| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Genitourinary Imaging |
1 From the Department of Diagnostic Radiology, Yale University School of Medicine, PO Box 208042, 333 Cedar St, New Haven, CT 06520-8042. Received March 29, 2006; revision requested May 30; revision received June 9; accepted July 7; final version accepted September 18. Address correspondence to G.M.I. (e-mail: gary.israel{at}yale.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
Materials and Methods: The Human Investigation Committee at our institution approved this study with waiver of informed consent. This study was compliant with the HIPAA. Fifty-four pathologically proved renal cell carcinomas in 54 patients (36 men and 18 women; average age, 53 years; range, 23–90 years) and 56 high-attenuation renal cysts in 51 patients (30 men and 21 women; average age, 63 years; range, 28–86 years) were retrospectively evaluated at unenhanced CT. Two independent readers reviewed randomized unenhanced CT images and obtained Hounsfield unit readings of each mass. A subjective determination of lesion heterogeneity was also performed by using a four-point scale (1: homogeneous, 2: mildly heterogeneous, 3: moderately heterogeneous, 4: markedly heterogeneous). Statistical analysis was performed by using Bland-Altman regression tree, classification and regression tree, and Shapiro-Wilk normality test.
Results: The average attenuation of cysts for reader 1 was 53.4 HU (range, 23–113 HU) and for reader 2 was 53.8 HU (range, 21–108 HU). The average attenuation of neoplasms for reader 1 was 34.7 HU (range, 21–60 HU) and for reader 2 was 38.4 HU (range, 22–60 HU). For cyst heterogeneity, a score of 1 was given in 55 of 56 (98%) cysts for reader 1 and in 53 of 56 (95%) cysts for reader 2. For neoplasm heterogeneity, a score of 1 was given in 35 of 54 (65%) neoplasms for reader 1 and in 36 of 54 (67%) for reader 2. Given the distribution of cyst and tumor attenuation values and lesion heterogeneity, a homogeneous mass measuring 70 HU or greater at unenhanced CT has a greater than 99.9% chance of representing a high-attenuation renal cyst.
Conclusion: The findings from this study may help differentiate high-attenuation renal cysts from renal cell carcinomas at unenhanced CT and may suggest the next appropriate imaging study for definitive characterization.
© RSNA, 2007
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
A computerized search of radiology reports from January 1, 2001, through January 1, 2005, in our radiology information system (IDX Imagecast version 10.3.9.91; IDX Systems, Burlington, Vt) was performed for high-attenuation renal cysts. Reports were searched for CT scans of the abdomen performed with and without intravenous (IV) contrast in which "hemorrhagic cyst," "high-attenuation cyst," or "proteinaceous cyst" was used to describe a renal mass. This yielded a total of 51 patients with 56 renal masses that met these criteria. The group consisted of 30 men and 21 women (mean age: 63 years; range: 28–86 years). The average size of the renal cyst (largest axial diameter) was 1.8 cm (range: 1.0–6.1 cm). Patients with autosomal dominant polycystic kidney disease or von Hippel–Lindau disease were excluded.
CT Technique
All CT examinations were performed on helical scanners. For the renal neoplasms, the scanners used were as follows: CTi (n = 15), LightSpeed (four-section) (n = 30), and LightSpeed (16-section) (n = 9) (GE Healthcare, Milwaukee, Wis). The section collimation for the renal neoplasms varied (10 mm, n = 1; 7 mm, n = 4; 5 mm, n = 19, 3.75 mm, n = 1; 2.5 mm, n = 29). For the renal cysts, the scanners used were as follows: CTi (n = 6), LightSpeed (four-section) (n = 39), and LightSpeed (16-section) (n = 6) (GE Healthcare). The section collimation for the high-attenuation renal cysts varied (10 mm, n = 4; 5 mm, n = 25; 2.5 mm, n = 27). For the high-attenuation renal cysts, unenhanced and contrast-enhanced images were performed by using the same parameters (peak voltage, milliampere-second, and section collimation) for both series. The contrast-enhanced images were obtained after 150 mL of IV iohexol (Omnipaque 300, Nycomed Amersham, Princeton, NJ) injected at 4 mL/sec and a scan delay of 70–100 seconds.
Image Interpretation
A single author (G.M.I.) with 10 years experience in genitourinary CT reviewed the high-attenuation renal cysts on the unenhanced and contrast-enhanced examinations to confirm their diagnosis. Region of interest (ROI) measurements (to obtain Hounsfield unit readings) from similar locations of the cysts were performed for both acquisitions. A high-attenuation cyst was confirmed if the change in attenuation (from the unenhanced to the contrast-enhanced images) was 10 HU or less (3).
Two readers (A.I.J. and A.N.R., with 2 and 10 years experience in body CT image interpretation, respectively), who were blinded to the final diagnoses, independently reviewed the randomized unenhanced CT images of the renal neoplasms and high-attenuation renal cysts. To avoid confusion with other lesions in the kidneys, the readers were made aware of the location of the lesions of interest. For each lesion, manually defined circular or oval ROI measurements were performed on the unenhanced CT images to obtain attenuation readings. For homogeneous lesions, the ROI included as much of the lesion as possible. For heterogeneous lesions, the largest possible ROI was placed on the portion of the lesion that subjectively had the highest attenuation. If a lesion contained calcification, this was avoided in the ROI analysis. The internal heterogeneity of each lesion was also subjectively determined by viewing the lesion with routine abdominal window and level settings and by using a four-point scale of 1 (homogeneous), 2 (mildly heterogeneous, characterized by slight mottling), 3 (moderately heterogeneous, characterized by small foci of high and low attenuation), and 4 (markedly heterogeneous, characterized by large foci of mixed attenuation).
Statistical Analysis
Statistical analysis involving standard parametric and nonparametric tests was performed using the software package Analyze-It version 1.73 (Analyze-It Software, Leeds, England), a statistical add-on for Microsoft Excel. Classification and regression tree analysis (CART) was performed by using CART software version 5.0 (Salford Systems, San Diego, Calif, 2006). The purpose of the statistical analysis was twofold: (a) to determine the agreement between the two readers and (b) to formulate a clinical guideline that would allow differentiation of a high-attenuation cyst from a renal cell carcinoma with a high degree of confidence if the appropriate criteria were satisfied.
For determining agreement, Bland-Altman regression analysis (4) was carried out for attenuation, the continuous variable, for both the high-attenuation cysts and tumors. The weighted 
statistic was used for internal heterogeneity, the categoric variable. The degree of observer agreement as indicated by values was interpreted as follows: 0–0.20 = slight agreement; 0.21–0.40 = fair agreement; 0.41–0.60 = moderate agreement; 0.61–0.80 = substantial agreement; and 0.81–1.00 = almost perfect agreement (5).
For formulation of a clinical guideline that could distinguish high-attenuation cysts from tumors, we pooled the data of the two readers in the following conservative worst-case manner. For cysts, we used the lower attenuation value and the more heterogeneous category assigned by either reader. For tumors, we used the higher attenuation value and more homogeneous category assigned by either reader. A visual analysis of the scatterplot, CART analysis, and the following tests were used to ascertain the normality of the tumor attenuation data: Shapiro-Wilk normality test (P < .05 is inconsistent with normality), and coefficients of skewness and kurtosis (P < .05 is inconsistent with normality).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
Carcinomas
For the 54 renal cell carcinomas, the average attenuation was 34.7 HU (range, 21–60 HU) for reader 1, who scored the internal heterogeneity as homogeneous (score = 1) in 35 of 54 (65%) renal cell carcinomas, as mildly heterogeneous (score = 1) in nine (17%), as moderately heterogeneous (score = 2) in four (7%), as markedly heterogeneous (score = 4) in six (11%).
For reader 2, the average attenuation of the 54 renal cell carcinomas was 38.4 (range, 22–60 HU). The internal heterogeneity was scored as homogeneous (score = 1) in 36 of 54 (67%) renal cell carcinomas, mildly heterogeneous (score = 2) in seven (13%), moderately heterogeneous (score = 3) in eight (15%), and markedly heterogeneous (score = 4) in three (5%).
Reader Agreement
There was substantial agreement between the two readers for heterogeneity of the high-attenuation cysts and tumors. For cysts, the readers assigned the same category in 54 of 56 cases; for the remaining two cases, the assigned categories differed by 1 ( = 0.66). For tumors, the readers agreed on the degree of heterogeneity in 37 of 54 cases and differed by just 1 category value in 15 of the remaining 17 cases ( = 0.62).
The Bland-Altman analysis showed good agreement between the readers on the attenuation of both the high-attenuation cysts and the neoplasm. The mean difference for attenuation values between readers for cysts was 0.6 HU for the 56 cysts (< 1 HU; 95% limits of agreement: –12.5 to +13.6 HU); for the 54 tumors, the mean difference was 3.7 HU (95% limits of agreement: –8.9 to +16.2 HU).
Overlap
The data ranges of the high-attenuation cysts and tumors indicated that there was overlap between cysts and tumors with respect to both attenuation and heterogeneity. However, no tumors had attenuation greater than 60 HU, and 16 cysts did. Cysts were also more homogeneous than tumors. Ninety-five percent (53 of 56) of high-attenuation cysts were classified as 1 (uniform) by both readers whereas 56% (30 of 54) of tumors were classified as such.
CART Analysis, Clinical Guideline
Since there was no overlap between the upper part of the range of attenuation between cysts and tumors, the first split suggested by the CART analysis was that if a lesion has an attenuation of 61 HU or greater (light dashed line, Fig 1), regardless of heterogeneity, it should be classified as a high-attenuation cyst. This rule is 100% effective in excluding tumors in our sample data. CART analysis did not yield any subsidiary rules to further distinguish high-attenuation cysts from tumors.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
Since the study by Zirinsky et al (6) was published, advances in US technology, including color and power Doppler techniques, have proved useful in evaluating renal masses because the vascularity of a mass can be assessed in some cases (7). More recently, compound spatial and tissue harmonic US imaging has been utilized, which can reduce background noise and eliminate unwanted low-level echoes in an otherwise simple renal cyst (8,9). With these technologic advances in US, high-attenuation renal cysts may be characterized as benign with greater frequency than in the past. On the other hand, if the mass measures less than 60 HU and is heterogeneous, our results suggest that there is a higher likelihood that the mass represents a renal cell carcinoma. In this scenario, CT (or MR) performed with and without IV contrast material may be the next appropriate examination, although it is recognized that some of these masses may also be characterized as benign with US.
The Hounsfield unit attenuation of high-attenuation renal cysts can be variable and is dependent on their content. The increased attenuation in these cysts may be secondary to hemorrhage or proteinaceous debris (1,6,10–12) and is also dependent on the state of hemorrhage within the cyst. The results of this study are concordant with those of studies that evaluated the CT attenuation of blood. New and Aronow (13) found that clotted blood measures 60–80 HU, while Wolverson et al (14) showed that fresh clot can measure up to 90 HU. However, one high-attenuation cyst in our series measured 113 HU (reader 1). The reason for this discrepancy is unclear, although the variability of attenuation readings may play a role (see below). It should be noted that the attenuation of a high-attenuation cyst is dependent on both the iron and the proteinaceous components of the blood within the cyst, and it has been demonstrated that the protein component accounts for the greater portion of the attenuation value (13). Therefore, the attenuation of the fluid within the cyst may also be dependent on the concentration of blood (hematocrit) or albumin in the patient.
The degree of heterogeneity also plays an important role in evaluating renal masses. In a study by Suh et al (2) in which 37 high-attenuation cysts were evaluated by two readers on portal venous phase contrast-enhanced CT images, all were graded as uniform (or mildly heterogeneous) in appearance by the first reader and 97% (36 of 37) were graded as uniform (or mildly heterogeneous) by the second reader. This is concordant with the results of our study, in which 98% (55 of 56) and 95% (53 of 56) (readers 1 and 2, respectively) of high-attenuation renal cysts were rated as homogeneous in attenuation. Although the study by Suh et al (2) evaluated contrast-enhanced images, this should not change the appearance of high-attenuation cysts since they do not enhance with contrast material and should thereby have an identical appearance before and after IV contrast material administration.
It must be emphasized that the use of absolute Hounsfield unit measurements for the characterization of specific tissue may be unjustified. Early in the history of CT, it was recognized that there was a wide range of Hounsfield unit attenuation values for a given tissue as a result of CT scanner performance alone (15). Variability of attenuation was found to be dependent on the type of scanner, position of an object within a scanner, change in peak voltage, and change in scan time. With the introduction of helical scanners, this unreliability was again recognized (16). Therefore, one must be cautious when using absolute Hounsfield units to make a specific diagnosis. Dunnick et al (17) reported a 2-cm homogeneous solid renal cell carcinoma that measured 72 HU on the unenhanced image. The results of our study suggest that this mass would likely represent a high-attenuation cyst. However, given the variability of Hounsfield unit readings, the true attenuation of the mass is not known and the diagnosis of a high-attenuation renal cyst cannot be made with absolute certainty. Therefore, these attributes should not substitute for definitive characterization of a renal mass by using US, CT, or MR imaging. The attenuation and degree of heterogeneity can be used to suggest what a lesion in question more likely represents and what would be the best follow-up test to characterize it.
There were limitations to this study. First, pathologic proof was not available for the high-attenuation renal cysts. However, they were characterized with standard methods used in radiologic practice today (demonstrating no enhancement by comparing unenhanced and contrast-enhanced CT images). It should be recognized that location within the kidney and size of the lesion also play a role in the characterization of high-attenuation renal cysts (18), and these factors were not evaluated in our study. Second, high-attenuation renal cysts were only compared with renal cell carcinomas. Angiomyolipomas, which do not contain macroscopic fat, may appear as homogeneous high-attenuation masses at unenhanced CT and may theoretically be confused with a high-attenuation cyst (19). However, in the study by Jinzaki et al (19), these types of angiomyolipomas measured only as high as 51 HU. Also, the masses in our study were evaluated by using different CT scanners and section thickness, which may produce variability in some of the measurements. Finally, a subjective determination of internal heterogeneity of each lesion was used, and interobserver variability in making this determination can exist.
In conclusion, high-attenuation renal cysts and renal cell carcinomas may have a similar appearance on unenhanced CT images. Our study demonstrated that a homogeneous renal mass with an attenuation of 70 HU or higher at unenhanced CT has a greater than 99.9% chance of representing a high-attenuation renal cyst. These findings may be helpful in suggesting the most appropriate next imaging study for definitive characterization.
|
ADVANCES IN KNOWLEDGE |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
|
FOOTNOTES |
|---|
Abbreviations: CART = classification and regression tree analysis • IV = intravenous • ROI = region of interest
Authors stated no financial relationship to disclose.
Author contributions: Guarantors of integrity of entire study, A.I.J., A.N.R., G.M.I.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; literature research, P.G.M., G.M.I.; clinical studies, A.I.J., A.N.R., G.M.I.; statistical analysis, P.G.M., G.M.I.; and manuscript editing, all authors
|
References |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONADVANCES IN KNOWLEDGEReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. G. Silverman, G. M. Israel, B. R. Herts, and J. P. Richie Management of the Incidental Renal Mass Radiology, October 1, 2008; 249(1): 16 - 31. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Siddiki, J. G. Fletcher, B. McFarland, N. Dajani, N. Orme, B. Koenig, M. Strobel, and S. M. Wolf Incidental Findings in CT Colonography: Literature Review and Survey of Current Research Practice. J. Law Med. Ethics, June 1, 2008; 36(2): 320 - 331. [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
) and tumors (
), with worst-case pooling for attenuation and heterogeneity (see text). The light dashed line (at 61 HU, the decision criterion recommended by visual inspection and CART analysis) effectively separates higher attenuation cysts from all tumors in this sample. Given the distribution of tumor attenuation values, a renal mass with attenuation of 70 HU (bold dashed line) or higher at unenhanced CT has a greater than 99.9% chance of representing a high-attenuation renal cyst.