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) 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 HighWire Citing Articles via Google Scholar Google Scholar Articles by Meisamy, S. Articles by Garwood, M. Search for Related Content PubMed PubMed Citation Articles by Meisamy, S. Articles by Garwood, M.

Adding in Vivo Quantitative ¹H MR Spectroscopy to Improve Diagnostic Accuracy of Breast MR Imaging: Preliminary Results of Observer Performance Study at 4.0 T¹

¹ From the Department of Radiology, Center for Magnetic Resonance Research Medical School, 2021 Sixth St SE, Minneapolis, MN 55455 (S.M., P.J.B., E.H.B., M.G.P., M.G.). The complete list of author affiliations appears at the end of this article. From the 2004 RSNA Annual Meeting. Received May 7, 2004; revision requested Jul 8; revision received Sep 15; accepted Oct 12. Supported by NIH grants RR08079, CA92004, and RR00400; Tickle Family Land Grant Endowment in Breast Cancer Research; PHS Cancer Center Support grant P30 CA77398; and Lillian Quist-Joyce Henline Chair in Biomedical Research. Address correspondence to M.G. (e-mail: gar{at}cmrr.umn.edu).

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION References

 
PURPOSE: To determine whether the addition of in vivo quantitative hydrogen 1 (¹H) magnetic resonance (MR) spectroscopy can improve the radiologist's diagnostic accuracy in interpreting breast MR images to distinguish benign from malignant lesions.

MATERIALS AND METHODS: The study was approved by the institutional review board and, where appropriate, was compliant with the Health Insurance Portability and Accountability Act. All patients provided written informed consent. Fifty-five breast MR imaging cases—one lesion each in 55 patients aged 24–66 years with biopsy-confirmed findings—were retrospectively evaluated by four radiologists. Patients were examined with contrast material–enhanced fat-suppressed T1-weighted 4.0-T MR imaging. The concentration of total choline–containing compounds (tCho) was quantified by using single-voxel ¹H MR spectroscopy. For each case, the radiologists were asked to give the percentage probability of malignancy, the Breast Imaging and Reporting Data System category, and a recommendation for patient treatment. Two interpretations were performed for each case: The initial interpretation was based on the lesion's morphologic features and time–signal intensity curve, and the second interpretation was based on the lesion's morphologic features, time–signal intensity curve, and tCho concentration. Receiver operating characteristic (ROC), Wilcoxon signed rank, statistic, and accuracy (based on the area under the ROC curve) analyses were performed.

RESULTS: Of the 55 lesions evaluated, 35 were invasive carcinomas and 20 were benign. The addition of ¹H MR spectroscopy resulted in higher sensitivity, specificity, accuracy, and interobserver agreement for all four radiologists. More specifically, two of the four radiologists achieved a significant improvement in sensitivity (P = .03, P = .03), and all four radiologists achieved a significant improvement in accuracy (P = .01, P = .05, P = .009, P < .001).

CONCLUSION: Current study results suggest that the addition of quantitative ¹H MR spectroscopy to the breast MR imaging examination may help to improve the radiologist's ability to distinguish benign from malignant breast lesions.

© RSNA, 2005

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION References

 
According to recent data published in the Journal of the National Cancer Institute (1), in the United States alone there is an estimated lifetime incidence of breast cancer of one in seven women. Nearly 40 000 women were predicted to die of breast cancer in 2004. Conventional modalities such as physical examination, mammography, and ultrasonography (US) are limited in both sensitivity for the detection of disease and specificity for the differentiation between benign and malignant breast lesions. Magnetic resonance (MR) imaging is increasingly being used to examine patients suspected of having breast cancer. More specifically, breast MR imaging has greatly influenced the surgical staging of breast cancer by enabling the identification of additional tumor foci either within the same quadrant (multifocal) or in different quadrants (multicentric) of the ipsilateral or contralateral breast (2–4). MR imaging of the breast is reported to have high sensitivity (94%–100%), variable specificity (37%–97%) (5–7), and poor positive predictive value, even in patients at high risk for breast cancer (8). As a result, definitive management of lesions detected at MR imaging can be a difficult task.

Currently, there are three primary approaches that radiologists use to interpret breast MR imaging cases: (a) determination of the presence or absence of lesion enhancement with gadolinium, (b) high-spatial-resolution imaging analysis of the morphologic features of the breast lesion (9,10), and (c) assessment of the evolution of lesion enhancement, which involves the evaluation of time–signal intensity curves (11–14). However, it has been shown that not all malignant lesions have a specific time–signal intensity curve pattern, and, thus, the evaluation of contrast enhancement dynamics alone can lead to false-negative findings (13,15,16). Furthermore, experiences with MR imaging and mammography have revealed that a small percentage of breast malignancies have benign morphologic features and that areas of trauma or postsurgical change in the breast can have morphologic features that are indistinguishable from those of invasive malignancies.

Hydrogen 1 (¹H) MR spectroscopy is increasingly being studied as a potential adjunct to breast MR imaging. Results of numerous ex vivo (17–21) and in vivo (22–31) studies have shown that neoplastic breast tissue contains elevated levels of total choline–containing compounds (tCho), which have methyl protons that resonate at a chemical shift of 3.2 ppm.

One group of researchers found that the use of ¹H MR spectroscopy, when implemented in an imaging protocol, may improve the specificity of breast MR imaging (32). However, performing in vivo ¹H MR spectroscopy of the breast can be technically challenging because the sensitivity is often limited and spectral artifacts can arise. Owing to the heterogeneous distribution of fat and glandular tissue in the breast, the ¹H spectra of the breast often have large lipid signals that can give rise to contaminant peaks around 3.2 ppm. However, these artifactual peaks, which are also known as lipid sidebands, can be suppressed by using a technique called echo-time averaging (33). Furthermore, performance testing in previous single-voxel ¹H MR spectroscopic studies at 1.5 T were based on the assumption that tCho is detectable only in malignant breast tissue (28). With the increased sensitivity achieved at high field (4.0 T), it has been shown that tCho is also detectable in benign lesions and in normal fibroglandular breast tissue (30). Thus, ¹H MR spectroscopy at high field strength may require implementation of a method to quantify spectra (30).

Although numerous studies have revealed that neoplastic breast tissue contains elevated levels of tCho, to our knowledge there are currently no published reports addressing the clinical efficacy of ¹H MR spectroscopy on the radiologist's accuracy in interpreting breast MR imaging cases. We hypothesized that the addition of quantitative single-voxel ¹H MR spectroscopy to the breast MR imaging examination would improve the radiologist's ability to distinguish benign from malignant breast lesions. Thus, the purpose of our study was to determine whether the addition of in vivo quantitative ¹H MR spectroscopy can improve the radiologist's diagnostic accuracy in interpreting breast MR images to distinguish benign from malignant lesions.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION References

 
Reference Standard and Imaging Case Preparation
Women aged 18–80 years who had suspicious physical examination, mammographic, and/or US findings in the breast and were scheduled for biopsy were eligible to enroll in our ongoing study. A total of 94 patients aged 24–66 years (mean age, 47 years) were enrolled in our study from February 2002 to February 2004. The study was approved by the institutional review board of the University of Minnesota School of Medicine and, where appropriate, was compliant with the Health Insurance Portability and Accountability Act. Patients were referred to the study by medical oncologists, surgeons, or radiologists at the University of Minnesota or in private practice. All patients provided written informed consent before enrolling in the study. One of the authors (S.M.) collected the following information from each patient: age, menopausal status, use or nonuse of hormone replacement therapy, use or nonuse of oral contraceptives, type of breast biopsy performed, histologic diagnosis, reference-standard examination used, lesion size, and tCho concentration.

In preparation for the observer performance study, a receiver operating characteristic (ROC) curve was generated from a set of spectra to provide the readers with an optimal cutoff point for the tCho concentration. The ROC curve was generated from spectra acquired from 68 (of the 94) patients whose tCho content included a lipid-to-water ratio of less than 0.33, as required with use of the tCho quantification method (30). Forty-one of these 68 patients had invasive breast carcinoma, and 27 had benign breast lesions. In this data set, the tCho concentration ranged from 0 to 8.5 mmol/kg (mean, 2.2 mmol/kg) for the malignant lesions and from 0 to 1.1 mmol/kg (mean, 0.21 mmol/kg) for the benign lesions. With equal weighting of sensitivity and specificity, the optimal cutoff point for the tCho concentration was 1.03 mmol/kg: A tCho concentration of 1.03 mmol/kg or greater suggested malignancy, whereas a concentration of less than 1.03 mmol/kg suggested benignity. The sensitivity and specificity at a tCho concentration cutoff point of 1.03 mmol/kg were 61% and 83%, respectively.

Fifty-five of the 94 patients met the following criteria to be examined in the observer performance study: (a) The patient had not undergone any chemotherapy or breast radiation therapy, and (b) ¹H MR spectroscopy revealed a lipid-to-water ratio of less than 0.33, as required with use of the tCho quantification method (30). One lesion in each of the 55 patients was individually evaluated at biopsy and at MR imaging by three authors (M.T.N., L.I.E., T.H.E.), with 3–11 years (mean, 7 years) of experience interpreting breast MR imaging studies, who did not participate in the observer performance study.

MR Imaging and MR Spectroscopy
All MR imaging measurements were performed by using a 4.0-T research unit that consisted of a 90-cm magnet bore (model 4T-900; Oxford Magnet Technology, Oxfordshire, England) with a clinical gradient system (Sonata; Siemens, Erlangen, Germany) interfaced with an imaging spectrometer (Unity Inova; Varian, Palo Alto, Calif). Several different single-breast quadrature transmit-receive radiofrequency surface coils of similar design were used to accommodate breasts of different sizes (34). The coils were mounted onto a custom-built patient table designed for prone unilateral breast examinations. The patient was placed in the prone position, with the breast centered horizontally in the magnet. After scout images were acquired to verify proper positioning, the coil was manually tuned and matched. A high-spatial-resolution three-dimensional (3D) fat-suppressed fast low-angle shot MR image (13.5/4.1 [repetition time msec/echo time msec], 256 x 256 x 64 matrix, 14–18-cm field of view, 30° flip angle, 0.6–0.7-mm in-plane resolution, 2.2–2.8-mm section thickness) and a two-dimensional fat-suppressed multisection fast low-angle shot image (390/5.1, 256 x 128 matrix, 30 sections, 2.5-mm section thickness, 14–18-cm field of view, 90° flip angle) were acquired before the injection of 0.1 mmol of gadopentetate dimeglumine (Magnevist; Berlex Laboratories, Wayne, NJ) per kilogram of body weight. In each patient, five sets of two-dimensional MR images were acquired immediately after injection. Then, a second high-spatial-resolution 3D fast low-angle shot MR image was acquired.

With the patient still in the magnet, the MR images were analyzed by using our own image-processing software, which was written in Matlab, version 6.1 (The Mathworks, Natick, Mass), language, so that the ¹H MR spectroscopic voxels could be selected. Each examination took 50–70 minutes. The total time required to collect the spectrum for each voxel was approximately 9 minutes, which included approximately 2 minutes for shimming, power calibrations, and the acquisition of a water spectrum, followed by approximately 6 minutes of averaging. Corrections for receiver gain, T1 and T2 relaxation values, and the numbers of protons per molecule were included in the quantification calculations (30).

One of the authors (S.M.) measured the lesion size according to the longest dimension of the enhancing lesion on the x-, y-, and z-axes on the high-spatial-resolution 3D subtraction MR image. Lesion size was recorded in cubic centimeters. Voxel placement was performed either jointly by two authors (among S.M., P.J.B., E.H.B., and M.G.) or by one of these authors (S.M. or E.H.B.) after that individual had acquired at least 6 months of experience in breast MR imaging. The MR spectroscopic voxel was positioned in the location of the breast that was expected to correspond to the eventual biopsy site, according to assessment of the lesion architecture, the pattern of dynamic contrast enhancement on the time–signal intensity curve, and the clinical information previously obtained from mammographic or US images that showed the biopsy site. The selected voxels were intended to maximize coverage of the enhancing lesion with minimal inclusion of adipose tissue.

Single-voxel ¹H MR spectroscopy was performed with a technique known as localization by adiabatic selective refocusing, or LASER (35). MR spectroscopic data were acquired by using 4096 complex points and a spectral width of 6 kHz. Each voxel measurement began with a calibration of the localized radiofrequency magnetic induction field (B₁) strength, followed by 30–60 seconds of manual adjustment of the linear shims. A fully relaxed, single-shot unsuppressed spectrum was acquired to measure the water and lipid signals. To suppress the water signal, the radiofrequency power was then manually adjusted by using the variable pulse power and optimized relaxation delays, or VAPOR, technique (36). The metabolite spectrum was acquired by using echo-time averaging with an echo time of 45–196 msec in 64 or 128 increments and a repetition time of 3 seconds (33). Each free-induction decay signal was individually saved: No averaging was performed until processing. We quantified tCho levels by fitting a Voigt line-shape model to the data and using the unsuppressed water signal as an internal reference (30).

Observer Performance Study
Four radiologists participated in the observer performance study: Two readers (K.R.B., K.K.A.) were from academic institutions, and two (R.A.C., B.A.L.) were from private practices. Each reader had 4–6 years of experience (mean, 5 years) interpreting breast MR images. Individual reading sessions were held for each reader, and all readings were performed during one session. The readers were blinded to all clinical data, including mammographic and US findings and medical histories. At least 1 week before the observer performance study, each reader was provided with a package of material that included the following: (a) a list of the breast MR imaging lexicons used in the American College of Radiology Breast Imaging and Reporting Data System (BI-RADS) to describe time–signal intensity curves, assess lesions, and recommend patient treatment (37); (b) the recommended optimal cutoff point for a tCho concentration of 1.03 mmol/kg with a sensitivity of 61% and a specificity of 83%; (c) a copy of the article describing the method of breast MR spectroscopic quantification used in this study (30); and (d) a description of the steps involved in the observer performance study.

The lexicons outlined by the American College of Radiology to describe the enhancement patterns depicted on time–signal intensity curves were as follows: slow, medium, or fast during the initial imaging phase and persistent, plateau, or washout during the delayed phase. The BI-RADS lexicon used to categorize lesions were as follows: Category 1 consisted of lesions with no abnormal enhancement; category 2, benign findings; category 3, lesions that were probably benign; category 4, lesions suspicious for malignancy; and category 5, lesions highly suspicious for malignancy. The lexicons outlined by the American College of Radiology for recommending patient treatment were as follows: no further work-up needed, follow-up MR imaging at 6 months to 1 year, follow-up MR imaging at less than 6 months, or biopsy.

Before the start of the study, each reader was given the opportunity to have any questions answered regarding the recommended cutoff point for the tCho concentration, the article describing the MR spectroscopic quantification method (30), and/or the steps involved in the observer performance study. Once the study was initiated, no discussion was held between the readers and the examiner (S.M.) who proctored the observer performance study.

Two separate interpretations of each breast MR imaging case were performed. For the initial interpretation, the reader was provided with only the 3D high-spatial-resolution MR images and the time–signal intensity curve with corresponding two-dimensional MR images. The 3D MR images used to evaluate the morphologic features of the lesion were displayed on a 21-inch liquid crystal display monitor. Sagittal views of the precontrast, postcontrast, and subtraction images from four contiguous sections centered on the lesion of interest were displayed. The lesion of interest appeared on at least two of the four postcontrast images and/or on two of the four subtraction images. For each case, an arrow indicated the location of the lesion in question. A sample display of the 3D MR images is shown in Figure 1a.

View larger version (147K): [in this window] [in a new window] [Download PPT slide]   Figure 1a. (a) Sample display of sagittal high-spatial-resolution 3D fat-suppressed fast low-angle shot breast MR images (13.5/4.1). Each row corresponds to a single section of the breast. Images acquired before (left column) and 7 minutes after (middle column) gadopentetate dimeglumine injection and with subtraction (right column) are shown. Arrow points to the lesion to be evaluated by each reader. (b) Time–signal intensity (SI) curve (top) for the lesion depicted in a and corresponding two-dimensional MR images (bottom). The signal intensity for each case was normalized to 1. Each two-dimensional image corresponds to one signal intensity data point on the time–signal intensity curve.

View larger version (47K): [in this window] [in a new window] [Download PPT slide]   Figure 1b. (a) Sample display of sagittal high-spatial-resolution 3D fat-suppressed fast low-angle shot breast MR images (13.5/4.1). Each row corresponds to a single section of the breast. Images acquired before (left column) and 7 minutes after (middle column) gadopentetate dimeglumine injection and with subtraction (right column) are shown. Arrow points to the lesion to be evaluated by each reader. (b) Time–signal intensity (SI) curve (top) for the lesion depicted in a and corresponding two-dimensional MR images (bottom). The signal intensity for each case was normalized to 1. Each two-dimensional image corresponds to one signal intensity data point on the time–signal intensity curve.

Once the initial interpretation was complete, the reader performed a second interpretation. For the second interpretation, the reader was shown the tCho concentration measured in the lesion and asked to reevaluate the case. After reevaluating the lesion's morphologic features, time–signal intensity curve, and tCho concentration, each reader was asked to do the following for the second interpretation: (a) give an assessment of the lesion in terms of BI-RADS category and percentage probability of malignancy (0%–100%) and (b) provide a recommendation for patient treatment. After completing the second interpretation, the reader began an initial interpretation of the next case.

Statistical Analyses
Statistical analyses were performed with standard statistical methods by using a computer software package (SAS for Windows, version 8.02; SAS Institute, Cary, NC). For each interpretation, there were four possible result scenarios: (a) a true-positive case, defined as a biopsy-proved diagnosis of cancer and the reader's recommendation of biopsy; (b) a false-positive, or "benign biopsy," case, defined as a biopsy-proved diagnosis of a benign lesion and the reader's recommendation of biopsy; (c) a false-negative, or "missed cancer," case, defined as a biopsy-proved diagnosis of cancer and the reader's recommendation not to perform biopsy; and (d) a true-negative case, defined as a biopsy-proved diagnosis of a benign lesion and the reader's recommendation not to perform biopsy. The sensitivity and specificity of each reader's interpretation, with corresponding standard errors and variance, were calculated (38). The weighted mean values of sensitivity and specificity, which were derived by using the inverse of the variance as weight, were calculated across all four readers (38).

ROC curves were generated from the tCho concentrations measured in the 55 patients whose MR imaging cases were used for the observer performance study and from the assessments based on BI-RADS category and percentage probability of malignancy assigned by each reader. Accuracy, as represented by the area under the ROC curve, and the cutoff point for the tCho concentration were calculated and recorded (39–41). The change in each reader's accuracy, based on the lesion's BI-RADS category and percentage probability of malignancy, between the initial and second interpretations was evaluated for statistical significance by using the Wilcoxon signed rank test (42–44). The change in each reader's sensitivity and specificity between the initial and second interpretations was evaluated for statistical significance by using the exact test for matched binary data (38).

Results regarding the recommendations for patient treatment were analyzed by using statistics for paired reader agreement (45). Statistic results were categorized as follows: A value greater than or equal to 0 but less than or equal to 0.20 indicated slight agreement; a value greater than or equal to 0.21 but less than or equal to 0.40, fair agreement; a value greater than or equal to 0.41 but less than or equal to 0.60, moderate agreement; a value greater than or equal to 0.61 but less than or equal to 0.80, substantial agreement; and a value greater than or equal to 0.81 but less than or equal to 1.00, excellent agreement (46).

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION References

 
Patient and Lesion Characteristics
Thirty-two of the 55 patients were premenopausal, and 13 of these 32 patients had a history of taking oral contraceptives. Twenty-three patients were postmenopausal, and six of these 23 patients were receiving hormone replacement therapy. The histologic diagnosis was made by using US-guided core-needle biopsy in 48, excisional biopsy in five, and mammographic stereotactically guided biopsy in two patients. For the observer performance study, a single lesion corresponding to the biopsy site was identified in each patient. All 55 lesions that were evaluated in this study demonstrated enhancement after gadopentetate dimeglumine administration.

Of the 55 breast lesions evaluated, 20 were benign and 35 were malignant. Of the 20 benign lesions, seven were fibroadenomas, four were proliferative fibrocystic changes, four were nonproliferative fibrocystic changes, three were normal benign breast tissue, and two were cases of atypical ductal hyperplasia. Of the 35 malignant lesions, 28 were invasive ductal carcinomas, and 12 of these 28 lesions also had a ductal carcinoma in situ component. Two lesions were pure ductal carcinomas in situ. Three lesions were invasive lobular carcinomas, and one of these three lesions also had a lobular carcinoma in situ component. Two lesions were composed of invasive ductal carcinoma and invasive lobular carcinoma components. The mean size of the benign lesions was 1.10 cm³ (range, 0.43–3.20 cm³), and the mean size of the malignant lesions was 6.30 cm³ (range, 0.39–28.00 cm³).

tCho Concentration Analysis in Observer Performance Study
An ROC curve was generated from the tCho concentrations measured in each of the 55 lesions in the observer performance study (Fig 2). The mean tCho concentration was 0.39 mmol/kg (range, 0–1.40 mmol/kg) for the benign lesions and 1.90 mmol/kg (range, 0–8.50 mmol/kg) for the malignant lesions. With equal weighting of sensitivity and specificity, the optimal cutoff point for the tCho concentration was 1.05 mmol/kg: A tCho concentration of 1.05 mmol/kg or greater suggested malignancy, whereas a tCho concentration of less than 1.05 mmol/kg suggested benignity. The accuracy of tCho concentration alone for lesion differentiation with a cutoff point of 1.05 mmol/kg was 83%, with a sensitivity of 69% and a specificity of 90%.

View larger version (22K): [in this window] [in a new window] [Download PPT slide]   Figure 2. ROC curve generated from the tCho concentration measured in the 55 lesions evaluated in the observer performance study. With a tCho concentration cutoff point of 1.05 mmol/kg, the accuracy—expressed as the area under the ROC curve (Az)—was 83%; the sensitivity, 69%; and the specificity, 90%.

During the initial interpretations, a total of 39 (mean, 10 per reader) benign cases were considered to be suspicious for malignancy ("benign lesions" in Table 1) by the readers. All of these cases were described in association with a time–signal intensity curve showing medium or fast enhancement during the initial phase followed by an enhancement plateau during the delayed phase. After the addition of ¹H MR spectroscopy for the second interpretation, the readers correctly classified four (10%) of the 39 benign cases as benign. Of these four cases, two were fibroadenomas and two were proliferative fibrocystic changes. The mean tCho concentration in these four cases was 0.24 mmol/kg (range, 0–0.86 mmol/kg).

Reader Accuracy and Interobserver Agreement
The addition of ¹H MR spectroscopy for the second breast MR imaging case interpretations resulted in all four readers having higher accuracy in assigning lesions to BI-RADS categories (Table 2). The mean accuracy of the BI-RADS category assignments was 66% (range, 62%–71%) in the initial interpretations and 75% (range, 67%–82%) in the second interpretations. Three of the four readers had significantly higher accuracy in assigning BI-RADS categories in the second interpretation than in the initial interpretation (P = .02, P = .1, P = .003, P = .002; Wilcoxon signed rank test). Figure 3 shows the ROC curves generated from the BI-RADS categories assigned to the lesions by each reader.

View larger version (26K): [in this window] [in a new window] [Download PPT slide]   Figure 3a. (a–d) ROC curves for the four readers based on the BI-RADS categories assigned to the lesions by each reader. On each graph, the dashed line represents the ROC curve generated from the initial interpretations, which were based on the morphologic features and time–signal intensity curves of the lesions, and the solid line represents the ROC curve generated from the second interpretations, which were based on the morphologic features, time–signal intensity curves, and tCho concentrations of the lesions. Note that the accuracy (expressed as area under ROC curve) of the interpretations made by also assessing tCho measurements (solid lines) is higher than the accuracy of the interpretations made without assessing tCho measurements (dashed lines).

View larger version (25K): [in this window] [in a new window] [Download PPT slide]   Figure 3b. (a–d) ROC curves for the four readers based on the BI-RADS categories assigned to the lesions by each reader. On each graph, the dashed line represents the ROC curve generated from the initial interpretations, which were based on the morphologic features and time–signal intensity curves of the lesions, and the solid line represents the ROC curve generated from the second interpretations, which were based on the morphologic features, time–signal intensity curves, and tCho concentrations of the lesions. Note that the accuracy (expressed as area under ROC curve) of the interpretations made by also assessing tCho measurements (solid lines) is higher than the accuracy of the interpretations made without assessing tCho measurements (dashed lines).

View larger version (26K): [in this window] [in a new window] [Download PPT slide]   Figure 3c. (a–d) ROC curves for the four readers based on the BI-RADS categories assigned to the lesions by each reader. On each graph, the dashed line represents the ROC curve generated from the initial interpretations, which were based on the morphologic features and time–signal intensity curves of the lesions, and the solid line represents the ROC curve generated from the second interpretations, which were based on the morphologic features, time–signal intensity curves, and tCho concentrations of the lesions. Note that the accuracy (expressed as area under ROC curve) of the interpretations made by also assessing tCho measurements (solid lines) is higher than the accuracy of the interpretations made without assessing tCho measurements (dashed lines).

View larger version (27K): [in this window] [in a new window] [Download PPT slide]   Figure 3d. (a–d) ROC curves for the four readers based on the BI-RADS categories assigned to the lesions by each reader. On each graph, the dashed line represents the ROC curve generated from the initial interpretations, which were based on the morphologic features and time–signal intensity curves of the lesions, and the solid line represents the ROC curve generated from the second interpretations, which were based on the morphologic features, time–signal intensity curves, and tCho concentrations of the lesions. Note that the accuracy (expressed as area under ROC curve) of the interpretations made by also assessing tCho measurements (solid lines) is higher than the accuracy of the interpretations made without assessing tCho measurements (dashed lines).

View larger version (26K): [in this window] [in a new window] [Download PPT slide]   Figure 4a. (a–d) ROC curves for the four readers based on the percentage probabilities of malignancy (0%–100%) assigned to the lesions by each reader. On each graph, the dashed line represents the ROC curve generated from the initial interpretations, which were based on the morphologic features and time–signal intensity curves of the lesions, and the solid line represents the ROC curve generated from the second interpretations, which were based on the morphologic features, time–signal intensity curves, and tCho concentrations of the lesions. Note that the accuracy (expressed as area under ROC curve) of the interpretations made by also assessing tCho measurements (solid lines) is higher than the accuracy of the interpretations made without assessing tCho measurements (dashed lines).

View larger version (26K): [in this window] [in a new window] [Download PPT slide]   Figure 4b. (a–d) ROC curves for the four readers based on the percentage probabilities of malignancy (0%–100%) assigned to the lesions by each reader. On each graph, the dashed line represents the ROC curve generated from the initial interpretations, which were based on the morphologic features and time–signal intensity curves of the lesions, and the solid line represents the ROC curve generated from the second interpretations, which were based on the morphologic features, time–signal intensity curves, and tCho concentrations of the lesions. Note that the accuracy (expressed as area under ROC curve) of the interpretations made by also assessing tCho measurements (solid lines) is higher than the accuracy of the interpretations made without assessing tCho measurements (dashed lines).

View larger version (29K): [in this window] [in a new window] [Download PPT slide]   Figure 4c. (a–d) ROC curves for the four readers based on the percentage probabilities of malignancy (0%–100%) assigned to the lesions by each reader. On each graph, the dashed line represents the ROC curve generated from the initial interpretations, which were based on the morphologic features and time–signal intensity curves of the lesions, and the solid line represents the ROC curve generated from the second interpretations, which were based on the morphologic features, time–signal intensity curves, and tCho concentrations of the lesions. Note that the accuracy (expressed as area under ROC curve) of the interpretations made by also assessing tCho measurements (solid lines) is higher than the accuracy of the interpretations made without assessing tCho measurements (dashed lines).

View larger version (26K): [in this window] [in a new window] [Download PPT slide]   Figure 4d. (a–d) ROC curves for the four readers based on the percentage probabilities of malignancy (0%–100%) assigned to the lesions by each reader. On each graph, the dashed line represents the ROC curve generated from the initial interpretations, which were based on the morphologic features and time–signal intensity curves of the lesions, and the solid line represents the ROC curve generated from the second interpretations, which were based on the morphologic features, time–signal intensity curves, and tCho concentrations of the lesions. Note that the accuracy (expressed as area under ROC curve) of the interpretations made by also assessing tCho measurements (solid lines) is higher than the accuracy of the interpretations made without assessing tCho measurements (dashed lines).

View larger version (51K): [in this window] [in a new window] [Download PPT slide]   Figure 5a. Breast MR imaging case in which 1H MR spectroscopic findings led to altered treatment recommendations. (a) Sagittal high-spatial-resolution 3D fat-suppressed fast low-angle shot MR images of the breast (13.5/4.1) obtained before (left) and 7 minutes after (middle) gadopentetate dimeglumine injection and with subtraction (right) show an 8.3-cm3 lesion. The box surrounding the lesion depicts the MR spectroscopic voxel. (b) Time–signal intensity (SI) curve measured from the lesion depicted in a. All four readers described this curve as showing slow enhancement during the initial phase followed by an enhancement plateau or persistent enhancement during the delayed phase. After evaluating the morphologic features and time–signal intensity curve of the lesion, three of the four readers did not recommend biopsy; rather, they recommended a 6-month follow-up examination. The fourth reader recommended biopsy. (c) 1H MR spectra measured from the lesion. The spectral peaks of mobile lipid, water, and tCho are labeled. The lines above and below the tCho peak represent the fitted tCho peak and the residual of the fit, respectively. The mean tCho concentration measured from this lesion was 1.78 mmol/kg ± 0.56. When the tCho measurement was presented to the readers in the second interpretation, three of the four readers changed their decision and recommended biopsy; the fourth reader kept the recommendation of biopsy. This patient received a diagnosis of invasive ductal carcinoma.

View larger version (7K): [in this window] [in a new window] [Download PPT slide]   Figure 5b. Breast MR imaging case in which 1H MR spectroscopic findings led to altered treatment recommendations. (a) Sagittal high-spatial-resolution 3D fat-suppressed fast low-angle shot MR images of the breast (13.5/4.1) obtained before (left) and 7 minutes after (middle) gadopentetate dimeglumine injection and with subtraction (right) show an 8.3-cm3 lesion. The box surrounding the lesion depicts the MR spectroscopic voxel. (b) Time–signal intensity (SI) curve measured from the lesion depicted in a. All four readers described this curve as showing slow enhancement during the initial phase followed by an enhancement plateau or persistent enhancement during the delayed phase. After evaluating the morphologic features and time–signal intensity curve of the lesion, three of the four readers did not recommend biopsy; rather, they recommended a 6-month follow-up examination. The fourth reader recommended biopsy. (c) 1H MR spectra measured from the lesion. The spectral peaks of mobile lipid, water, and tCho are labeled. The lines above and below the tCho peak represent the fitted tCho peak and the residual of the fit, respectively. The mean tCho concentration measured from this lesion was 1.78 mmol/kg ± 0.56. When the tCho measurement was presented to the readers in the second interpretation, three of the four readers changed their decision and recommended biopsy; the fourth reader kept the recommendation of biopsy. This patient received a diagnosis of invasive ductal carcinoma.

View larger version (16K): [in this window] [in a new window] [Download PPT slide]   Figure 5c. Breast MR imaging case in which 1H MR spectroscopic findings led to altered treatment recommendations. (a) Sagittal high-spatial-resolution 3D fat-suppressed fast low-angle shot MR images of the breast (13.5/4.1) obtained before (left) and 7 minutes after (middle) gadopentetate dimeglumine injection and with subtraction (right) show an 8.3-cm3 lesion. The box surrounding the lesion depicts the MR spectroscopic voxel. (b) Time–signal intensity (SI) curve measured from the lesion depicted in a. All four readers described this curve as showing slow enhancement during the initial phase followed by an enhancement plateau or persistent enhancement during the delayed phase. After evaluating the morphologic features and time–signal intensity curve of the lesion, three of the four readers did not recommend biopsy; rather, they recommended a 6-month follow-up examination. The fourth reader recommended biopsy. (c) 1H MR spectra measured from the lesion. The spectral peaks of mobile lipid, water, and tCho are labeled. The lines above and below the tCho peak represent the fitted tCho peak and the residual of the fit, respectively. The mean tCho concentration measured from this lesion was 1.78 mmol/kg ± 0.56. When the tCho measurement was presented to the readers in the second interpretation, three of the four readers changed their decision and recommended biopsy; the fourth reader kept the recommendation of biopsy. This patient received a diagnosis of invasive ductal carcinoma.

View larger version (59K): [in this window] [in a new window] [Download PPT slide]   Figure 6a. Breast MR imaging case in which the measured tCho concentration did not lead to a change in the recommended lesion management. (a) Sagittal high-spatial-resolution 3D fat-suppressed fast low-angle shot MR images of the breast (13.5/4.1) obtained before (left) and 7 minutes after (middle) gadopentetate dimeglumine injection and with subtraction (right) show a 0.39-cm3 lesion. The box surrounding the lesion depicts the MR spectroscopic voxel. (b) Time–signal intensity (SI) curve measured from the lesion depicted in a. All four readers described this curve as showing fast enhancement during the initial phase followed by an enhancement plateau during the delayed phase. After evaluating the morphologic features and time–signal intensity curve of the lesion, all four readers recommended biopsy. (c) 1H MR spectra measured from the lesion. The spectral peaks of mobile lipids, water, and tCho are labeled. The line above the tCho peak represents the minimal tCho concentration that is detectable with use of the quantification procedure. The mean tCho concentration measured from this lesion was 0 mmol/kg ± 1.73. When the tCho measurement was presented to the four readers in the second interpretation, none of them changed the recommendation of biopsy. This patient received a diagnosis of invasive ductal carcinoma.

View larger version (8K): [in this window] [in a new window] [Download PPT slide]   Figure 6b. Breast MR imaging case in which the measured tCho concentration did not lead to a change in the recommended lesion management. (a) Sagittal high-spatial-resolution 3D fat-suppressed fast low-angle shot MR images of the breast (13.5/4.1) obtained before (left) and 7 minutes after (middle) gadopentetate dimeglumine injection and with subtraction (right) show a 0.39-cm3 lesion. The box surrounding the lesion depicts the MR spectroscopic voxel. (b) Time–signal intensity (SI) curve measured from the lesion depicted in a. All four readers described this curve as showing fast enhancement during the initial phase followed by an enhancement plateau during the delayed phase. After evaluating the morphologic features and time–signal intensity curve of the lesion, all four readers recommended biopsy. (c) 1H MR spectra measured from the lesion. The spectral peaks of mobile lipids, water, and tCho are labeled. The line above the tCho peak represents the minimal tCho concentration that is detectable with use of the quantification procedure. The mean tCho concentration measured from this lesion was 0 mmol/kg ± 1.73. When the tCho measurement was presented to the four readers in the second interpretation, none of them changed the recommendation of biopsy. This patient received a diagnosis of invasive ductal carcinoma.

View larger version (15K): [in this window] [in a new window] [Download PPT slide]   Figure 6c. Breast MR imaging case in which the measured tCho concentration did not lead to a change in the recommended lesion management. (a) Sagittal high-spatial-resolution 3D fat-suppressed fast low-angle shot MR images of the breast (13.5/4.1) obtained before (left) and 7 minutes after (middle) gadopentetate dimeglumine injection and with subtraction (right) show a 0.39-cm3 lesion. The box surrounding the lesion depicts the MR spectroscopic voxel. (b) Time–signal intensity (SI) curve measured from the lesion depicted in a. All four readers described this curve as showing fast enhancement during the initial phase followed by an enhancement plateau during the delayed phase. After evaluating the morphologic features and time–signal intensity curve of the lesion, all four readers recommended biopsy. (c) 1H MR spectra measured from the lesion. The spectral peaks of mobile lipids, water, and tCho are labeled. The line above the tCho peak represents the minimal tCho concentration that is detectable with use of the quantification procedure. The mean tCho concentration measured from this lesion was 0 mmol/kg ± 1.73. When the tCho measurement was presented to the four readers in the second interpretation, none of them changed the recommendation of biopsy. This patient received a diagnosis of invasive ductal carcinoma.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION References

It is important to note that in this study, the sensitivity and specificity of breast MR imaging were defined not on the basis of the presence or absence of lesion enhancement but rather on the basis of the reader's ability to correctly identify a malignant breast lesion and recommend biopsy. Our results are consistent with those of other studies in that all malignant lesions were enhanced after gadopentetate dimeglumine administration (5–7); this result represents 100% sensitivity according to the definition used in previous breast MR imaging studies. Overall, the addition of ¹H MR spectroscopy led to an improvement in reader sensitivity of as much as 15% in this study setting. Even more important, the reader who had the highest sensitivity (94%) in the initial set of interpretations also had improved sensitivity after assessing the ¹H MR spectroscopic data.

Although the use of ¹H MR spectroscopy improved the specificity for all four readers, the absolute values of specificity were low. The low specificity observed in this study is thought to be a result of the small number of benign cases and the consequent context bias. One study (47) revealed that the inclusion of a large number of malignant cases relative to benign cases tends to lead to increased sensitivity and decreased specificity. Thus, a study with a larger number of benign cases is needed to ascertain the actual specificity that ¹H MR spectroscopy can yield in a clinical setting.

In addition, for all six pairs of readers, the addition of ¹H MR spectroscopy led to improved interobserver agreement regarding recommendations for patient treatment. Although it was not a primary purpose of this study, this result is clinically valuable because agreement between radiologists regarding the management of lesions is important for standardizing patient care. The Lesion Diagnosis Working Group is currently evaluating various breast MR imaging descriptors for lesion morphologic features and time–signal intensity curves to improve interobserver agreement regarding lesion assessment and patient treatment recommendations (37).

The precise mechanism(s) by which neoplastic tissue exhibits elevated levels of tCho is not fully understood. It has been proposed that the increased level of phosphocholine, the primary metabolite responsible for the tCho peak in neoplastic tissue, is a result of the increased synthesis of membranes by replicating cells (48,49). Elevated levels of tCho may also reflect a change in the balance between the biosynthetic and catabolic pathways in which choline compounds serve as both precursors and catabolites (48).

In agreement with the findings of earlier work (9–14), our study results show that the morphologic features and time–signal intensity curves of lesions are valuable tools for guiding radiologists in the MR imaging–based diagnosis of breast cancer. Although we did not quantitatively analyze the morphologic features and time–signal intensity curves of the lesions in this study, some of the malignant lesions did not show any detectable tCho levels even though they had morphologic features and/or time–signal intensity curve patterns that were characteristic of cancer. For example, one of the small (0.39-cm³) lesions that served as an example case in this study had no detectable tCho content. However, the morphologic features and time–signal intensity curve of this lesion were strong evidence of malignancy and resulted in a recommendation of biopsy by all four readers. The inability to detect tCho in this lesion most likely resulted from its small voxel size and hence low signal-to-noise ratio. It has been shown that voxels smaller than 1 cm³ yield large errors in tCho measurements (30).

Radiologists do not routinely assess breast MR images without initially viewing mammographic and/or US breast image findings. Therefore, our readers possibly could have had higher sensitivity and/or specificity if they had had access to the mammographic or US findings. Also, our readers were not provided with dual breast images; rather, they were shown sagittal images of one breast. Thus, side-to-side breast symmetry or asymmetry could not be assessed. Furthermore, reader sensitivity and specificity may have been negatively affected because some of the readers were not accustomed to evaluating the morphologic features of breast lesions on sagittal fat-suppressed MR images. Additionally, we assumed that the features of time–signal intensity curves generated at 4.0-T MR imaging are not different from the features of curves generated at lower-field-strength (1.5-T) examinations. Further studies are needed to determine whether the shape of the time–signal intensity curve is influenced by the field strength. Finally an important limitation to note is that owing to time constraints, we did not acquire T2-weighted MR images, which have been shown to be helpful in differentiating benign from malignant breast lesions (50). Thus, we possibly could have achieved higher specificity if we had assessed T2-weighted MR image findings.

The results of this study suggest that the addition of quantitative ¹H MR spectroscopy to the breast MR imaging examination may be valuable for improving the radiologist's ability to distinguish benign from malignant breast lesions. The additional information yielded by ¹H MR spectroscopy may be valuable when the morphologic features or time–signal intensity curve of the lesion is indeterminate. With the addition of ¹H MR spectroscopy, it may be possible to reduce the number of missed cancers and benign lesion biopsies and to improve the accuracy of surgical staging. The promising findings in this study were observed in a moderately small group of patients; thus, an observer performance study involving a larger patient series is needed. To the best of our knowledge, our investigation is the first observer performance study in which quantitative ¹H MR spectroscopy was assessed for its clinical value—specifically, in terms of improving reader accuracy in the interpretation of breast MR imaging cases.

	ACKNOWLEDGMENTS

 
The authors are grateful to Bibi Husain, Lou Forsythe, RN, Julliette Gay, RN, Susan Pappas-Varco, RN, and the General Clinical Research Center nursing staff for their help in coordinating the described study.

	FOOTNOTES

 

Abbreviations: BI-RADS = Breast Imaging and Reporting Data System • ROC = receiver operating characteristic • 3D = three-dimensional • tCho = total choline–containing compounds

Authors stated no financial relationship to disclose.

Author affiliations: Department of Radiology, Center for Magnetic Resonance Research Medical School, Minneapolis, Minn (S.M., P.J.B., E.H.B., M.G.P., M.G.); Departments of Radiology (S.M., P.J.B., E.H.B., M.T.N., L.I.E., T.H.E., M.G.), Biostatistics (C.T.L.), Surgery (T.M.T.), and Medicine (D.Y.), and Cancer Center (C.T.L., D.Y., M.G.), University of Minnesota, Minneapolis, Minn; Dept of Radiology, University of Wisconsin Hospital and Clinics, Madison, Wis (F.K.); Department of Radiology, Park Nicollet Breast Center, Minneapolis, Minn (M.C.L., B.A.L.); Department of Radiology, Suburban Radiologic Consultants, Minneapolis, Minn (R.A.C.); and Department of Radiology, Mayo Clinic, Rochester, Minn (K.R.B., K.K.A.).

Author contributions: Guarantors of integrity of entire study, S.M., P.J.B., M.G.; study concepts and design, S.M., P.J.B., M.G.; literature research, S.M., P.J.B., C.T.L., D.Y., M.G.; clinical studies, S.M., P.J.B., E.H.B., M.G.P., L.I.E., M.G.; data acquisition, S.M., P.J.B., E.H.B., M.G.P., M.G.; data analysis/interpretation, all authors; statistical analysis, S.M., C.T.L.; manuscript preparation, S.M., M.G.; manuscript definition of intellectual content, S.M., P.J.B., M.G.; manuscript editing, S.M., P.J.B., E.H.B., M.G.P., C.T.L., F.K., R.A.C., K.R.B., K.K.A., M.T.N., D.Y., M.G.; manuscript revision/review and final version approval, all authors

	References

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION References

 

1. Weir HK, Thun MJ, Hankey BF, et al. Annual report to the nation on the status of cancer, 1975–2000, featuring the uses of surveillance data for cancer prevention and control. J Natl Cancer Inst 2003; 95:1276–1299.[Abstract/Free Full Text]
2. Rodenko GN, Harms SE, Pruneda JM, et al. MR imaging in the management before surgery of lobular carcinoma of the breast: correlation with pathology. AJR Am J Roentgenol 1996; 167:1415–1419.[Abstract/Free Full Text]
3. Weinstein SP, Orel SG, Heller R, et al. MR imaging of the breast in patients with invasive lobular carcinoma. AJR Am J Roentgenol 2001; 176:399–406.[Abstract/Free Full Text]
4. Bedrosian I, Mick R, Orel SG, et al. Changes in the surgical management of patients with breast carcinoma based on preoperative magnetic resonance imaging. Cancer 2003; 98:468–473.[CrossRef][Medline]
5. Harms SE, Flamig DP, Hesley KL, et al. MR imaging of the breast with rotating delivery of excitation off resonance: clinical experience with pathologic correlation. Radiology 1993; 187:493–501.[Abstract/Free Full Text]
6. 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]
7. Orel SG, Schnall MD. MR imaging of the breast for the detection, diagnosis, and staging of breast cancer. Radiology 2001; 220:13–30.[Abstract/Free Full Text]
8. Warner E, Plewes DB, Shumak RS, et al. Comparison of breast magnetic resonance imaging, mammography, and ultrasound for surveillance of women at high risk for hereditary breast cancer. J Clin Oncol 2001; 19:3524–3531.[Abstract/Free Full Text]
9. Orel SG, Schnall MD, Powell CM, et al. Staging of suspected breast cancer: effect of MR imaging and MR-guided biopsy. Radiology 1995; 196:115–122.[Abstract/Free Full Text]
10. Nunes LW, Schnall MD, Siegelman ES, et al. Diagnostic performance characteristics of architectural features revealed by high spatial-resolution MR imaging of the breast. AJR Am J Roentgenol 1997; 169:409–415.[Abstract/Free Full Text]
11. 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]
12. Boetes C, Barentsz JO, Mus RD, et al. MR characterization of suspicious breast lesions with a gadolinium-enhanced TurboFLASH subtraction technique. Radiology 1994; 193:777–781.[Abstract/Free Full Text]
13. 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.
14. Knopp MV, Weis E, Sinn HP, et al. Pathophysiologic basis of contrast enhancement in breast tumors. J Magn Reson Imaging 1999; 10:260–266.[CrossRef][Medline]
15. Weinstein D, Strano S, Cohen P, Fields S, Gomori JM, Degani H. Breast fibroadenoma: mapping of pathophysiologic features with three-time-point, contrast-enhanced MR imaging—pilot study. Radiology 1999; 210:233–240.[Abstract/Free Full Text]
16. Wurdinger S, Kamprath S, Eschrich D, Schneider A, Kaiser WA. False-negative findings of malignant breast lesions on preoperative magnetic resonance mammography. Breast 2001; 10:131–139.[CrossRef][Medline]
17. Gribbestad IS, Peterson SB, Fjosne HE, Kvinnsland S, Krane J. ¹H NMR spectroscopic characterization of perchloric acid extracts from breast carcinomas and non-involved breast tissue. NMR Biomed 1994; 7:181–194.[Medline]
18. Mackinnon WB, Barry PA, Malycha PL, et al. Fine-needle biopsy specimens of benign breast lesions distinguished from invasive cancer ex vivo with proton MR spectroscopy. Radiology 1997; 204:661–666.[Abstract/Free Full Text]
19. Cheng LL, Chang IW, Smith BL, Gonzalez RG. Evaluating human breast ductal carcinomas with high-resolution magic-angle spinning proton magnetic resonance spectroscopy. J Magn Reson 1998; 135:194–202.[CrossRef][Medline]
20. Katz-Brull R, Seger D, Rivenson-Segal D, Rushkin E, Degani H. Metabolic markers of breast cancer: enhanced choline metabolism and reduced choline-ether-phospholipid synthesis. Cancer Res 2002; 62:1966–1970.[Abstract/Free Full Text]
21. Gribbestad IS, Sitter B, Lundgren S, Krane J, Axelson D. Metabolic composition in breast tumors examined by proton nuclear resonance spectroscopy. Anticancer Res 1999; 19:1737–1746.[Medline]
22. Gribbestad IS, Singstad TE, Nilsen G, et al. In vivo ¹H MRS of normal breast and breast tumors using a dedicated double breast coil. J Magn Reson Imaging 1998; 8:1191–1197.[Medline]
23. Roebuck JR, Cecil KM, Schnall MD, Lenkinski RE. Human breast lesions: characterization with proton MR spectroscopy. Radiology 1998; 209:269–275.[Abstract/Free Full Text]
24. Kvistad KA, Bakken IJ, Gribbestad IS, et al. Characterization of neoplastic and normal human breast tissues with in vivo ¹H MR spectroscopy. J Magn Reson Imaging 1999; 10:159–164.[CrossRef][Medline]
25. Jagannathan NR, Kumar M, Seenu V, et al. Evaluation of total choline from in-vivo volume localized proton MR spectroscopy and its response to neoadjuvant chemotherapy in locally advanced breast cancer. Br J Cancer 2001; 84:1016–1022.[CrossRef][Medline]
26. Yeung DK, Cheung HS, Tse GM. Human breast lesions: characterization with contrast-enhanced in vivo proton MR spectroscopy—initial results. Radiology 2001; 220:40–46.[Abstract/Free Full Text]
27. Cecil KM, Schnall MD, Siegelman ES, Lenkinski RE. The evaluation of human breast lesions with magnetic resonance imaging and proton magnetic resonance spectroscopy. Breast Cancer Res Treat 2001; 68:45–54.[CrossRef][Medline]
28. Katz-Brull R, Lavin PT, Lenkinski RE. Clinical utility of proton magnetic resonance spectroscopy in characterizing breast lesions. J Natl Cancer Inst 2002; 94:1197–1203.[Abstract/Free Full Text]
29. Tse GM, Cheung HS, Pang LM, et al. Characterization of lesions of the breast with proton MR spectroscopy: comparison of carcinomas, benign lesions, and phyllodes tumors. AJR Am J Roentgenol 2003; 181:1267–1272.[Abstract/Free Full Text]
30. Bolan PJ, Meisamy S, Baker EH, et al. In vivo quantification of choline compounds in the breast with ¹H MR spectroscopy. Magn Reson Med 2003; 50:1134–1143.[CrossRef][Medline]
31. Kim JK, Park SH, Hyun ML, et al. In vivo ¹H-MRS evaluation of malignant and benign breast disease. Breast 2003; 12:179–182.[CrossRef][Medline]
32. Huang W, Fisher PR, Dulaimy K, Tudorica LA, O'Hea B, Button TM. Detection of breast malignancy: diagnostic MR protocol for improved specificity. Radiology 2004; 232:585–591.[Abstract/Free Full Text]
33. Bolan PJ, DelaBarre L, Baker EH, et al. Eliminating spurious lipid sidebands in ¹H MRS of breast lesions. Magn Reson Med 2002; 48:215–222.[CrossRef][Medline]
34. Merkle H, DelaBarre L, Bolan PJ, et al. Transceive quadrature breast coils and applications at 4 Tesla (abstr). In: Proceedings of the Ninth Meeting of the International Society for Magnetic Resonance in Medicine. Berkeley, Calif: International Society for Magnetic Resonance in Medicine, 2001; 1114.
35. Garwood M, DelaBarre L. The return of the frequency sweep: designing adiabatic pulses for contemporary NMR. J Magn Reson 2001; 153:155–177.[CrossRef][Medline]
36. Tkac I, Starcuk Z, Choi IY, Gruetter R. In vivo ¹H NMR spectroscopy of rat brain at 1 msec echo time. Magn Reson Med 1999; 41:649–656.[CrossRef][Medline]
37. Ikeda DM, Hylton NM, Kinkel K, et al. Development, standardization, and testing of a lexicon for reporting contrast-enhanced breast magnetic resonance imaging studies. J Magn Reson Imaging 2001; 13:889–895.[CrossRef][Medline]
38. Le CT. Screening test. In: Introductory biostatistics. Hoboken, NJ: Wiley, 2003; 5–8.
39. Bamber D. The area above the ordinal dominance graph and the area below the receiver operating graph. J Math Psychol 1975; 12:387–415.[CrossRef]
40. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristics (ROC) curve. Radiology 1982; 143:29–36.[Abstract/Free Full Text]
41. Le CT. Receiver operating characteristics curve. In: Introductory biostatistics. Hoboken, NJ: Wiley, 2003; 336–338.
42. Hanley JA, McNeil BJ. Method for comparing the areas under the receiver operating characteristic curves derived from the same cases. Radiology 1983; 148:839–843.[Abstract/Free Full Text]
43. Le CT, Lindgren BR. Construction and comparison of two receiving operating characteristic curves derived from the same samples. Biometric J 1995; 37:869–877.
44. Le CT. Evaluation of confounding effects in ROC studies. Biometrics 1997; 53:998–1007.[CrossRef][Medline]
45. Le CT. Reader agreement. In: Introductory biostatistics. Hoboken, NJ: Wiley, 2003; 118–120.
46. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33:159–174.[CrossRef][Medline]
47. Egglin T, Fienstein A. Context bias: a problem in diagnostic radiology. JAMA 1996; 276:1752–1755.[Abstract/Free Full Text]
48. Podo F. Tumour phospholipid metabolism. NMR Biomed 1999; 12:413–439.[CrossRef][Medline]
49. Aboagye EO, Bhujwalla ZM. Malignant transformation alters membrane choline phospholipid metabolism of human mammary epithelial cells. Cancer Res 1999; 59:80–84.[Abstract/Free Full Text]
50. Kuhl CK, Klaschik S, Mielcarek P, Gieseke J, Wardelmann E, Schild HH. Do T2-weighted pulse sequences help the differential diagnosis of enhancing lesions in dynamic breast MRI? J Magn Reson Imaging 1999; 9:187–196.[CrossRef][Medline]

This article has been cited by other articles:

				F. Sardanelli, A. Fausto, G. Di Leo, R. de Nijs, M. Vorbuchner, and F. Podo In Vivo Proton MR Spectroscopy of the Breast Using the Total Choline Peak Integral as a Marker of Malignancy Am. J. Roentgenol., June 1, 2009; 192(6): 1608 - 1617. [Abstract] [Full Text] [PDF]

				C. K. Kuhl, F. Traber, J. Gieseke, W. Drahanowsky, N. Morakkabati-Spitz, W. Willinek, M. von Falkenhausen, C. Manka, and H. H. Schild Whole-Body High-Field-Strength (3.0-T) MR Imaging in Clinical Practice * Part II. Technical Considerations and Clinical Applications Radiology, April 1, 2008; 247(1): 16 - 35. [Abstract] [Full Text] [PDF]

				L. Bartella and W. Huang Proton (1H) MR Spectroscopy of the Breast RadioGraphics, October 1, 2007; 27(suppl_1): S241 - S252. [Abstract] [Full Text] [PDF]

				P. Stanwell and C. Mountford In Vivo Proton MR Spectroscopy of the Breast RadioGraphics, October 1, 2007; 27(suppl_1): S253 - S266. [Abstract] [Full Text] [PDF]

				C. K. Kuhl Current Status of Breast MR Imaging * Part 2. Clinical Applications Radiology, September 1, 2007; 244(3): 672 - 691. [Abstract] [Full Text] [PDF]

				V. S. Lee, E. M. Hecht, B. Taouli, Q. Chen, K. Prince, and N. Oesingmann Body and Cardiovascular MR Imaging at 3.0 T Radiology, September 1, 2007; 244(3): 692 - 705. [Abstract] [Full Text] [PDF]

				T. G. ODLE Breast MR Radiol. Technol., September 1, 2006; 78(1): 45M - 66M. [Abstract] [Full Text] [PDF]

This Article Abstract Figures Only Full Text (PDF) 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 HighWire Citing Articles via Google Scholar Google Scholar Articles by Meisamy, S. Articles by Garwood, M. Search for Related Content PubMed PubMed Citation Articles by Meisamy, S. Articles by Garwood, M.