, 0.8–3.5 cm).
Renal cell carcinoma.—During the same period, 662 patients were diagnosed with histologically confirmed RCCs. From those patients, 167 had tumors 3.5 cm or less in maximum diameter (mean, 2.5 cm; range, 1.0–3.5 cm) and were chosen to perform a size-matched comparison with the AML group.
Of these 167 patients, 46 were excluded, including those who had not undergone CT examinations at our institution (n = 29), had no unenhanced CT images (n = 13), and had multiple RCCs related to von Hippel Lindau disease (n = 4). In addition, 11 patients were excluded because their tumors were embedded in the renal parenchyma, which made it difficult to draw regions of interest (ROIs) on tumors. Therefore, 110 patients with 110 pathologically proved RCCs were included in our study (mean age, 54 years ± 11; range, 25–78 years; 24 women, 86 men).
CT Examinations
All CT examinations were performed with either a four-channel multi–detector row helical CT scanner (LightSpeed QX/i; GE Medical Systems, Milwaukee, Wis) (n = 112; 89 RCCs, 23 AMLs) or a 16-channel multi–detector row helical CT scanner (Sensation 16; Siemens Medical System, Erlangen, Germany) (n = 32; 21 RCCs, 11 AMLs).
All patients underwent biphasic CT scanning that included unenhanced, corticomedullary phase, and early excretory phase scanning. Unenhanced scanning, which covered the entire volume of the spleen and both kidneys, was performed after administration of the oral contrast material. All patients received 500–900 mL oral contrast material (2% barium sulfate suspension, E-Z-CAT; E-Z-Em, Westbury, NY) 30 minutes prior to their CT examination. Then, intravenous contrast material (iopromide, Ultravist 300, Schering, Berlin, Germany; or iopamidol, Iopamiro 300, Bracco, Milan, Italy) was administered in an antecubital vein with a power injector (Percupump II; E-Z-Em, Westbury, NY) at a dose of 2 mL per kilogram of body weight at a rate of 3 mL/sec to a maximum of 160 mL. Corticomedullary phase scans were obtained from the kidneys through the aorta bifurcation by using an automatic bolus triggering technique that started scanning when the CT number of an ROI in the aorta at the renal artery level reached 100 HU. The scan delay for early excretory phase scanning was 120–150 seconds.
The scanning parameters for unenhanced scanning with four–detector row helical CT were as follows: a four-detector array at 20-mm collimation; table speed, 30 mm per rotation (37.5 mm/sec); pitch, 1.5 (equivalent to a section pitch of 6 in high-speed mode); reconstruction interval, 5 mm; tube voltage, 120 kV; and tube current, 180–230 mA.
The scanning parameters for unenhanced scanning with a 16-channel multi–detector row helical CT were as follows: a 16-detector array at 24-mm collimation; table speed, 24–36 mm per rotation (48 mm/sec); pitch, 1.0–1.5; reconstruction interval, 5 mm; tube voltage, 120 kV; and tube current (determined by using an automatic dose-modulation technique), 170–250 mA.
CT Histogram Analysis
To reduce the variability in the CT number according to various patient factors and CT acquisition techniques, our study only included patients with CT numbers of air ranging from –1005 to –997 HU. For this process, a radiologist (J.Y.K.) measured the CT number of the air by drawing an ROI anteriorly to the inferior edge of the xyphoid process in each patient, the area of which was constant in all patients (314 mm2).
A radiologist (J.Y.K.) measured noise level (defined as the standard deviation of the mean attenuation) in a standard 1 x 1 cm circular ROI in the abdominal aorta, at the level of the renal mass. Noise level was the quantitative parameter used to assess image quality and was compared between the AML and RCC groups.
In addition, we calculated effective milliampere-second levels for each scan at the specific location of each tumor. Effective tube current levels were calculated by multiplying the gantry rotation period by the tube current and dividing the total by the beam pitch.
For CT histogram analysis, in-house software (AView 10; MIL, Seoul, Korea) was developed by the Medical Imaging Laboratory at our institution. A staff radiologist (J.K.K., with 10 years experience in renal tumor imaging), who was unaware of the final diagnosis in each patient, drew ROIs on the scans including the entire tumor area. During this process, the CT scans were magnified three- or four-fold, and the window width and level were arbitrarily set to maximize tumor visualization on the software.
To exclude surrounding renal parenchyma and perinephric fat from the ROIs, the line was carefully drawn while trying to maintain an approximate distance of 2–3 mm from the tumor margin by referring to contrast-enhanced CT images. Furthermore, in the upper- and lowermost images containing the tumors, the ROIs were not drawn to prevent the partial volume averaging artifact by perinephric fat. Time spent in postprocessing by the radiologist varied according to tumor volume but was less than 5 minutes in all patients.
The distribution of the CT numbers in the tumors was analyzed by using two methods. First, stacks of ROIs in each tumor were turned into a three-dimensional volume of interest and the distribution of the CT numbers according to voxel unit was calculated. Second, an ROI at the middle level of the tumor was chosen and the distribution of the CT number according to pixel units was calculated.
To systematically evaluate the distribution of the CT numbers in the tumors, the CT number ranges were arbitrarily classified as –1024 to –30 HU, –1024 to –20 HU, –1024 to –10 HU, and –1024 to 0 HU. Then the percentages of voxels or pixels were calculated and classified according to these categories.
It has been postulated (14) that RCC has various components, according to each subtype of RCC (ie, clear cell RCC), each with various amounts of intratumoral fat. Thus, we compared the CT number distribution in various subtypes of RCC. The subtype classification followed the classification system of the World Health Organization (15,16).
Statistical Analysis
The age and sex distribution were compared between the AML and RCC groups by using the independent sample t test and the Fisher exact test, respectively. The noise level of the image and effective milliampere-seconds for each scan at the level of each tumor were then compared between the AML and RCC groups by using the independent sample t test.
The number of patients with certain percentages of voxels was compared between the two groups according to CT number categories by using the Fisher exact test. The numbers and percentages of voxels and pixels, also classified according to CT number categories, were compared between the two groups by using the independent-sample t test.
By using the receiver operating characteristic (ROC) curve analysis, the diagnostic performance of the percentages of voxels and pixels, according to the CT number categories, was evaluated. From the ROC curves, the optimal cutoff values were extracted, which showed the best separation (the minimal false-negative and false-positive results) of the two groups. In addition, a cutoff value indicating a specificity of 100% for differentiating AML from RCC was extracted and the corresponding sensitivity was calculated from the ROC curves.
To evaluate the difference between CT histograms in various subtypes of RCC, the percentages of voxels and pixels, according to the CT number classification (see above), were compared by using the one-way analysis of variance.
|
RESULTS
|
|---|
The flow diagram for patient enrollment is shown in Figure 1.
Signal Intensity and Effective Tube Charge
The noise level was similar between the AML (mean, 11.8 HU ± 1.6; range, 9–15 HU) and RCC (mean, 11.8 HU ± 1.6; range, 9–15 HU) (P = .545) groups. The effective milliampere-second level was also similar between the AML (mean, 102.4 mAs ± 11.5; range, 80–127 mAs) and RCC (mean, 101.6 mAs ± 9.9; range, 83–144 mAs) groups (P = .127).
Age, Sex, and Tumor Size
The patient ages were similar in the AML and RCC groups (P = .072). However, the female–male ratio was significantly greater in the AML group (15:19) than in the RCC group (24:86) (P < .001).
CT Number Distribution in the Tumor
Voxels with respective CT numbers of less than –30 HU, less than –20 HU, less than –10 HU, and less than 0 HU were noted in eight (24%) patients, 14 (41%) patients, 20 (59%) patients, and 24 (71%) patients in the AML group and in five (5%) patients, 12 (11%) patients, 30 (27%) patients, and 62 (56%) patients in the RCC group. More patients in the AML group had a tumor with a CT number less than –30 HU (P = .003), less than –20 HU (P < .001), and less than –10 HU (P = .002), than did the RCC group. The number of patients who had tumors with a CT number less than 0 HU was similar between the two groups (P = .203).
The numbers and percentages of voxels (Table 1) with a CT number less than –30 HU, less than –20 HU, and less –10 HU were significantly greater in the AML group (P < .05) (Fig 2) than in the RCC group (P < .01 for percentages) (Fig 3). In addition, the numbers and percentages of pixels (Table 1) with a CT number less than –30 HU, less than –20 HU, less –10 HU, and less than 0 HU were significantly greater in the AML group than in the RCC group (P < .001).
Differentiation of AML from RCC
The ROC analysis for the percentages of voxels and pixels according to the CT number categories (Table 2; Figs 4, 5) showed that the voxel percentage area under the ROC curve (AUC) ranged from 0.598 to 0.695 and the pixel percentage AUC ranged from 0.595 to 0.706.
View larger version (23K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4: Graph shows ROC curve for voxel percentages in differentiating AML without visible fat from RCC at CT. AUCs with voxel percentage with CT number from –1024 to –30 HU, from –1024 to –20 HU, from –1024 to –10 HU, and from –1024 to 0 HU were 0.598, 0.660, 0.695, and 0.672, respectively.
|
|
View larger version (23K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5: Graph shows ROC curve for pixel percentage in differentiating AML without visible fat from RCC at CT. AUCs with voxel percentage with CT number from –1024 to –30 HU, from –1024 to –20 HU, from –1024 to –10 HU, and from –1024 to 0 HU were 0.595, 0.621, 0.706, and 0.673, respectively.
|
|
The cutoff voxel percentages corresponding to a specificity of 100% were 2.0% with a CT number less than –30 HU, 3.0% with a CT number less than –20 HU, 8.0% with a CT number less than –10 HU, and 27.0% with a CT number less than 0 HU. The sensitivities corresponding to those threshold values were 14% (five of 34) for a CT number less than –30 HU, 17% (six of 34) for a CT number less than –20 HU, 17% (six of 34) for a CT number less than –10 HU, and 17% (six of 34) for a CT number less than 0 HU.
The cutoff pixel percentages corresponding to a specificity of 100% were 1.0% with a CT number less than –30 HU, 2.0% with a CT number less than –20 HU, 6.0% with a CT number less than –10 HU, and 22.0% with a CT number less than 0 HU. The sensitivities corresponding to those threshold values were 17% (six of 34) for a CT number less than –30 HU, 17% (six of 34) for a CT number less than –20 HU, 20% (seven of 34) for a CT number less than –10 HU, and 20% (seven of 34) for a CT number less than 0 HU.
CT Histogram of RCC by Subtype
The RCC group was composed of 95 clear cell, 10 papillary, and five chromophobe RCCs; there were no collecting duct carcinomas. There was no statistical difference (Table 3) between the voxel percentages according to the CT number categories in three subtypes of RCC (P > .05).
|
DISCUSSION
|
|---|
In our study, CT histogram analysis showed significantly different distribution of CT numbers in tumors in AML without visible fat and in RCC at CT. According to our data, the percentages of voxels and pixels with a CT number less than –30 HU, less than –20 HU, less than –10 HU, and less than 0 HU were significantly greater in the AML group than in the RCC group. The AUC was as high as 0.706 when a pixel percentage with a CT number less than –10 HU was used as a differentiating parameter. Corresponding to a specificity of 100% for differentiating AML from RCC, the sensitivity was as high as 20% when a pixel percentage of 6% with a CT number less than –10 HU was used as a criterion.
Our data show that the absence of visible fat in a tumor does not necessarily indicate that the tumor has no fat. More than half the tumors in our study, 71% (24 of 34) of the AMLs and 56% (62 of 110) of the RCCs, had negative CT numbers for voxels. Therefore, when the decision regarding the presence of fat in a tumor is crucial for diagnosis, quantitative measurement of the CT number would be preferred to a decision made on the basis of visual inspection because a small amount of scattered fat cannot be readily identified on visual inspection.
Kim et al (3) have shown that chemical shift MR imaging provided satisfactory diagnostic performance, as the AUCs were 0.952–0.976 and the sensitivity and specificity were both greater than 90%. While the diagnostic performance of chemical shift MR imaging seems to be more accurate than that of CT histogram analysis, CT is the first-line study for evaluating renal masses because of its widespread availability, high speed, and lower cost. Therefore, when histogram analysis can be used to accurately diagnose AML without visible fat at CT (ie, with a specificity of 100%), additional MR imaging or unnecessary surgery can be avoided. In addition, we expect that CT diagnostic performance may improve when a decision is made on the combination of CT histogram analysis and biphasic contrast-enhanced CT findings.
To our knowledge, clear cell RCC has been most frequently noted of the various RCC subtypes as having intratumoral fat (8,17–20). However, according to our review, unusual RCCs containing fat have been sporadically reported and there has been no systematic evaluation with regard to the subtype of RCC and the frequency of intratrumoral fat. Regarding this issue, the results of our study, in which we compared the distribution of CT numbers in the tumor between various subtypes of RCCs, observed no significant difference.
In our study, the percentages of voxels and pixels with negative CT numbers were so small that the maximum percentages of voxels and pixels with a CT number less than –10 HU were only 7.0% and 8.1%, respectively. This small amount of negative CT numbers for voxels and pixels may be affected by various factors or biases. This problem is an inherent limitation of our study because only tumors without visible fat at CT were included.
The CT number may vary even in the same tissue according to various factors, including patient body habitus, motion artifact, ROI location, tube voltage and current, collimation, section thickness, and reconstruction kernel (12). Moreover, due to the recent rapid development of multi–detector row helical CT scanners, it is often not possible to obtain images in daily practice by using the same scanner with a single standardized protocol.
In our study, we also obtained CT images from two types of scanners (ie, four- and 16–detector row helical CT scanners), another study limitation. Furthermore, because the percentages of voxels and pixels in our study were so small, they might have been easily affected by various factors or biases.
To reduce this potential error, we tried to apply the same imaging acquisition parameters, including section thickness (5 mm on unenhanced scan), tube voltage, and reconstruction interval, and we selected patients by using the strict criterion that the CT number in the air was –1005 to –997 HU. In addition, our institution performs CT scanner calibration on a monthly basis to accurately determine the reference attenuation of the histogram analysis and to decrease the fluctuation of CT scanners from different vendors.
To obtain an accurate comparison of CT numbers, it is necessary to verify the noise level, which we did by measuring the standard deviation of CT attenuation in the aorta, which ranged from 9 to 15 HU. In spite of no significant difference of noise levels between the RCC and AML groups, our methods did not completely verify the noise level because individual normalization of noise levels was not performed.
To prevent inclusion of perinephric fat in the ROIs, we drew them while trying to keep a distance of approximately 2–3 mm from the tumor margin and excluded the upper- and lowermost sections. On the other hand, this attempt might have posed a risk that negative CT numbers for voxels existing in the tumor periphery were also excluded from the ROIs (yet another limitation).
We only evaluated tumors with a greatest diameter of 3.5 cm or less. Therefore, our criteria for differentiating AML from RCC cannot be used for large tumors because they may have a greater number of negative CT numbers for voxels. Furthermore, our study selected only cases with a CT number from –1005 to –997 of air. Consequently, our data cannot be generalized for all cases.
As a final limitation, our study included 13 patients who had been clinically but not pathologically diagnosed as having AML without visible fat. Although we applied strict criteria for the diagnosis of AML, we cannot exclude these tumors as indolent RCCs without pathologic proof.
In summary, our CT histogram analysis showed that AML without visible fat at CT had greater percentages of voxels and pixels with a CT number from –1024 to –30 HU, from –1024 to –20 HU, from –1024 to –10 HU, and from –1024 to 0 HU. By using these characteristics, some AMLs without visible fat can be differentiated from RCC at CT with a specificity of 100%. Thus, CT histogram analysis may be useful for differentiating AML without visible fat from RCC at CT.
|
ADVANCES IN KNOWLEDGE
|
|---|
- CT histogram analysis can be used to detect a small amount of fat that is not identified by visual inspection at unenhanced CT.
- CT histogram analysis showed that angiomyolipomas (AMLs) without visible fat at CT had a greater percentage of negative CT numbers for voxels than did renal cell carcinomas (RCCs).
|
IMPLICATION FOR PATIENT CARE
|
|---|
- CT histogram analysis may be useful for differentiating AML from RCC.
|
ACKNOWLEDGMENTS
|
|---|
The authors thank Bonnie Hami, MA, Department of Radiology, University Hospitals Health System, Cleveland, Ohio, for her editorial assistance in preparing the manuscript.
|
FOOTNOTES
|
|---|
Abbreviations: AML = angiomyolipoma • AUC = area under the ROC curve • RCC = renal cell carcinoma • ROC = receiver operating characteristic • ROI = region of interest
Author contributions: Guarantor of integrity of entire study, J.K.K.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, all authors; clinical studies, all authors; statistical analysis, J.K.K.; and manuscript editing, J.K.K.
Authors stated no financial relationship to disclose.
|
References
|
|---|
- Jinzaki M, Tanimoto A, Narimatsu Y, et al. Angiomyolipoma: imaging findings in lesions with minimal fat. Radiology 1997;205:497–502. [Abstract/Free Full Text]
- Kim JK, Park SY, Shon JH, Cho KS. Angiomyolipoma with minimal fat: differentiation from renal cell carcinoma at biphasic helical CT. Radiology 2004;230:677–684. [Abstract/Free Full Text]
- Kim JK, Kim SH, Jang YJ, et al. Renal angiomyolipoma with minimal fat: differentiation from other neoplasms on double-echo GRE chemical shift MR imaging. Radiology 2006;239:174–180. [Abstract/Free Full Text]
- Obuz F, Karabay N, Secil M, Igci E, Kovanlikaya A, Yorukoglu K. Various radiological appearances of angiomyolipomas in the same kidney. Eur Radiol 2000;10:897–899. [CrossRef][Medline]
- Ohkawa M, Kadoya M, Nonomura A. Renal angiomyolipoma composed primarily of smooth muscle element: diagnostic consideration. Urol Int 1995;54:230–233. [CrossRef][Medline]
- Hosokawa Y, Kinouchi T, Sawai Y, et al. Renal angiomyolipoma with minimal fat. Int J Clin Oncol 2002;7:120–123. [Medline]
- Lemaitre L, Claudon M, Dubrulle F, Mazeman E. Imaging of angiomyolipomas. Semin Ultrasound CT MR 1997;18:100–114. [CrossRef][Medline]
- Valvassori GE. CT densitometry in otosclerosis. Adv Otorhinolaryngol 1987;37:47–49. [Medline]
- Muller NL, Staples CA, Miller RR, Abboud RT. "Density mask": an objective method to quantitate emphysema using computed tomography. Chest 1988;94:782–787. [CrossRef][Medline]
- Boland GW, Lee MJ, Gazelle GS, Halpern EF, McNicholas MM, Mueller PR. Characterization of adrenal masses using unenhanced CT: an analysis of the CT literature. AJR Am J Roentgenol 1998;171:201–204. [Abstract/Free Full Text]
- Bae KT, Fuangtharnthip P, Prasad SR, Joe BN, Heiken JP. Adrenal masses: CT characterization with histogram analysis method. Radiology 2003;228:735–742. [Abstract/Free Full Text]
- Namimoto T, Yamashita Y, Mitsuzaki K, et al. Adrenal masses: quantification of fat content with double-echo chemical shift in-phase and opposed-phase FLASH MR images for differentiation of adrenal adenomas. Radiology 2001;218:642–646. [Abstract/Free Full Text]
- Haider MA, Ghai S, Jhaveri K, Lockwood G. Chemical shift MR imaging of hyperattenuating (>10 HU) adrenal asses: does it still have a role? Radiology 2004;231:711–716.
- Outwater EK, Bhatia M, Siegelman ES, Burke MA, Mitchell DG. Lipid in renal clear cell carcinoma: detection on opposed-phase gradient-echo MR images. Radiology 1997;205:103–107. [Abstract/Free Full Text]
- Bostwick DG, Eble JN, Murphy GP. Conference summary: diagnosis and prognosis of renal cell carcinoma—1997 workshop [abstr]. Cancer 1997;80:975–976. [CrossRef]
- Bostwick DG, Eble JN. Diagnosis and classification of renal cell carcinoma. Urol Clin North Am 1999;26:627–635. [CrossRef][Medline]
- Helenon O, Merran S, Paraf F, et al. Unusual fat-containing tumors of the kidney: a diagnostic dilemma. RadioGraphics 1997;17:129–144. [Abstract]
- Helenon O, Chretien Y, Paraf F, et al. Renal cell carcinoma containing fat: demonstration with CT. Radiology 1993;188:429–430. [Abstract/Free Full Text]
- Schuster TG, Ferguson MR, Baker DE, Schaldenbrand JD, Solomon MH. Papillary renal cell carcinoma containing fat without calcification mimicking angiomyolipoma on CT. AJR Am J Roentgenol 2004;183:1402–1404. [Free Full Text]
- Lesavre A, Correas JM, Merran S, Grenier N, Viellefond A, Helenon O. CT of papillary renal cell carcinomas with cholesterol necrosis mimicking angiomyolipomas. AJR Am J Roentgenol 2003;181:143–145. [Free Full Text]
TOP
ABSTRACT
INTRODUCTION
–30 HU but < –20 HU), 48 and 2% (CT number 
