Positive and Negative Predictive Values of BI-RADS}-MRI Descriptors for Focal Breast Masses

Transcription

1 Magnetic Resonance in Medical Sciences, Vol. 5, No. 1, pp. 7 15, 2006 MAJOR PAPER Positive and Negative Predictive Values of BI-RADS}-MRI Descriptors for Focal Breast Masses Mitsuhiro TOZAKI*, Takao IGARASHI, and Kunihiko FUKUDA Department of Radiology, The Jikei University School of Medicine Nishi-Shimbashi, Minato-ku, Tokyo , Japan (Received January 13, 2006; Accepted February 8, 2006) Purpose: The purposes of this study were to evaluate the positive and negative predictive valuesofthebi-rads}-mr descriptors of focal masses and to develop an interpretation model based on the kinetic and morphologic parameters. Methods: Retrospective review was performed of 171 consecutive focal breast masses. MR imaging was performed on a 1.5T system using the volumetric interpolated breathhold examination sequence (mean partition thickness, 1.2 mm; time of acquisition, 35 s). Kinetic enhancement patterns were assessed by visually comparing signal intensity on the dynamic images acquired at 60 s and 4 min (washout, plateau, and persistent). Results: There were 126 malignant and 45 benign lesions. The most frequent morphological ˆnding among the malignant lesions was heterogeneous internal enhancement in the delayed phase (96z; Pº0.001); the most frequent ˆnding in benign lesions was smooth margin or smooth shapewmargin (80 to 82z; Pº0.001). The features with the highest positive predictive value for carcinoma were spiculated margin (100z), delayed central enhancement (100z), delayed enhancing internal septations (97z), and irregular shape (97z). The characteristics with the highest diagnostic accuracy for malignancy were spiculated margin (100z) and heterogeneous enhancement following washout in the smooth shapewmargin group (100z). The sensitivity, speciˆcity, and positive and negative predictive values of an interpretation model based on a combination of the morphologic characteristics and kinetic information were 99z, 89z, 96z, and98z, respectively. Conclusion: A combination of morphologic criteria, particularly lesion shapewmargin and internal heterogeneity, and kinetic information is useful for dišerentiating benign and malignant lesions. Keywords: breast neoplasm, MR imaging, interpretation, BI-RADS} Introduction Magnetic resonance (MR) imaging of the breast has shown diagnostic sensitivity of 94 to 99z for invasive breast cancer, 1 5 whereas varying speciˆcity has been reported (37 to 86z). 1 4,6,7 To improve speciˆcity, detailed assessment of lesion morphology using three-dimensional (3D) MR imaging and of kinetic patterns using dynamic protocols has been reported useful. 8 Several recent studies have reported diagnostic criteria combining the morphologic and kinetic characteristics The recently published ``Breast Imaging Reporting and Data System } (BI-RADS } )'' 12 included the ˆrst edition of a lexicon of terms for breast MR. However, a standardized classiˆcation scheme for the interpretation of lesions does not exist. Our goals were to evaluate the positive and negative predictive values of the BI-RADS } -MR descriptors of focal masses and develop an interpretation model based on the kinetic and morphologic parameters. *Corresponding author, Phone: , Fax: , e-tozaki@keh.biglobe.ne.jp Present address: Division of Diagnostic Imaging, Breast Center, Kameda Medical Center, 929 Higashi-cho, Kamogawa, Chiba , Japan Materials and Methods Patients We retrospectively reviewed 425 consecutive patients in two hospitals who had undergone breast 7

2 8 M. Tozaki et al. MR imaging between July 2000 and February Inclusion criteria were: breast MR imaging conducted in a prone position, no administration of neoadjuvant chemotherapy, and diagnosis of focal breast masses established histologically. Excluded were 25 patients with breast carcinoma who had undergone MR imaging in the supine position 13 and one patient with a poorly enhancing mass (mucinous carcinoma). A mass was deˆned as a 3D, space-occupying (of any size), enhancing lesion with convex borders. 12 In all, 155 consecutive patients (age range, 23 to 83 years; mean, 49 years) with 171 histopathologically diagnosed breast masses were selected for the study. Magnetic resonance imaging MR imaging was performed using a 1.5T system (Symphony; Siemens Medical Solutions, Erlangen, Germany; maximum gradient ˆeld strength, 30 mtwm). All patients were examined in a prone position using a double CP breast array coil. A transverse, fat-suppressed, T2-weighted fast spinecho sequence was performed with: repetition timewecho time (TRWTE), 3500W78; ˆeld of view (FOV), 20 cm; matrix size, ; and slice thickness, 5 mm with a 1 mm gap. T 2-weighted imaging was performed on one side, and when lesions were bilateral, separate imaging was conducted for each side. T2*-weighted ˆrst-pass perfusion images were obtained in the transverse plane before, during, and after bolus injection of 0.1 mmol Gd-DTPAWkg at a rate of 3 mlws followed by a 20-mL saline ush using an automatic injector. The T 2 *-weighted perfusion data were collected with a multisection echo-planar sequence (151W61; ip angle [FA], 709; FOV,25cm;5-mmsection thickness; and 1.5-mm intersection gap). A 3D fat-suppressed volumetric interpolated breath-hold examination (VIBE) sequence was obtained before and 60 s, 100 s, and 4 min after the start of the intravenous administration. The ašected single breast was examined on the ˆrst- and third-phase dynamic images acquired at 60 s and 4 min, respectively, and both breasts were examined on images obtained in the second phase at 100 s. If incidental suspicious enhancement was detected in the contralateral breast during the second phase, additional images of both breasts were obtained immediately during the subsequent third phase. For VIBE, the images were prepared mainly from the coronal cross section; but for the sake of uniformity with the perfusion images, transverse sections were also prepared for the third phase only. The parameters for the VIBE sequence of the single breast examination were: TR WTE, 3.7 W1.7; FA, 259; FOV,27cm;matrix, ; receiver bandwidth, 490 HzWpixel; mean partition thickness, 1.2 mm; and time of acquisition, 35 s. Section thickness varied depending on the size of the breast and ranged from 1 to 1.5 mm without a gap. Image interpretation Two radiologists evaluated all cases retrospectively and arrived at a consensus; the radiologists were unaware of conventional imaging ˆndings or the histopathological diagnosis. The morphological parameters evaluated were lesion shape (round, oval, lobular, irregular); mass margin (smooth, irregular, spiculated); and pattern of internal enhancement (homogeneous, heterogeneous, rim enhancement, enhancing internal septations, central enhancement, dark internal septations). 12 In this study, the lesion shapewmargin was also classiˆed into 3 categories: smooth (smooth Wround, smoothwoval, smoothwlobulated); irregular (irregular margin or irregular shape); or spiculated (spiculated margin). 11 Internal enhancement was classiˆed as homogeneous or heterogeneous, and the presence or absence of MR signs of the lesions, such as rim enhancement, enhancing internal septations, central enhancement, and dark internal septations, was also evaluated. In the cases showing positive rim enhancement, the pattern of internal enhancement was assessed after excluding the enhancing rim. 11 The lesion shapewmargin was evaluated on the coronal images and transverse, sagittal multiplanar reformations (MPRs) acquired at 60 s (early phase) and 4 min (delayed phase). If the margin of the lesion became progressively more indistinct with time, the lesion shapewmargin was evaluated in the early phase images. The decision on the pattern of internal enhancement was based on images obtained during both the early and delayed phases. Kinetic enhancement patterns were assessed by visually comparing signal intensity on the dynamic images acquired at 60 s and 4 min using the same window and level settings. By deˆnition, any decline in signal intensity between 60 s and 4 min after injection of contrast material was considered a ``washout'' enhancement pattern. ``Plateau'' enhancement was considered to be stabilized enhancement without change in signal intensity between 60 s and 4 min. ``Persistent'' enhancement was considered to be an increase in signal intensity throughout the dynamic period. Each lesion was characterized according to the strongest enhancement pattern visible over the entire lesion. Statistical analysis To analyze group dišerences from dichotomous variables, chi-square and Fisher's exact tests were Magnetic Resonance in Medical Sciences

3 BI-RADS}-MRI Descriptors for Breast Masses 9 Table 1. Frequency of MR imaging parameters in breast lesions MR Descriptor Benign (n=45) Malignant (n=126) P-value a Lesion shape º0.001 Round 9 (20) 2 (2) Oval 14 (31) 2 (2) Lobular 19 (42) 39 (31) Irregular 3 (7) 83 (66) Type of margin º0.001 Smooth 37 (82) 13 (10) Irregular 8 (18) 49 (39) Spiculated 0 64 (51) Lesion shapewmargin º0.001 Smooth 36 (80) 12 (10) Irregular 9 (20) 50 (40) Spiculated 0 64 (51) Internal enhancement (Homogeneous or Heterogeneous) Early phase º0.001 Homogeneous 32 (71) 12 (10) Heterogeneous 13 (29) 114 (90) Delayed phase º0.001 Homogeneous 34 (76) 5 (4) Heterogeneous 11 (24) 121 (96) Internal enhancement (MR signs of the lesion) Rim enhancement Early phase 15 (33) 69 (55) 0.03 Delayed phase 13 (29) 93 (74) º0.001 Enhancing internal septation Early phase 4 (9) 73 (58) º0.001 Delayed phase 3 (7) 95 (75) º0.001 Central enhancement Early phase 0 0 n.s. Delayed phase 0 6 (5) n.s. Dark internal septation Early phase 5 (11) 3 (2) n.s. Delayed phase 6 (13) 1 (1) Kinetic pattern º0.001 Persistent 32 (71) 27 (21) Plateau 7 (16) 18 (14) Washout 6 (13) 81 (64) Note: Percentages are shown in parentheses. a Chi-square and Fisher's exact tests. n.s.=not signiˆcant employed. Fisher's exact or chi-square test was used to analyze a 2 2 contingency table and the latter only for 2 2 orgreater.ap value less than 0.05 was considered a statistically signiˆcant dišerence. Results Histological results Histological examination demonstrated 126 malignant and 45 benign lesions. The histological types Vol. 5 No. 1, 2006

4 10 M. Tozaki et al. Table 2. Relationship of the histological to MR ˆndings Lesion shapewmargin Internal Enhancement* Kinetic Pattern Histological Results Smooth Irregular Spiculated Homo Hetero RE EIS CE DIS Persistent Plateau Washout Malignant n= Ductal carcinoma in situ (100) 0 2 (29) 5 (71) 4 (57) 3 (43) (14) 6 (86) Invasive ductal carcinoma 99 9 (9) 32 (32) 58 (59) 1 (1) 98 (99) 77 (78) 83 (84) 6 (6) 0 18 (18) 16 (16) 65 (66) Invasive lobular carcinoma 5 1 (20) 1 (20) 3 (60) 1 (20) 4 (80) 1 (20) 3 (60) (60) 1 (20) 1 (20) Tubular carcinoma (50) 2 (50) 0 4 (100) 3 (75) 2 (50) (25) 0 3 (75) Medullary carcinoma (100) 0 1 (100) 1 (100) (100) Mucinous carcinoma 6 1 (17) 5 (83) 0 1 (17) 5 (83) 4 (67) 2 (33) (83) 0 1 (17) Invasive micropapillary carcinoma (100) (100) 1 (100) 1 (100) (100) Metaplastic carcinoma (100) (100) 2 (100) 1 (50) (100) Malignant phyllodes tumor 1 1 (100) (100) (100) (100) Benign n= Intraductal papilloma 9 6 (67) 3 (33) 0 8 (89) 1 (11) 4 (44) (22) 1 (11) 6 (67) Benign epithelial proliferation 7 7 (100) (100) (71) 2 (29) 0 Fibroadenoma (76) 6 (24) 0 19 (76) 6 (24) 7 (28) 2 (8) 0 3 (12) 22 (88) 3 (12) 0 Phyllodes tumor 3 3 (100) (100) 1 (33) (100) 3 (100) 0 0 Nipple adenoma 1 1 (100) (100) 1 (100) 1 (100) (100) 0 Note: Percentages are shown in parentheses. *Internal enhancement was evaluated on delayed phase. Homo: Homogeneous enhancement Hetero: Heterogeneous enhancement RE: Rim enhancement EIS: Enhancing internal septation CE: Central enhancement DIS: Dark internal septation of carcinoma included ductal carcinoma in situ (n=7), invasive ductal carcinoma (n=99), invasive lobular carcinoma (n=5), tubular carcinoma (n= 4), medullary carcinoma (n=1), mucinous carcinoma (n=6), invasive micropapillary carcinoma (n=1), metaplastic carcinoma (n=2), and malignant phyllodes tumor (n=1). The 45 benign lesions included intraductal papilloma (n=9), epithelial proliferation (n=7), ˆbroadenoma (n=25), phyllodes tumor (n=3), and nipple adenoma (n =1). The average tumor size of the malignant lesions was 22.9 mm (range, 6 to 80 mm), and the average tumor size of the benign lesions was 11.9 mm (range, 3 to 75 mm). Frequency of recognition of the BI-RADS } -MR descriptors Table 1 shows the frequencies of the morphological parameters and kinetic patterns. The most frequent morphological ˆnding among the malignant lesions was heterogeneous internal enhancement in the delayed phase (96z; Pº0.001), followed by heterogeneous internal enhancement in the early phase (90z; Pº0.001); smooth margin or smooth shape Wmargin were the most frequent ˆndings in benign lesions (80 to 82z; Pº0.001). Two lesions with a smooth margin were classiˆed into the irregular shape Wmargin group because of irregular shape, one a ˆbroadenoma and the other an invasive lobular carcinoma. Regarding kinetic enhancement, the most frequent pattern observed in the malignant lesions was a washout pattern (64z); on the other hand, 71z of the benign lesions exhibited a persistent pattern (Pº0.001). The features with the highest positive predictive value for carcinoma were spiculated margin (100z), delayed central enhancement (100z), delayed enhancing internal septations (97z), and irregular shape (97z). The features with the highest positive predictive value for benign lesions were oval shape (88z), homogeneous internal enhancement in the delayed phase (87z),anddarkinternal septations (86z). Relationship of histological to MR findings Table 2 shows the relationship between the histological characteristics and the frequencies of the various morphological and kinetic parameters. All 7 cases with ductal carcinoma in situ showed irregular shapewmargin and nonpersistent kinetic pattern. Central enhancement was only observed in cases of invasive ductal carcinoma; all 6 cases showing central enhancement showed a washout pattern and delayed rim enhancement. Histologically, all lesions were characterized by the presence of large solid clusters of cancer cells with expansive growth forming distinct boundaries. Delayed rim enhancement correlated well with marginal ˆbrosis, and pooling of contrast material following washout was explained by central ˆbrosis. Four of the 5 cases of Magnetic Resonance in Medical Sciences

5 BI-RADS}-MRI Descriptors for Breast Masses 11 Table 3. Positive and negative predictive values and accuracy of MR imaging parameters Smooth shapewmargin (n=48) Descriptor PPV (z) NPV (z) Accuracy (z) Washout 75 (12W16) 100 (32W32) 92 (44W48) Delayed rim enhancement following washout 77 (10W13) 94 (33W35) 90 (43W48) Delayed enhancing internal septation following washout 100 (10W10) 95 (36W38) 96 (46W48) Heterogeneous on delayed phase 57 (12W21) 100 (27W27) 81 (39W48) Heterogeneous internal enhancement following washout 100 (12W12) 100 (36W36) 100 (48W48) Irregular shapewmargin (n=59) Washout 95 (35W37) a 32 (7W22) 71 (42W59) Delayed enhancing internal septation following washout 100 (29W29) 30 (9W30) 64 (38W59) Heterogeneous on delayed phase 96 (46W48) b 64 (7W11) c,d 90 (53W59) Spiculated shapewmargin (n=64) 100 (64W64) 100 (64W64) Washout (n=87) Nonsmooth shapewmargin (irregular and spiculated) 97 (69W71) a 25 (4W16) 84 (73W87) Heterogeneous internal enhancement following washout 100 (79W79) 75 (6W8) c 98 (85W87) Nonwashout (n=84) Spiculated shapewmargin 100 (30W30) 72 (39W54) 82 (69W84) Heterogeneous on delayed phase 79 (42W53) 90 (28W31) 83 (70W84) a Intraductal papilloma (n=2) b Fibroadenoma (n=1) and intraductal papilloma (n=1) c Ductal carcinoma in situ (n=2) d Invasive lobular carcinoma (n=1) and mucinous carcinoma (n=1) invasive lobular carcinoma showed a non-washout pattern (80z). All 6 mucinous carcinomas showed non-spiculated margin, and ˆve of the 6 lesions exhibited a persistent pattern (83z). One case of malignant phyllodes tumor exhibited smooth shapew margin, dark internal septations, and washout pattern. Among the benign lesions, all the lesions showing washout pattern were intraductal papillomas. Four cases of intraductal papilloma also showed delayed rim enhancement (44z). Dark internal septations were observed in the cases of ˆbroadenoma and phyllodes tumor. All 3 phyllodes tumors showed smooth shapewmargin, dark internal septations, and persistent pattern. Suggested interpretation model Table 3 shows the positive and negative predictive values and accuracy of the combinations of the MR imaging parameters. The characteristics with the highest diagnostic accuracy for malignancy were spiculated margin (100z) and heterogeneous enhancement following washout in the smooth shapewmargin group (100z). On the other hand, in the irregular shapewmargin group, the feature with the highest positive predictive value for carcinoma was heterogeneous enhancement in the delayed phase (96z). Figure 1 illustrates the proposed interpretation model based on the results of this study. The histological diagnoses in the 2 false-positive lesions with heterogeneous enhancement in the irregular shapew margin group were ˆbroadenoma and intraductal papilloma. The remaining 3 false-positive lesions included one ˆbroadenoma and 2 intraductal papillomas; on the other hand, the false-negative lesion was histologically established to be a mucinous carcinoma. Sensitivity, speciˆcity, and positive and negative predictive values were 99z, 89z, 96z, and 98z, respectively. Discussion Breast MR imaging has emerged for several purposes: local staging before surgery, dišerentiation between recurrence and surgical scar, search for primary carcinoma in patients with axillary adenopathy, evaluation of ešect of chemotherapy, and screening of women at high risk. Although the recently published BI-RADS } includes the ˆrst edition of a breast MR lexicon, 12 astandardized classiˆcation scheme for the interpretation of lesions does not exist. The lack of standardized Vol. 5 No. 1, 2006

6 12 M. Tozaki et al. Fig. 1. The interpretation model proposed on the basis of the results in this study. Benign terminal nodes are shaded. The histological diagnoses in the 2 false-positive lesions with heterogeneous enhancement in the irregular shapewmargin group were ˆbroadenoma and intraductal papilloma. The remaining 3 false-positive lesions included one ˆbroadenoma and 2 intraductal papillomas; the false-negative lesion was histologically established to be a mucinous carcinoma. Positive and negative predictive values were 96z (125W130) and 98z (40W41). interpretation criteria has been ascribed to dišerences in image acquisition protocols, from dynamic scanning 1,3,6,7 to high spatial resolution MR imaging However, Orel and Schnall 5 predicted that images of both high spatial and high temporal resolution may ultimately come to dominate MR imaging protocols of the breast. Recently, the development of new MR imaging protocols of the breast has enabled simultaneous acquisition of imagesofhighspatialandhightemporalresolution. One of the representative advanced protocols is a 3D fat-suppressed gradient-recalled echo technique with volumetric interpolation, ˆrst described by Rofsky and associates. 19 The 3D VIBE sequence has enabled evaluation of the distribution and morphological characteristics of breast tumors using near-isotropic MPRs. 11,13,20 In the present study, the morphological characteristics based on the BI-RADS } -MRI descriptors and semidynamic information using the 3- timepoint technique with the VIBE sequence were assessed for dišerentiating benign and malignant focal breast masses. Among the morphological parameters, mass margin was found the most useful for dišerentiating benign and malignant lesions. 9 11,15 18 Irregular borders and spiculated margins were the features primarily associated with breast carcinoma at ultrasound as well. 21 However, because it is di cult to entirely dišerentiate mass margin from mass shape, 22 we assessed the morphological parameters on the basis of both lesion margin and shape. Because smooth or lobulated margin on MR images has been reported to be highly predictive of benignity, 15 smooth margin and lobulated margin were included in the same group in our interpretation method. Ultimately, 2 lesions with smooth margins were classiˆed into the irregular shapewmargin group because of irregular shape, one a ˆbroadenoma and the other an invasive lobular carcinoma. Our categorization of the lesion shape Wmargin was intended to minimize malignant lesions in the smooth margin group. In our study, the most frequent morphological ˆnding among the malignant lesions was heterogeneous internal enhancement (in the delayed phase, 96z; in the early phase, 90z). Wedegartner and colleagues 23 also reported that the most reliable morphological ˆnding suggestive of breast carcinoma was irregular contour and heterogeneous enhancement. In interpreting images of high spatial resolution, combined assessment of the lesion conˆguration, kinetic pattern, and internal heterogeneity is considered useful for the dišerential diagnosis of breast masses because the morphological appearance on MR images depends on the spatial resolution. Magnetic Resonance in Medical Sciences

7 BI-RADS}-MRI Descriptors for Breast Masses In contrast, the features with the highest positive predictive value for carcinoma were spiculated margin (100z), delayed central enhancement (100z), enhancing internal septations in the delayed phase (97z), and irregular shape (97z). Contrary to the results reported by Liberman's group, 18 rim enhancement was not a strong indicator of malignancy, whereas central enhancement and the presence of enhancing internal septations in the delayed phase were useful indicators of malignancy in our study. This discrepancy may be related to the dišerence in mean size of the lesions between the 2 studies. The mean diameter of the malignant lesions (23 mm) in our study was greater than that (5 mm) reported in Liberman's. 18 Central enhancement and the presence of enhancing internal septations could also depend on lesion diameter because these signs are assessed within the tumor in contrast to rim enhancement, which is assessed at the peripheral or marginal region of the lesion. Numerous studies have reported the ˆnding of rim enhancement suggestive of malignancy. 3,14 18 Buadu and associates 24 reported dišerent patterns of rim enhancement and concluded that early enhancement of the rim with progression to the center was fairly speciˆc for carcinomas. However, almost 30z of the benign lesions reported by Wedegartner's group 23 exhibited positive rim enhancement. Also in our study, early rim enhancement was found in 33z of benign lesions and 55z of malignant lesions, whereas delayed rim enhancement was observed in 29z of benign lesions and 74z of malignant lesions. Thus, we found that delayed rim enhancement to be a highly frequent morphological ˆnding among breast carcinomas (Pº0.001), although it was not found to have a high positive predictive value (88z; 93W106). Regarding kinetic pattern, visual assessment of washout using the 3-timepoint technique was ˆrst reported by Kinkel and colleagues. 9 The positive predictive value of the visually assessed washout pattern in high spatial resolution sequences was reported in a previous study to be 91z (29W32). 9 In the present study, the positive predictive value of the washout pattern for the diagnosis of malignancy was 93z (81W87). In addition, Kuhl's group 7 reported the usefulness of evaluating the shape of the intensity curve of the time signal for the dišerential diagnosis of breast lesions; the positive predictive value of a washout curve was reported to be 87z (58W67). However, Liberman's group 18 reported that a visually assessed washout pattern was not a signiˆcant predictor of breast carcinoma (33z;8W24). The apparently higher positive predictive value of the washout pattern for malignancy in 13 our study may be attributable to two major factors: (1) dišerence in the mean size of the lesions and (2) dišerences in the protocols employed for dynamic scanning between the 2 studies. We visually assessed the kinetic characteristics of the tumor by comparing the signal intensity on dynamic scans obtained during the ˆrst and third phases. The scans of the third phase were acquired at 4 min after the start of intravenous administration of the contrast material. It is possible that 4 min could be a reasonable timepoint for stating if a malignant washout is present or not, excluding a benign washout. The ˆnal goal of this study was the development of an interpretation model on the basis of a combination of the morphologic characteristics of the lesions and kinetic information (Fig. 1). In terms of internal enhancement and kinetic patterns, the features most suggestive of a benign lesion are persistent pattern and homogeneous enhancement, determined according to the scoring system proposed by Fischer (persistent, 0 points; homogeneous enhancement, 0 points). 25 Therefore, among the lesions showing an irregular shape Wmargin, downstaging was performed using a combination of homogeneous internal enhancement in the delayed phase and persistent pattern. Finally, the sensitivity, speciˆcity, and positive and negative predictive values of these features for the diagnosis of malignancy were 99z, 89z, 96z, and98z, respectively. Recently, several studies have reported diagnostic criteria using a combination of morphologic and kinetic characteristics An interpretation model using such a combination was ˆrst reported by Kinkel and colleagues, 9 in which lesions with smooth margin were classiˆed as benign and lesions with spiculated margin were classiˆed as malignant; lesions with lobulated or irregular margins were dišerentiated only on the basis of washout information. The most important dišerence between our interpretation model and that proposed by Kinkel is the order in which the lesion shapewmargin and the washout pattern were assessed. Another dišerence is that internal heterogeneity is evaluated in our diagnostic criteria. From comparing these interpretation models, it may be suggested that the dišerences in characteristics between benign and malignant lesions may depend on not only the margin characteristics and kinetic information, but also the internal heterogeneity as assessed using high spatial resolution images. Using our proposed interpretation model, we found one false-negative lesion. This case was eventually diagnosed as a mucinous carcinoma and Vol. 5 No. 1, 2006

8 14 M. Tozaki et al. exhibited a very high signal intensity and no internal septations on T2-weighted images. MR imaging References ˆndings for mucinous carcinomas include persistent dynamic curves and very high signal intensity on T 2-weightedimagescomparedwithˆndingsfor other histological types of invasive ductal carcinoma. 26 However, because most lesions with high signal intensity on T2-weighted images are benign, the ˆnding of hyperintensity was treated as a benign sign according to the scoring system proposed by Fischer (hyperintensity in T2; -3 points). 27 In contrast, a mass with lobulated margin, internal septations, and very high signal intensity on T 2- weighted images may re ect the intrinsic growth pattern of ˆbroadenomas. 28 Thus, during categorization using data from dynamic contrastenhanced MR imaging, it must be pointed out that despite apparent benign characteristics, a strongly hyperintense lesion on T2-weighted images may represent a mucinous carcinoma. 11 To our knowledge, this is the ˆrst report of an interpretation model based on the results of the positive and negative predictive values of the BI- RADS } -MR descriptors. However, our study had several limitations. First, only a relatively small group of 45 histologically proved benign lesions was evaluated. Second, interobserver variability was not evaluated in this study. Kinkel and associates 9 reported that intra- and interobserver variability ranges from moderate to fair in evaluation of internal enhancement of lesions. The evaluation of this intra- and interobserver agreement is crucial in the preparation of an interpretation model. Therefore, further investigations and evaluations are needed to reˆne our interpretation model. Conclusion In conclusion, a combination of morphologic criteria, particularly lesion shapewmargin and internal heterogeneity, and kinetic information is useful for dišerentiating benign and malignant lesions. The characteristics with the highest diagnostic accuracy for malignancy were spiculated margin (100z) and heterogeneous enhancement following washout in the group with smooth shapewmargins (100z). In the group having irregular shapew margins, unnecessary biopsy may be avoided for lesions showing homogeneous internal enhancement and persistent pattern. 1. Kaiser WA, Zeitler E. MR imaging of the breast: fast imaging sequences with and without Gd- DTPA. Preliminary observations. Radiology 1989; 170: Harms SE, Flamig DP, Hesley KL, et al. MR imaging of the breast with rotating delivery of excitation oš resonance: clinical experience with pathologic correlation. Radiology 1993; 187: BoetesC,BarentszJO,MusRD,etal.MRcharacterization of suspicious breast lesions with a gadolinium-enhanced TurboFLASH subtraction technique. Radiology 1994; 193: Heywang-Kobrunner SH, Viehweg P, Heinig A, Kuchler C. Contrast-enhanced MRI of the breast: accuracy, value, controversies, solutions. Eur J Radiol 1997; 24: Orel SG, Schnall MD. MR imaging of the breast for the detection, diagnosis, and staging of breast cancer. Radiology 2001; 220: Sherif H, Mahfouz AE, Oellinger H, et al. Peripheral washout sign on contrast-enhanced MR images of the breast. Radiology 1997; 205: Kuhl CK, Mielcareck P, Klaschik S, et al. Dynamic breast MR imaging: are signal intensity time course data useful for dišerential diagnosis of enhancing lesions? Radiology 1999; 211: Bluemke DA, Gatsonis CA, Chen MH, et al. Magnetic resonance imaging of the breast prior to biopsy. JAMA 2004; 292: 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: 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: Tozaki M, Igarashi T, Matsushima S, Fukuda K. High-spatial-resolution MR imaging of focal breast masses: interpretation model based on kinetic and morphological parameters. Radiat Med 2005; 23: American College of Radiology. Breast imaging reporting and data system} (BI-RADS}), fourth ed. Reston, VA: American College of Radiology, Tozaki M, Fukuda K. Supine MR mammography using VIBE with parallel acquisition technique for the planning of breast-conserving surgery: clinical feasibility. Breast 2006; 15: Orel SG, Schnall MD, LiVolsi VA, Troupin RH. Suspicious breast lesions: MR imaging with radiologic-pathologic correlation. Radiology 1994; 190: Nunes LW, Schnall MD, Orel SG, et al. Breast MR Magnetic Resonance in Medical Sciences

9 BI-RADS}-MRI Descriptors for Breast Masses 15 imaging: interpretation model. Radiology 1997; 202: Nunes LW, Schnall MD, Orel SG. Update of breast MR imaging architectural interpretation model. Radiology 2001; 219: Kim SJ, Morris EA, Liberman L, et al. Observer variability and applicability of BI-RADS terminology for breast MR imaging: invasive carcinomas as focal masses. AJR Am J Roentgenol 2001; 177: Liberman L, Morris EA, Lee MJ, et al. Breast lesions detected on MR imaging: features and positive predictive value. AJR Am J Roentgenol 2002; 179: Rofsky NM, Lee VS, Laub G, et al. Abdominal MR imaging with a volumetric interpolated breathhold examination. Radiology 1999; 212: Tozaki M. Interpretation of breast MRI: correlation of kinetic and morphological parameters with pathological ˆndings. Magn Reson Med Sci 2004; 3: Stavros AT, Thickman D, Rapp CL, Dennis MA, Parker SH, Sisney GA. Solid breast nodules: use of sonography to distinguish between benign and malignant lesions. Radiology 1995; 196: 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: Wedegartner U, Bick U, Wortler K, Rummeny E, Bongartz G. DiŠerentiation between benign and malignant ˆndings on MR-mammography: usefulness of morphological criteria. Eur Radiol 2001; 11: BuaduLD,MurakamiJ,MurayamaS,etal.Patterns of peripheral enhancement in breast masses: correlation of ˆndings on contrast medium enhanced MRI with histologic features and tumor angiogenesis. J Comput Assist Tomogr 1997; 21: Fischer U, Kopka L, Grabbe E. Breast carcinoma: ešect of preoperative contrast-enhanced MR imaging on the therapeutic approach. Radiology 1999; 213: Kawashima M, Tamaki Y, Nonaka T, et al. MR imaging of mucinous carcinoma of the breast. AJR Am J Roentgenol 2002; 179: Fischer DR, Wurdinger S, Boettcher J, Malich A, Kaiser WA. Further signs in the evaluation of magnetic resonance mammography: a retrospective study. Invest Radiol 2005; 40: Hochman MG, Orel SG, Powell CM, Schnall MD, Reynolds CA, White LN. Fibroadenomas: MR imaging appearances with radiologic-histopathologic correlation. Radiology 1997; 204: Vol. 5 No. 1, 2006