HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) All Versions of this Article: 2441060634v1 244/1/94    most recent Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Williams, T. C. Articles by Lehman, C. D. Search for Related Content PubMed PubMed Citation Articles by Williams, T. C. Articles by Lehman, C. D.

Breast MR Imaging: Computer-aided Evaluation Program for Discriminating Benign from Malignant Lesions¹

¹ From the Department of Radiology, University of Washington Medical Center, Seattle Cancer Care Alliance, 825 Eastlake Ave E, Room G3-200, Seattle, WA 98109-1023. From the 2005 RSNA Annual Meeting. Received April 10, 2006; revision requested June 6; revision received August 8; accepted September 7; final version accepted November 8. Address correspondence to C.D.L. (e-mail: lehman{at}u.washington.edu).

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCE IN KNOWLEDGE References

 
Purpose: To retrospectively determine the sensitivity of kinetic features measured with computer-aided evaluation at breast magnetic resonance (MR) imaging in discriminating benign from malignant lesions, with histopathologic findings used as the reference standard.

Materials and Methods: Institutional review board approval was obtained for this HIPAA-compliant study. Informed consent was waived. Suspicious breast lesions visible only at MR imaging and in which biopsy had been performed with MR imaging guidance were retrospectively evaluated with a computer-aided evaluation program. Computer-generated kinetic features for each lesion were recorded, and those of benign and malignant lesions were compared. Features analyzed included the presence or absence of computer-aided evaluation "threshold enhancement" at 50% and 100% minimum thresholds; degree of initial peak enhancement; and enhancement profiles composed of lesion percentages of washout, plateau, and persistent enhancement. The Fisher exact test and Student t test were used to assess differences in these analyses.

Results: One hundred fifty-four consecutive lesions (41 malignant, 113 benign) in 125 women (age range, 27–86 years; mean age, 52 years) were evaluated. The presence of threshold enhancement at computer-aided evaluation was sensitive for malignancy, with 38 of 41 (93%) malignant lesions demonstrating enhancement at both the 50% and 100% thresholds. Absence of threshold enhancement at computer-aided evaluation helped improve the discrimination between benign and malignant lesions when compared with that at initial interpretation by the radiologists. False-positive rates were reduced by 8.8% at the 50% enhancement threshold (not significant) and by 23.0% at the 100% enhancement threshold (P = .02) when compared with that at initial interpretation. Analyses of initial peak enhancement values and enhancement profiles did not demonstrate further improvements in lesion discrimination.

Conclusion: The use of computer-aided evaluation for breast MR imaging significantly helped improve the discrimination of benign from malignant lesions when compared with that at initial interpretations by radiologists.

© RSNA, 2007

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCE IN KNOWLEDGE References

 
Current applications for breast magnetic resonance (MR) imaging include evaluation of ipsilateral extent of disease and screening of the contralateral breast in patients with newly diagnosed breast cancer (1–7), screening of women at high risk for breast cancer (8–14), evaluation of patients with metastatic axillary adenopathy and an unknown primary cancer (15–17), and evaluation of breast implants for rupture (18).

A particular challenge for interpretation of breast MR images is the assessment of a lesion's morphologic and kinetic features at multiple imaging series. To address some of the impediments to performing breast MR imaging, computer-aided evaluation programs for MR imaging of the breast have been developed (19,20). Such programs are now used by many practice sites to automatically perform image processing and analysis functions typically performed manually by the technologist and the radiologist. One key analysis function performed with computer-aided evaluation is automatic kinetic assessment. The detailed information on lesion kinetics provided with computer-aided evaluation differs substantially from that obtained with conventional manual placement of a region of interest. While manual placement of a region of interest provides kinetic information only for portions of a given lesion, computer-aided evaluation generates detailed data for all pixels in the lesion.

There are sparse data regarding the effect of kinetic features assessed with computer-aided evaluation on the diagnostic accuracy of breast MR imaging. Prior small studies of computer-aided evaluation of breast MR images have focused on computer-assessed morphologic features and enhancement kinetics that are best used to separate benign from malignant lesions; however, these studies have used a single MR imaging technique and institution-specific computer analysis systems (20,21). Results of one pilot study (22) of a commercially available computer-aided evaluation system found that kinetics assessed with computer-aided evaluation improved specificity in discriminating benign from malignant lesions. This study was limited by the small number of lesions, as well as by the inclusion of only invasive malignancies and the use of only two methods for acquiring MR images. Thus, the purpose of our study was to retrospectively determine the sensitivity of kinetic features measured with computer-aided evaluation at breast MR imaging in discriminating benign from malignant lesions, with histopathologic findings used as the reference standard.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCE IN KNOWLEDGE References

 
Patients and Lesions
Our study included all consecutive suspicious breast lesions depicted at MR imaging that were evaluated at our institution between January 2001 and September 2004 and led to a biopsy performed with MR imaging guidance. The lesions had been initially identified, evaluated, and recommended for biopsy by one of five radiologists (C.D.L., W.B.D.) at the time of original interpretation of MR images. All lesions were nonpalpable and were not visible at mammography or targeted "second-look" ultrasonography (US). The study lesions were identified at MR imaging performed for the following clinical indications: evaluation of extent of disease in patients with a new cancer diagnosis (64%), screening in high-risk patients (14%), follow-up of a probably benign lesion detected at MR imaging (3%), problem solving (14%), and other (5%). After MR imaging–guided biopsy, the original MR imaging examinations for all lesions were retrospectively analyzed with computer-aided evaluation to determine kinetic features.

Institutional review board approval was obtained for this Health Insurance Portability and Accountability Act–compliant study. Patient informed consent for this retrospective study was waived by the institutional review board.

Protocols for MR Image Acquisition
All study examinations were performed with an LX 1.5-T unit (GE Medical Systems, Milwaukee, Wis) by using a dedicated breast coil (MRI Devices, Waukesha, Wis). Gadolinium-based contrast material was used; 20 mL was hand-injected at 1 mL/sec. Three separate protocols were used. Each imaging protocol included a bilateral transverse localizer sequence and one precontrast and at least two postcontrast sagittal T1-weighted fat-suppressed three-dimensional fast gradient-spoiled acquisitions. Imaging parameters common to all protocols included a field of view of 18 to 22 cm depending on patient size, section thickness 3 mm or less, and a matrix of 256 x 192. The number of section partitions in each acquisition varied slightly with breast size, with at least 60 sagittal sections typically acquired for full coverage of a single breast. Transverse and coronal reformats, as well as maximum intensity projections, were constructed from the first postcontrast series.

From January 2001 through March 2002, a four-channel phased-array receive-only breast coil was used (4 Channel Breast Array; MRI Devices) with a unilateral acquisition protocol. The precontrast and postcontrast T1-weighted three-dimensional fast gradient-spoiled images were acquired with repetition time of 6.7 msec, echo time of 4.2 msec, and flip angle of 35°. Postcontrast images were acquired, with the first acquisition centered at 2.5 minutes after injection and the delayed acquisition centered at 7.5 minutes after injection.

From April 2002 through January 2004, a four-channel phased-array receive-only breast coil was used (4 Channel Breast Array; MRI Devices) with a sequential bilateral acquisition protocol, achieved by alternating the imaging sequences from one breast to the other. As a result, separate imaging volumes were prescribed for each breast; shimming was performed only for the breast of primary interest. T1-weighted three-dimensional fast gradient-spoiled imaging parameters were as follows: repetition time msec/echo time msec, 17/2.3; flip angle, 35°. Postcontrast acquisitions were centered at 1 minute and 5 minutes for the first breast and at 3 minutes and 7 minutes for the second breast.

From February 2004 through September 2004, a seven-channel phased-array receive-only breast coil was used (Excite 7 Channel Biopsy Breast Array; MRI Devices) with a volume imaging for breast assessment parallel imaging technique (GE Medical Systems). Sagittal imaging of both breasts was performed simultaneously, with independent shimming for each breast. A single imaging volume was placed to encompass both breasts, and in most cases 200 sagittal images were obtained per sequence. T1-weighted three-dimensional fast gradient-spoiled imaging parameters were as follows: 6.7/4.2; flip angle, 10°. Postcontrast acquisitions were centered at 90 seconds, 180 seconds, and 270 seconds for each breast; the early and first delayed postcontrast examinations were completed within approximately 2 and 4 minutes, respectively.

Initial Interpretation
Each MR image was initially interpreted by a single radiologist specializing in breast imaging. Five fellowship-trained breast imagers (including W.B.D. and C.D.L.) with 2 to 8 years of breast MR imaging experience performed interpretations during the study period. MR imaging interpretation criteria, assessments, and recommendations were based on the American College of Radiology Breast Imaging and Reporting Data System (BI-RADS) Magnetic Resonance Imaging lexicon (23). All lesions were assessed as suspicious (BI-RADS category 4) or highly suggestive of malignancy (BI-RADS category 5). Targeted second-look US was recommended and performed for all lesions. Biopsy was performed on the lesions visible at US by using ultrasound guidance and were not included in this study. MR imaging–guided biopsies performed before May 2003 were performed by using a 14-gauge spring-loaded core biopsy needle (C.R. Bard, Covington, Ga). MR imaging–guided biopsies conducted after May 2003 were performed by using a 9-gauge vacuum-assisted device (Suros, Indianapolis, Ind).

Retrospective Interpretation of MR Images Obtained with Computer-aided Evaluation
All MR images were retrospectively processed by using a computer-aided evaluation system (CADstream, version 3.0; Confirma, Kirkland, Wash). This software is designed to automate image processing and analysis functions currently performed manually by the MR imaging technologist and radiologist. All processed data are saved and presented with the original images for radiologist interpretation.

The software incorporates three MR imaging series into its calculations: one precontrast T1-weighted series, one immediate postcontrast T1-weighted series, and one delayed postcontrast T1-weighted series. The program compares pixel signal intensity values on the precontrast and immediate postcontrast series. If a pixel value increases above a user-specified minimum enhancement threshold, such as a 50% or a 100% increase in enhancement, the pixel is said to meet threshold enhancement. Once a pixel has been identified as enhancing above the established threshold, the program then compares pixel signal intensity values on the immediate and delayed postcontrast series. If a pixel value on the delayed series decreases by more than 10% compared with the immediate postcontrast series, it is color-coded red, indicating a washout pattern of enhancement. If a pixel value increases by more than 10%, it is color-coded blue, indicating persistent enhancement. If a pixel value does not change in either direction by more than 10%, it is color-coded green, for plateau enhancement. One of three colors (blue, green, or red) is applied to that pixel based on the delayed enhancement pattern. This results in a color overlay map that is displayed on each MR image, indicating regions of threshold enhancement. Areas of threshold enhancement determined with the computer-aided evaluation software algorithm to be "connected" are summed and constitute a lesion. A synopsis of the kinetic enhancement details of the total connected area or lesion is automatically generated.

The kinetic data include an initial peak enhancement value, calculated as the highest percent enhancement in the pixel with the most suspicious curve type (washout > plateau > persistent) on the immediate postcontrast series. The kinetic data also include an enhancement profile, composed of the percentages of the total lesion volume that demonstrates washout, plateau, and persistent enhancement.

In the version of computer-aided evaluation software used in this study, enhancement characteristics are automatically summarized for only the 30 largest independent lesions depicted with computer-aided evaluation. For a lesion lacking automatically generated enhancement characteristics, an enhancement curve can be generated by manual placement of a cursor over a pixel in the lesion. This curve demonstrates the initial and delayed enhancement kinetics for the pixel.

Data Collection
All data were collected and recorded by a radiology resident (T.C.W.). The resident, who had undergone training in the computer-aided evaluation system by the vendor and through 1 month of clinical experience, was blinded to lesion biopsy results and reviewed all of the MR imaging examinations processed with computer-aided evaluation on a computer monitor. The presence or absence of threshold enhancement with computer-aided evaluation at 50% and 100% minimum enhancement levels was recorded for each lesion. Threshold enhancement was said to be present if any portion of a lesion achieved the specified minimum threshold, either 50% or 100%, as indicated by the presence of computer-aided evaluation–generated color within the lesion. Initial data analysis showed three false-negative lesions on the basis of lack of threshold enhancement at automatic computer-aided evaluation. Then, a curve was generated for each of the three lesions by using manual placement of a cursor on the lesion on the computer-aided evaluation images and the extent of initial enhancement was recorded. For lesions with enhancement meeting the 100% minimum threshold, the resident identified the lesions for which the extent of enhancement assessed with computer-aided evaluation agreed well with the visual extent of the lesion (enhancement at computer-aided evaluation not visually greater or less than the size of the lesion). For these lesions, the initial peak enhancement and the percentage values for washout, plateau, and persistent enhancement were recorded. All data entered were reviewed by two research staff assistants and re-evaluated by using the computer-aided evaluation system for a randomly selected 10% of lesions (T.C.W.).

Characteristics of the lesions, including BI-RADS lesion type (focus, mass, non-masslike enhancement), size, MR imaging acquisition method, and histopathologic results, were obtained from a database that we maintain for our clinical breast MR imaging program.

Statistical Analysis
Relative false-positive fractions (change in the percentage of false-positive findings), sensitivities, and positive predictive values for the initial interpretations by the radiologists and for computer-aided evaluation alone at both the 50% and the 100% enhancement thresholds were computed and compared by two of the authors (S.P. and S.C.P.) (24). Fisher exact tests were performed, and P values were calculated to compare the likelihood of malignancy based on presence of threshold enhancement at the 50% and the 100% enhancement thresholds. Fisher exact tests and calculated P values were also used to compare computer-aided evaluation threshold enhancement of ductal carcinoma in situ (DCIS) and invasive malignancy. Kinetic enhancement profiles and initial peak enhancement of malignant versus benign lesions were compared by using a Student t test. Descriptive statistics of lesion characteristics, including frequencies, means, range, and standard deviations, were also computed. All computations were conducted by using statistical software (SAS version 8; SAS Institute, Cary, NC). All graphs and charts were produced by using software (Excel 2002; Microsoft, Redmond, Wash).

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCE IN KNOWLEDGE References

 
Patients and Lesions
One hundred fifty-five lesions in 126 female patients (age range, 27–86 years; mean age, 52 years) were identified. One lesion was excluded because the MR imaging examination could not be processed with computer-aided evaluation as a result of a technical issue on a postcontrast series. One hundred fifty-four lesions in 125 patients make up the analysis set (Fig 1).

View larger version (17K): [in this window] [in a new window] [Download PPT slide]   Figure 1: Flow chart of study profile according to the Standards of Reporting of Diagnostic Accuracy recommendations. CAE = computer-aided evaluation.

Additional benign lesions included focal fibrosis, fibromatosis, adenosis, sclerosing adenosis, lymph node, benign cyst, fibroadenoma, epithelial hyperplasia without atypia, papilloma, apocrine metaplasia, radial scar, inflammation, and pseudoangiomatous stroma hyperplasia.

The following histologic types were identified for the 41 malignant lesions: DCIS (n = 18), invasive ductal carcinoma (n = 15), invasive lobular carcinoma (n = 6), and two lesions of carcinoma not otherwise specified.

Threshold Enhancement at Computer-aided Evaluation
Thirty-eight of 41 malignant lesions showed threshold enhancement at the computer-aided evaluation minimum of 50%. All 38 malignant lesions also demonstrated threshold enhancement at the computer-aided evaluation minimum of 100%. The sensitivity of computer-aided evaluation on the basis of the presence of threshold enhancement was 93% for both the 50% and the 100% minimum levels (Fig 2). Three of 41 (7%) malignant lesions did not demonstrate automated computer-aided evaluation threshold enhancement at either enhancement threshold and were false-negative lesions. Further evaluation with manual placement of a cursor over each false-negative lesion demonstrated computer-aided evaluation kinetic curves with enhancement above 100% in two of the three lesions. The possible explanations for lack of automated assessment of these lesions are detailed in the Discussion section. The false-negative lesions were two foci and one area of non-masslike enhancement, and the histopathologic types were DCIS in two lesions (measuring 3 mm and 24 mm, respectively) and invasive ductal carcinoma in one lesion (measuring 5 mm) (Fig 3).

View larger version (79K): [in this window] [in a new window] [Download PPT slide]   Figure 2a: MR images of left breast in a 56-year-old woman with a newly diagnosed carcinoma in the left breast at the 6 o'clock position. (a) Sagittal fat-suppressed T1-weighted three-dimensional fast gradient-spoiled (17/2.3) image obtained immediately after administration of gadolinium-based contrast material depicts an additional enhancing irregular mass at the 3 o'clock position (arrowhead). (b) On same image with computer-aided evaluation applied, lesion meets the minimum enhancement threshold of 100%, indicated by presence of color overlay at the site. Computer-aided evaluation–generated enhancement synopsis in lower left demonstrates lesion enhancement profile (long arrow), with 100% persistent enhancement. Computer-aided evaluation–generated initial peak enhancement is also displayed (short arrow). Lesion also met the enhancement threshold of 50% (not shown). MR imaging–guided biopsy results depicted infiltrating ductal carcinoma, confirming multicentric malignancy.

View larger version (81K): [in this window] [in a new window] [Download PPT slide]   Figure 2b: MR images of left breast in a 56-year-old woman with a newly diagnosed carcinoma in the left breast at the 6 o'clock position. (a) Sagittal fat-suppressed T1-weighted three-dimensional fast gradient-spoiled (17/2.3) image obtained immediately after administration of gadolinium-based contrast material depicts an additional enhancing irregular mass at the 3 o'clock position (arrowhead). (b) On same image with computer-aided evaluation applied, lesion meets the minimum enhancement threshold of 100%, indicated by presence of color overlay at the site. Computer-aided evaluation–generated enhancement synopsis in lower left demonstrates lesion enhancement profile (long arrow), with 100% persistent enhancement. Computer-aided evaluation–generated initial peak enhancement is also displayed (short arrow). Lesion also met the enhancement threshold of 50% (not shown). MR imaging–guided biopsy results depicted infiltrating ductal carcinoma, confirming multicentric malignancy.

View larger version (133K): [in this window] [in a new window] [Download PPT slide]   Figure 3a: MR images of the left breast in a 30-year-old woman with newly diagnosed carcinoma in the left breast at the five o'clock position. (a) Sagittal fat-suppressed T1-weighted three-dimensional fast gradient-spoiled (6.7/4.2) image obtained immediately after administration of gadolinium-based contrast material shows additional non-masslike ductal enhancement at the 7 o'clock position (arrowheads). (b) On same image with computer-aided evaluation applied, lesion does not meet the minimum enhancement threshold of 100%, indicated by lack of color overlay at the site. Lesion also did not meet the enhancement threshold of 50% (not shown). MR imaging–guided biopsy results depicted DCIS, confirming multifocal malignancy.

View larger version (68K): [in this window] [in a new window] [Download PPT slide]   Figure 3b: MR images of the left breast in a 30-year-old woman with newly diagnosed carcinoma in the left breast at the five o'clock position. (a) Sagittal fat-suppressed T1-weighted three-dimensional fast gradient-spoiled (6.7/4.2) image obtained immediately after administration of gadolinium-based contrast material shows additional non-masslike ductal enhancement at the 7 o'clock position (arrowheads). (b) On same image with computer-aided evaluation applied, lesion does not meet the minimum enhancement threshold of 100%, indicated by lack of color overlay at the site. Lesion also did not meet the enhancement threshold of 50% (not shown). MR imaging–guided biopsy results depicted DCIS, confirming multifocal malignancy.

View larger version (15K): [in this window] [in a new window] [Download PPT slide]   Figure 4: Bar graph compares true-positive, false-positive, and false-negative values for initial radiologist interpretation versus computer-aided evaluation at 50% and 100% enhancement thresholds. Application of computer-aided evaluation at a 100% threshold reduces the number of false-positive lesions. There is no change in sensitivity of computer-aided evaluation at 50% threshold compared with that at 100% threshold. Three malignant lesions identified at initial radiologist's interpretation were false-negative findings at application of computer-aided evaluation. CAE = computer-aided evaluation; FN = false-negative finding; FP = false-positive finding; TP = true-positive finding.

View larger version (66K): [in this window] [in a new window] [Download PPT slide]   Figure 5a: MR images of left breast in a 51-year-old asymptomatic woman with a history of prior right breast carcinoma. (a) Sagittal fat-suppressed T1-weighted three-dimensional fast gradient-spoiled (6.7/4.2) image obtained immediately after administration of gadolinium-based contrast material shows non-masslike segmental enhancement in upper inner quadrant (arrow). (b) On same image with computer-aided evaluation applied, lesion does not meet minimum enhancement threshold of 100%, indicated by lack of color overlay at the site. Lesion also did meet the enhancement threshold of 50% (not shown). MR imaging–guided biopsy results depicted benign fibrosis and ductal hyperplasia.

View larger version (41K): [in this window] [in a new window] [Download PPT slide]   Figure 5b: MR images of left breast in a 51-year-old asymptomatic woman with a history of prior right breast carcinoma. (a) Sagittal fat-suppressed T1-weighted three-dimensional fast gradient-spoiled (6.7/4.2) image obtained immediately after administration of gadolinium-based contrast material shows non-masslike segmental enhancement in upper inner quadrant (arrow). (b) On same image with computer-aided evaluation applied, lesion does not meet minimum enhancement threshold of 100%, indicated by lack of color overlay at the site. Lesion also did meet the enhancement threshold of 50% (not shown). MR imaging–guided biopsy results depicted benign fibrosis and ductal hyperplasia.

False-positive cases were evenly distributed across the three acquisition protocols. The three false-negative cases occurred with use of the bilateral simultaneous acquisition protocol. The small number of false-negative cases in this study precluded analysis of the effect of acquisition protocol on the rate of false-negative findings.

Enhancement Profiles and Initial Peak Enhancement with Computer-aided Evaluation
Enhancement profiles and peak initial enhancement were analyzed for cases with threshold enhancement confined to the study lesion at the 100% threshold. This yielded 79 lesions (50 benign and 29 malignant).

There were no significant differences between the enhancement profiles of benign and malignant lesions, with lesions demonstrating a variety of washout, persistent, and plateau patterns of enhancement (P = .9 for washout pattern, P = .5 for persistent pattern, and P = .2 for plateau pattern) (Fig 6). Thirty-four percent of malignant lesions demonstrated no washout according to computer-aided evaluation assessment. Sixty-four percent of benign lesions demonstrated washout with computer-aided evaluation assessment. Of 11 lesions with exclusively persistent enhancement, eight were benign and three were malignant.

View larger version (8K): [in this window] [in a new window] [Download PPT slide]   Figure 6: Scatterplot illustrates delayed enhancement profiles of benign and malignant lesions. Wide overlap of the enhancement profiles generated by computer-aided evaluation among the benign and malignant lesions shows lack of significant differences in the persistent, plateau, and washout profiles between these two groups.

View larger version (8K): [in this window] [in a new window] [Download PPT slide]   Figure 7: Scatterplot illustrates peak initial enhancements for benign and malignant lesions, according to MR imaging acquisition protocol. Benign and malignant lesions demonstrate a similarly wide distribution of peak enhancement percentages.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCE IN KNOWLEDGE References

Results of previous studies (20,21,25–32) have demonstrated the utility of measuring contrast enhancement rates to discriminate between different types of lesions. In general, malignant lesions have been found to exhibit higher levels of initial enhancement than benign lesions at 1–2 minutes after injection of contrast material. The optimal diagnostic enhancement threshold or cutoff has varied between studies and probably depends on such factors as technique for administration of contrast material and acquisition timing. Gribbestad et al (25) found an enhancement threshold of 70% successfully discriminated benign fibroadenomata from malignant tumors at 1 minute after injection. Kaiser and Zeitler (33) reported that malignant lesions enhanced approximately 100% within 2 minutes after injection while benign breast abnormalities demonstrated much lower enhancements. Similarly, Stomper et al (26) maximized their diagnostic accuracy by using an enhancement threshold of 100% at 2 minutes after injection. Our results—a statistically significant difference in the presence of enhancement above 100% between malignant and benign lesions—are in keeping with the findings of prior studies.

In our study, a subset of lesions exhibiting enhancement greater than 100% were further characterized by their levels of initial peak enhancement and delayed enhancement profiles. We found that the initial peak enhancement values did not vary significantly between benign and malignant lesions. Others have reported the utility of measuring peak enhancement for discriminating lesions (25,27,29,31,33). However, our analysis differed from the prior investigations in that we evaluated peak enhancement in only those lesions that enhanced above 100%. Our results coincide with those of Stomper et al (26), who also found no significant differences in contrast enhancement amplitude, rate, and washout for benign and malignant lesions that enhanced by more than 100%, and with other study results that have revealed overlap in peak enhancement values of benign and malignant lesions (28). Our results for initial peak enhancement were similar for the three different MR imaging protocols included in our study. Some of the prior studies used shorter acquisition times (25,27,29), and it is possible that measurement of peak enhancement at an earlier time point may help improve discrimination between lesion types.

The delayed enhancement profiles (washout, plateau, or persistent) also did not significantly differ between benign and malignant lesions, akin to findings published by Stomper et al (26). However, our results differ from the findings of multiple prior studies that showed a significant correlation between malignancy and washout kinetics (20,21,27,30). The disparity in results may be because of the higher temporal resolution of MR imaging acquisition protocols used in the prior studies, with postcontrast sequences acquired as frequently as every 15 seconds. Application of computer-aided evaluation to higher-temporal-frequency MR imaging examinations may reveal differences in computer-aided evaluation enhancement profiles between benign and malignant lesions. Differences between our results and those of prior studies are also probably due to the fact that computer-aided evaluation provides enhancement information for all pixels in a lesion rather than for a portion of a lesion as measured by using manual region-of-interest placement. The computer-aided evaluation method is similar to the lesion analysis systems described by Gilhuijs et al (21) and Chen et al (20) and more fully and accurately demonstrate lesion kinetics. However, comparison of our computer-aided evaluation enhancement profile results to kinetic curves obtained by region-of-interest placement is challenging because of the differences in the measurement techniques.

While the presence of threshold enhancement in malignant lesions was highly sensitive, there were three malignant lesions that were false-negative findings at computer-aided evaluation and did not demonstrate enhancement at the 100% or the 50% thresholds. Computer-aided evaluation enhancement curves produced manually showed enhancement greater than 100% in two of the three lesions. These cases illustrate several technical limitations to the computer-aided evaluation program because threshold enhancement was not automatically detected in some lesions despite their levels of enhancement above the specified minimum. The software includes a noise-filtering process that can lead to failure of automatic analysis of small areas of enhancement. Also, the computer-aided evaluation version that we studied automatically generates kinetic information for only the largest 30 lesions, some of which may be nonbreast structures, such as the heart and chest wall. A more recent version of this computer-aided evaluation program has corrected this restriction. Nevertheless, these technical limitations underscore the importance of using computer-aided evaluation as a complement but not a replacement to the radiologist's assessment.

Our study had several limitations. Although three breast MR imaging acquisition protocols were analyzed, there are multiple other methods of acquisition in clinical use. Our study also included only lesions determined by a radiologist to be suspicious. Thus, our results may not apply to all enhancing lesions at MR imaging.

In summary, our findings suggest that computer-aided evaluation has the potential to improve the discrimination of benign from malignant breast lesions at MR imaging. We believe that computer-aided evaluation is useful as a tool to supplement the radiologist's subjective interpretation, but it should not be exclusively relied on to guide management. Specifically, given the presence of three false-negative lesions, a finding deemed suspicious by the radiologist should be further evaluated regardless of the enhancement features determined by using computer-aided evaluation.

	ADVANCE IN KNOWLEDGE

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCE IN KNOWLEDGE References

 

• The use of computer-aided evaluation significantly improved (P = .02) the discrimination between benign and malignant lesions when compared with that of the initial interpretations by the radiologists.
  

	FOOTNOTES

 

Abbreviations: BI-RADS = Breast Imaging and Reporting Data System • DCIS = ductal carcinoma in situ

Authors stated no financial relationship to disclose.

Author contributions: Guarantor of integrity of entire study, C.D.L.; 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, T.C.W., S.C.P.; clinical studies, W.B.D.; statistical analysis, S.C.P., S.P.; and manuscript editing, all authors

	References

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION ADVANCE IN KNOWLEDGE References

 

1. Lee SG, Orel SG, Woo IJ, et al. MR imaging screening of the contralateral breast in patients with newly diagnosed breast cancer: preliminary results. Radiology 2003;226:773–778.[Abstract/Free Full Text]
2. Sardanelli F, Giuseppetti GM, Panizza P, et al. Sensitivity of MRI versus mammography for detecting foci of multifocal, multicentric breast cancer in fatty and dense breasts using the whole-breast pathologic examination as a gold standard. AJR Am J Roentgenol 2004;183:1149–1157.[Abstract/Free Full Text]
3. Fischer U, Kopka L, Brinck U, Korabiowska M, Schauer A, Grabbe E. Prognostic value of contrast-enhanced MR mammography in patients with breast cancer. Eur Radiol 1997;7:1002–1005.[CrossRef][Medline]
4. Harms SE, Flamig DP. MR imaging of the breast. J Magn Reson Imaging 1993;3:277–283.[Medline]
5. Liberman L, Morris EA, Kim CM, et al. MR imaging findings in the contralateral breast of women with recently diagnosed breast cancer. AJR Am J Roentgenol 2003;180:333–341.[Abstract/Free Full Text]
6. Fischer U, Kopka L, Grabbe E. Breast carcinoma: effect of preoperative contrast-enhanced MR imaging on the therapeutic approach. Radiology 1999;213:881–888.[Abstract/Free Full Text]
7. Lehman CD, Blume JD, Thickman D, et al. Added cancer yield of MRI in screening the contralateral breast of women recently diagnosed with breast cancer: results from the International Breast Magnetic Resonance Consortium (IBMC) trial. J Surg Oncol 2005;92:9–15.[CrossRef][Medline]
8. Kriege M, Brekelmans CT, Boetes C, et al. Efficacy of MRI and mammography for breast-cancer screening in women with a familial or genetic predisposition. N Engl J Med 2004;351:427–437.[Abstract/Free Full Text]
9. Warner E, Plewes DB, Hill KA, et al. Surveillance of BRCA1 and BRCA2 mutation carriers with magnetic resonance imaging, ultrasound, mammography, and clinical breast examination. JAMA 2004;292:1317–1325.[Abstract/Free Full Text]
10. Morris EA. Screening for breast cancer with MRI. Semin Ultrasound CT MR 2003;24:45–54.[CrossRef][Medline]
11. Burke W, Daly M, Garber J, et al. Recommendations for follow-up care of individuals with an inherited predisposition to cancer. II. BRCA1 and BRCA2. Cancer Genetics Studies Consortium. JAMA 1997;277:997–1003.
12. Claus EB, Risch N, Thompson WD. Autosomal dominant inheritance of early-onset breast cancer: implications for risk prediction. Cancer 1994;73:643–651.[CrossRef][Medline]
13. Easton DF, Ford D, Bishop DT. Breast and ovarian cancer incidence in BRCA1-mutation carriers. Breast Cancer Linkage Consortium. Am J Hum Genet 1995;56:265–271.
14. Lehman CD, Blume JD, Weatherall P, et al. Screening women at high risk for breast cancer with mammography and magnetic resonance imaging. Cancer 2005;103:1898–1905.[CrossRef][Medline]
15. Orel SG, Weinstein SP, Schnall MD, et al. Breast MR imaging in patients with axillary node metastases and unknown primary malignancy. Radiology 1999;212:543–549.[Abstract/Free Full Text]
16. Morris EA, Schwartz LH, Dershaw DD, van Zee KJ, Abramson AF, Liberman L. MR imaging of the breast in patients with occult primary breast carcinoma. Radiology 1997;205:437–440.[Abstract/Free Full Text]
17. Brenner RJ, Rothman BJ. Detection of primary breast cancer in women with known adenocarcinoma metastatic to the axilla: use of MRI after negative clinical and mammographic examination. J Magn Reson Imaging 1997;7:1153–1158.[Medline]
18. Cher DJ, Conwell JA, Mandel JS. MRI for detecting silicone breast implant rupture: meta-analysis and implications. Ann Plast Surg 2001;47:367–380.[CrossRef][Medline]
19. Wood C. Computer aided detection (CAD) for breast MRI. Technol Cancer Res Treat 2005;4:49–53.[Medline]
20. Chen W, Giger ML, Lan L, Bick U. Computerized interpretation of breast MRI: investigation of enhancement-variance dynamics. Med Phys 2004;31:1076–1082.[CrossRef][Medline]
21. Gilhuijs KG, Deurloo EE, Muller SH, Peterse JL, Schultze Kool LJ. Breast MR imaging in women at increased lifetime risk of breast cancer: clinical system for computerized assessment of breast lesions initial results. Radiology 2002;225:907–916.[Abstract/Free Full Text]
22. Lehman CD, Peacock S, DeMartini WB, Chen X. A new automated software system to evaluate breast MR examinations: improved specificity without decreased sensitivity. AJR Am J Roentgenol 2006;187:51–56.[Abstract/Free Full Text]
23. American College of Radiology. ACR BI-RADS: Magnetic Resonance Imaging. In: ACR Breast Imaging Reporting and Data System, Breast Imaging Atlas. Reston, Va: American College of Radiology, 2003.
24. Pepe M. The statistical evaluation of medical tests for classification and prediction. New York, NY: Oxford University Press, 2003.
25. Gribbestad IS, Nilsen G, Fjosne H, et al. Contrast-enhanced magnetic resonance imaging of the breast. Acta Oncol 1992;31:833–842.[Medline]
26. Stomper PC, Herman S, Klippenstein DL, et al. Suspect breast lesions: findings at dynamic gadolinium-enhanced MR imaging correlated with mammographic and pathologic features. Radiology 1995;197:387–395.[Abstract/Free Full Text]
27. Kuhl CK, Mielcareck P, Klaschik S, et al. Dynamic breast MR imaging: are signal intensity time course data useful for differential diagnosis of enhancing lesions? Radiology 1999;211:101–110.
28. Knopp MV, Weiss E, Sinn HP, et al. Pathophysiologic basis of contrast enhancement in breast tumors. J Magn Reson Imaging 1999;10:260–266.[CrossRef][Medline]
29. Szabo BK, Aspelin P, Wiberg MK, Bone B. Dynamic MR imaging of the breast: analysis of kinetic and morphologic diagnostic criteria. Acta Radiol 2003;44:379–386.[CrossRef][Medline]
30. Bluemke DA, Gatsonis CA, Chen MH, et al. Magnetic resonance imaging of the breast prior to biopsy. JAMA 2004;292:2735–2742.[Abstract/Free Full Text]
31. Mussurakis S, Gibbs P, Horsman A. Primary breast abnormalities: selective pixel sampling on dynamic gadolinium-enhanced MR images. Radiology 1998;206:465–473.[Abstract/Free Full Text]
32. Kinkel K, Helbich TH, Esserman LJ, et al. Dynamic high-spatial-resolution MR imaging of suspicious breast lesions: diagnostic criteria and interobserver variability. AJR Am J Roentgenol 2000;175:35–43.[Abstract/Free Full Text]
33. Kaiser WA, Zeitler E. MR imaging of the breast: fast imaging sequences with and without Gd-DTPA—preliminary observations. Radiology 1989;170:681–686.[Abstract/Free Full Text]

This Article Abstract Figures Only Full Text (PDF) All Versions of this Article: 2441060634v1 244/1/94    most recent Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Williams, T. C. Articles by Lehman, C. D. Search for Related Content PubMed PubMed Citation Articles by Williams, T. C. Articles by Lehman, C. D.