| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Genitourinary Imaging |
1 From the Department of Radiology, New York University Medical Center, 560 First Ave, Suite HW 202, New York, NY 10016 (E.M.H., G.M.I., G.A.K., W.Y.H., D.C.K., V.S.L.); and Department of Environmental Medicine, Division of Biostatistics, New York University School of Medicine, New York, NY (I.B.L.). Received August 6, 2003; revision requested October 13; revision received October 29; accepted December 15. Address correspondence to G.M.I. (e-mail: gary.israel@med.nyu.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: MR imaging was performed by using a gadolinium-enhanced breath-hold fat-suppressed three-dimensional T1-weighted gradient-echo sequence in 71 patients (48 men and 23 women; mean age, 62 years; age range, 26–87 years) with 93 renal lesions for which pathologic correlation was available. For quantitative measurements of enhancement, the relative increase in signal intensity values was measured by one investigator with manually defined regions of interest, and the threshold of an increase of 15% or greater was used to distinguish malignant from benign masses. For qualitative assessment, two investigators independently reviewed the subtracted images of all lesions and subjectively determined whether enhancement was present or absent. The sensitivity, specificity, and positive and negative predictive values for each method were calculated and compared. Mean (± standard deviation) and median values of relative enhancement were also calculated for benign and malignant lesions.
RESULTS: At pathologic analysis, 74 (80%) of the 93 lesions were malignant, and 19 (20%)—including seven oncocytomas—were benign. For diagnosing malignancy based on enhancement alone, sensitivity and specificity, respectively, were 95% (70 of 74 lesions) and 53% (10 of 19 lesions) at quantitative analysis and 99% (73 of 74 lesions) and 58% (11 of 19 lesions) at qualitative analysis. All seven oncocytomas were considered to be malignant with both methods. When the oncocytomas were excluded, specificities increased to 83% (10 of 12 lesions) and 92% (11 of 12 lesions) for the quantitative and qualitative evaluations, respectively. Three of the four malignant lesions incorrectly characterized as benign at quantitative assessment were hyperintense on unenhanced MR images; all were diagnosed correctly at qualitative evaluation.
CONCLUSION: Image subtraction enables accurate assessment of renal tumor enhancement, particularly in the setting of masses that are hyperintense on unenhanced MR images.
© RSNA, 2004
Index terms: Images, processing • Kidney neoplasms, diagnosis, 81.31, 81.32 • Kidney neoplasms, MR, 81.12143 • Magnetic resonance (MR), contrast enhancement, 81.12143
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Recently, investigators have applied the practice of measuring relative enhancement of renal masses at CT to magnetic resonance (MR) imaging, with promising results (6). Image subtraction is widely used in MR imaging for improving the visibility of contrast enhancement in applications such as MR angiography and breast imaging (7), but, to our knowledge, its role in assessing renal masses has not yet been reported. Thus, the purpose of our study was to retrospectively compare quantitative and qualitative methods of assessing MR contrast enhancement as the basis for diagnosing renal malignancy.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
MR Imaging Protocol
MR imaging was performed at 1.5 T (Vision or Symphony; Siemens Medical Systems, Erlangen, Germany) by using a torso phased-array coil. All patients underwent coronal or transverse breath-hold fat-suppressed T1-weighted imaging with a 3D spoiled gradient-echo sequence, coronal breath-hold T2-weighted imaging with a half-Fourier rapid acquisition with relaxation enhancement sequence, and transverse in-phase and opposed-phase T1-weighted imaging with a two-dimensional gradient-echo sequence. In this study, all analyses of renal mass enhancement were performed by using images obtained with the 3D T1-weighted sequence.
Parameters for the 3D sequence were as follows: repetition time msec/echo time msec, 3.4–6.8/1.3–2.3; flip angle, 12°–35°; matrix, 80–256 x 256–512; field of view, 300–450 mm; and section thickness, interpolated to 1.3–3.1 mm. The sequence was repeated at least three times after 19 mL of gadopentetate dimeglumine (Magnevist; Berlex Laboratories, Wayne, NJ) was injected intravenously at a rate of 2 mL/sec with a power injector (Spectris; Medrad, Pittsburgh, Pa) and followed by a 20-mL saline flush.
Timing for the first contrast material–enhanced (arterial) phase was based on administration of a 1-mL test bolus of contrast material with standard methods (8). The second acquisition (venous phase) was performed approximately 90 seconds after the start of the contrast agent injection. The final phase was performed approximately 3–5 minutes after the administration of the contrast material. All sequences were performed during suspended respiration at end expiration to optimize image coregistration for subtraction algorithms. Patients were routinely offered supplemental oxygen through a nasal cannula to improve breath-holding capabilities, and breath-holding instructions were reviewed both during patient preparation and during the examination.
Quantitative Analysis and Calculations
One investigator (E.M.H.) with 1
years of experience in interpreting renal MR images—who was blinded to results of pathologic analysis and image subtraction—manually defined multiple circular ROIs on the unenhanced and contrast-enhanced 3D T1-weighted images at a workstation (Syngo; Siemens). ROIs were placed either in the center of the lesion or over any soft-tissue components that appeared to enhance based on the unenhanced and contrast-enhanced images. For all analyses, the ROIs were copied from the enhanced images onto the unenhanced images and were visually confirmed to be positioned similarly within the lesion on both kinds of images.
The percentage relative enhancement of each lesion was calculated as [(SIpost – SIpre)/SIpre] x 100%, where SIpost is the signal intensity on the contrast-enhanced image and SIpre is the signal intensity on the unenhanced image. For heterogeneous lesions, multiple ROIs were placed over different areas, and the region that showed maximum relative enhancement was used for analysis. The ROI size was similar for all measurements within a single patient (range, 0.9–3.9 cm2; mean, 0.88 cm2).
On the basis of previous reports, the criterion used to diagnose malignancy on the basis of relative enhancement was 15% or greater enhancement (6). Lesions with enhancement of less than 15% were considered to be benign. In the event enhancement was less than 0%, the lesion was considered to be benign.
The same investigator also noted all lesions that were predominantly hyperintense relative to the rest of the renal parenchyma on unenhanced images. The number of lesions that were hyperintense on unenhanced fat-suppressed T1-weighted MR images was determined. A lesion was considered hyperintense if 50% or more of it was subjectively higher in signal intensity than the renal parenchyma.
Qualitative Analysis
As part of our routine protocol, voxel-by-voxel subtraction of unenhanced images from contrast-enhanced images was performed on a satellite console (the same workstation used in the quantitative analysis). Two independent reviewers (W.Y.H., D.C.K.)—each of whom had 1 years of experience in interpreting renal MR images—who were blinded to results at pathologic and quantitative analysis, evaluated only the subtracted T1-weighted data sets on the commercially available workstation and visually assessed the presence or absence of enhancement.
Lesions that demonstrated any soft-tissue enhancement or enhancement of thickened walls or septa were considered to be malignant. Any disagreement between reviewers was resolved by consensus. Characteristics of lesions that required consensus assessment were also recorded. If the readers found that misregistration made the subtracted data sets uninterpretable, this was recorded and they were permitted to review the raw data and to perform individual section subtraction, if needed. The assessment of enhancement, however, was based solely on the subtracted data.
Comparison of Imaging Findings in Misdiagnosed Lesions
After all lesions were evaluated for enhancement quantitatively and qualitatively, two authors in consensus (E.M.H., with 1 years of experience, and G.M.I., with 5 years of experience) performed a side-by-side comparison of imaging findings for all masses given false-positive or false-negative diagnoses to identify possible causes of misdiagnoses.
Statistical Analysis
The pathologic diagnosis of benign versus malignant tumor was used as the reference for the assessment of the accuracy of quantitative and qualitative analyses of enhancement. For quantitative analysis, relative enhancement of 15% or greater was considered indicative of malignancy. For qualitative analysis, any visible signal intensity on subtracted images was considered to indicate malignancy. A separate analysis of accuracy was performed when oncocytomas were excluded from the category of benign lesions. The sensitivity and specificity of the two methods were compared by using the McNemar exact test. The positive and negative predictive values for each method were also calculated.
The mean (± standard deviation) and median values of relative enhancement for benign and malignant lesions were also calculated. The means were compared by using a two-sample t test and assuming unequal variance. P < .05 was considered to indicate a statistically significant difference.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
|
|
Qualitative Analysis
At review of subtracted images (ie, on the basis of qualitative assessment), 81 (87%) of 93 lesions showed visible enhancement and were considered to be malignant. Of these, 73 (90%) proved to be malignant at pathologic analysis and eight (10%)—including all seven oncocytomas—proved to be benign (Table 2). Of the 12 lesions that did not visibly enhance, 11 (92%) proved to be benign at pathologic analysis and one (8%) proved to be malignant (Table 2). There were no cases of misregistration that required the readers to view the raw data sets or to perform individual section subtraction.
|
|
Although the qualitative method appeared to be more sensitive and more specific for the detection of enhancement than the quantitative method, the differences between the methods were not statistically significant (P = .25 for sensitivity and P > .999 for specificity).
Comparison of Imaging Findings in Misdiagnosed Lesions
When oncocytomas were excluded, qualitative analysis of subtracted images was found to have resulted in incorrect characterization of one malignant and one benign lesion; both of these lesions were also incorrectly characterized at quantitative analysis. The malignant lesion was the previously described papillary renal cell carcinoma. The benign lesion was found to be a complex renal cyst with nodular wall thickening of up to 8 mm.
Quantitative analysis failed to reveal either the presence (n = 3) or the absence of enhancement (n = 1) in four additional lesions: three malignant lesions (sizes: 22, 45, and 69 mm) that were all hyperintense on unenhanced fat-suppressed T1-weighted MR images (Figs 3, 4) and one 8-mm benign lesion. These lesions were all correctly categorized by using subtracted MR images.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Ho et al (6) recently applied the strategy of using manually defined ROIs to measure relative enhancement in renal masses on the basis of the percentage increase in signal intensity following intravenous contrast agent administration. They determined that the optimal percentage threshold of enhancement for distinguishing cysts from solid renal lesions was 15% or greater (6). Applying this cutoff value, they found a sensitivity of 100% (50 of 50 lesions) and a specificity of 94% (47 of 50 lesions) for distinguishing cysts from solid lesions by using enhancement as the sole criterion.
We implemented similar methods of analysis, including a similar ROI size and shape, a similar formula for calculation of percentage enhancement, a similar threshold of enhancement, and a similar timing of the image acquisition after contrast agent injection that ranged from 1.5 to 5.0 minutes. However, in this study, we used a different MR magnet manufacturer, an alternative pulse sequence (interpolated fat-suppressed 3D T1-weighted spoiled gradient echo in the present study vs two-dimensional spoiled gradient echo without fat suppression in the study of Ho et al), thinner sections (interpolated to 2 mm in the present study vs 6–7 mm in the study of Ho et al), and a different rate of contrast agent injection (2 mL/sec in the present study vs 1 mL/sec in the study of Ho et al). Last, our study included lesions of all sizes, including three that measured less than 1 cm in diameter.
Despite these differences, we obtained a similarly high sensitivity (95% [70 of 74 lesions]). However, in our study, we found a lower specificity (53% [10 of 19 lesions]) for detection of enhancement in renal masses, due in part to the relatively large number of oncocytomas (n = 7). When oncocytomas were excluded, specificity improved to 83% (10 of 12 lesions).
A quantitative approach involves some practical considerations. First, it requires that the MR imaging system gain be held constant between unenhanced and enhanced acquisitions. The use of a fixed percentage of increased signal intensity as a cutoff involves the assumption that signal intensity will have an approximately linear relationship with the concentration of gadolinium-based contrast material. This assumption may not hold in certain circumstances, particularly for lesions that are hyperintense on unenhanced T1-weighted MR images.
Because ROIs are placed manually, assessment of enhancement with the quantitative approach may be subject to operator dependence and may be affected by spatial misregistration across acquisitions. In the case of lesions that are hyperintense on unenhanced T1-weighted MR images, placement of the ROI may be challenging because further increases in signal intensity on the contrast-enhanced images may be difficult to detect visually. In fact, in our series, three of the four malignant lesions incorrectly characterized as benign with signal intensity measurements were hyperintense on unenhanced T1-weighted MR images.
Last, lesion size can play an important role in the accuracy of quantitative measurements. Partial volume artifacts are more problematic with smaller lesions and can falsely elevate signal intensity measurements (9). In this study, quantitative analysis falsely indicated enhancement in an 8-mm simple cyst, despite the use of thin sections (interpolated section thickness, 2 mm). For small simple cysts, we routinely obtain T2-weighted MR images to aid in lesion characterization; however, we did not analyze these images as part of this study.
Image subtraction is a technique that has been widely used for image postprocessing in MR. In particular, it has been used to facilitate the detection of enhancing masses in breast MR imaging and is routinely used for postprocessing in contrast-enhanced MR angiography. To our knowledge, its efficacy for revealing enhancement in renal masses has not been previously reported.
There are several advantages to the subtraction approach for assessing renal mass enhancement. It enables a global assessment of renal enhancement that is not dependent on user-positioned ROIs. Also, the subtraction approach may be particularly well suited for the assessment of lesions that are hyperintense on unenhanced images. Although we did not specifically evaluate this in the present study, in our experience, image subtraction is also faster and easier than quantitative assessment for evaluating renal mass enhancement.
Like the quantitative approach, the image subtraction approach also involves the assumption that system gain between acquisitions will be fixed. It is also susceptible to motion and other misregistration artifacts, typically owing to the variability of patient breath holding. This variability can be minimized by stressing the importance of breath holding to the patients before MR imaging and by coaching them during imaging. In addition, end-expiratory breath holding has been shown to be more reproducible compared with end-inspiratory breath holding (10,11). In cases of obvious misregistration between images, it is possible to perform subtraction on an image-by-image basis, selecting those images that are matched in position.
Last, image subtraction does require a subjective determination by the radiologist. This does not pose a problem when a mass is clearly enhancing. However, in some neoplasms, enhancement can be very subtle, and there may be interobserver variability regarding the presence or absence of enhancement within a lesion. In our series, interreader variability was small—consensus readings were required for only two of 93 lesions. When interpretations are equivocal regarding the presence or absence of enhancement, a direct comparison of the lesion in question with a coexisting simple renal cyst is often helpful.
There were recognized limitations of this study, including its retrospective nature and a case-selection bias for pathologically proved lesions, which are far more likely to be malignant. This study did not allow for long-term follow-up of lesions to prove their benignity, and, therefore, statistical results may be different in a series for which long-term follow-up was available.
For all analyses, interpretation of whether a lesion was benign or malignant was based on enhancement only, although in clinical practice other features such as T2 signal intensity are routinely used. Because we aimed to compare two methods of assessing enhancement, the complete set of sequences for each lesion was not used in this study. The delay times between contrast agent injection and imaging were not standardized for all patients. Ho et al (6) showed that the optimal timing for measurements was between 2 and 4 minutes after gadolinium chelate injection, and for all of our patients, measurements were performed on images obtained within 5 minutes after injection. Our MR imaging protocol varied during the course of the study, as technical advances resulted in improved spatial resolution. However, in the comparison of qualitative and quantitative assessments of enhancement, both approaches involved use of the same MR imaging sequences in each patient, and, therefore, errors attributable to spatial resolution would be expected to apply equally to both methods.
In conclusion, because the presence or absence of enhancement within a renal mass is the most important factor in its proper characterization, a standardized approach for determining enhancement should be defined. According to our findings, both quantitative and qualitative methods are sensitive in the detection of enhancement within a renal mass. However, in lesions that are hyperintense on unenhanced T1-weighted MR images, qualitative assessment based on image subtraction should be performed to avoid false-negative quantitative results.
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, E.M.H., G.M.I., I.B.L., V.S.L.; study concepts, G.A.K., E.M.H., G.M.I., V.S.L.; study design, E.M.H., G.M.I., G.A.K., W.Y.H., D.C.K., V.S.L.; literature research, E.M.H., G.M.I.; clinical studies, E.M.H., W.Y.H., D.C.K.; data acquisition, E.M.H., W.Y.H., D.C.K., I.B.L.; data analysis/interpretation, all authors; statistical analysis, E.M.H., I.B.L., V.S.L.; manuscript preparation, definition of intellectual content, and final version approval, all authors; manuscript editing, E.M.H., G.M.I., G.A.K., V.S.L.; manuscript revision/review, E.M.H., G.M.I., I.B.L., V.S.L.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. R. Oliva, J. N. Glickman, K. H. Zou, S. Y. Teo, K. J. Mortele, M. S. Rocha, and S. G. Silverman Renal Cell Carcinoma: T1 and T2 Signal Intensity Characteristics of Papillary and Clear Cell Types Correlated with Pathology Am. J. Roentgenol., June 1, 2009; 192(6): 1524 - 1530. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Kim, M. Jain, A. B. Harris, V. S. Lee, J. S. Babb, E. E. Sigmund, L. E. Rueff, and B. Taouli T1 Hyperintense Renal Lesions: Characterization with Diffusion-weighted MR Imaging versus Contrast-enhanced MR Imaging Radiology, June 1, 2009; 251(3): 796 - 807. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. Taouli, R. K. Thakur, L. Mannelli, J. S. Babb, S. Kim, E. M. Hecht, V. S. Lee, and G. M. Israel Renal Lesions: Characterization with Diffusion-weighted Imaging versus Contrast-enhanced MR Imaging Radiology, May 1, 2009; 251(2): 398 - 407. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. R. M. Sun, L. Ngo, E. M. Genega, M. B. Atkins, M. E. Finn, N. M. Rofsky, and I. Pedrosa Renal Cell Carcinoma: Dynamic Contrast-enhanced MR Imaging for Differentiation of Tumor Subtypes--Correlation with Pathologic Findings Radiology, March 1, 2009; 250(3): 793 - 802. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. M. Israel and M. A. Bosniak Pitfalls in Renal Mass Evaluation and How to Avoid Them RadioGraphics, September 1, 2008; 28(5): 1325 - 1338. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Zhang, Y. Mazaheri Tehrani, L. Wang, N. M. Ishill, L. H. Schwartz, and H. Hricak Renal Masses: Characterization with Diffusion-weighted MR Imaging--A Preliminary Experience Radiology, May 1, 2008; 247(2): 458 - 464. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Newatia, G. Khatri, B. Friedman, and J. Hines Subtraction Imaging: Applications for Nonvascular Abdominal MRI Am. J. Roentgenol., April 1, 2007; 188(4): 1018 - 1025. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. M. Israel and M. A. Bosniak How I Do It: Evaluating Renal Masses Radiology, August 1, 2005; 236(2): 441 - 450. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
