The institutional review board approved this study and did not require the approval of patients nor their informed consent for review of their images and records. From January 2002 to July 2004, 303 women underwent breast MRI examinations were initially considered for this study. The primary inclusion criterion of this retrospective analysis was a preoperative MRI in patients with histologically confirmed breast cancer. Among the 303 women, 213 patients had confirmed breast cancer. Of these patients, 61 who underwent neoadjuvant chemotherapy were excluded because of a possible chemotherapeutic effect causing different kinetics and morphology of suspicious lesions (5). Three patients who did not undergo surgery in our institution were also excluded. The remaining 149 patients were included in this study.

The average age of the patients in the present study was 50 years (range 23 to 79 years). Indications for an MR examination included preoperative staging before planned breast conserving therapy to exclude multicentricity in patients who had one mammographic or sonographic suspicious lesion (n = 123), a search for an occult tumor with axillary metastases of suspected breast origin (n = 5) and a postoperative examination to rule out residual disease (n = 21).

A review of the MR images and final reports by an experienced radiologist specialized in breast imaging was conducted. We searched the cases with additional suspicious breast lesions that were initially detected by MRI. By the definition of Liberman et al. (6), MR lesions were considered to be additional sites if they were located in a different breast quadrant than the index cancer, if they were in the same quadrant but were separated from the index cancer by at least 1.0 cm of intervening normal-appearing tissue on MRI, or if they were in the same quadrant and contiguous with the index cancer but extended at least 4.0 cm beyond the site of the index cancer. Typically benign appearing enhancing lesions, i.e. those with a circumscribed margin or delayed enhancement were excluded.

If MRI revealed additional suspicious breast lesions other than the index cancer, a targeted US performed by the radiologist who interpreted the MR images was recommended to look for a US correlates amenable to further biopsy or localization. Although there were no standardized protocols to follow, a targeted US examination by the radiologists who had knowledge of the MRI findings was performed for the clinical and mammographical occult lesions. When targeted US was performed, the additional suspicious breast lesions were classified into two groups based on whether or not they had a US correlates. We investigated their pathological outcomes and changes in treatment for these patients. We compared the rate of identifying additional cancers between the lesions with and without a US correlate using the Fisher's exact test.

The 31 patients that did not receive targeted US consisted of two groups; one of the groups in whom the breast US was performed before the breast MR imaging and did not correlate with any additional detectable lesion by MRI and the other group with mammographic microcalcifications, which were not detected initially and did correlate with the MRI lesions. These lesions were confirmed by surgical excision or were followed with a subsequent examination.

Breast MRI was performed with the patient prone in a 1.5-T commercially available imager (Signa; GE Medical System, Milwaukee, WI) with the use of a dedicated surface breast coil. The MRI sequence used in this study included a fat suppressed axial fast spin echo T2-weighted sequence (4000/120, repetition time msec/echo time msec) and fat-suppressed unilateral sagittal dynamic imaging. Contralateral sagittal dynamic imaging was obtained on the next day. Dynamic imaging was performed with a T1-weighted three-dimensional, fat-suppressed fast spoiled gradient-echo sequence (17.3/1.3 ms; flip angle, 30°) three times (one before and two after a rapid bolus injection of 0.1 mmol/kg gadopentetate dimeglumine [Magnevist; Berlex, Wayne , NJ]) (7). Section thickness was held between 1.0-2.0 mm without an intersection gap with a 256 × 192 matrix, an 18-24 cm field of view, and a scan time of 3-4 minutes. Two sequential post-contrast scans were obtained without a break, beginning immediately after the saline flush. Standard subtraction images were obtained by subtracting the precontrast images from the early peak (or serial) postcontrast image on a pixel-by-pixel basis. Reverse subtraction images were obtained by subtracting the last postcontrast image from the early peak postcontrast image. If the lesion demonstrated a kinetic pattern showing an early rise and an early washout, the lesion would be observed with the remaining high signal intensity on reverse subtraction images. We regarded these lesions as kinetically suspicious lesions. If any kinetic curves were equivocal, the morphologic features were considered together. If the lesion has a spiculate or an irregular margin, we regarded it as a morphological suspicious lesion. If a lesion had morphologic features that indicated it was probably benign but reverse subtraction images demonstrated a high signal intensity lesion, this was also considered to be suspicious (8). Criteria for the features of suspicious non-mass-like enhancements included linear, ductal, and segmental patterns. Maximum intensity projection images were reformed from the subtraction images.

Breast US examinations were performed by one of four breast imaging radiologists using a 7-10 MHz linear transducer (Logiq 700; General Electric, Milwaukee, WI, or HDI 5000; Philips Medical Systems, Bothell, WA). At the time of breast US examination, prior MR images were made available for direct correlation.

Thirty-eight lesions in 31 patients were performed with targeted US. Of 21 patients with correlated suspicious enhancing lesions that were evident on US, eight underwent US-guided core biopsy for the suspicious lesion and 10 underwent US-guided localization, one with US-guided hook-wire localization and the remaining nine with US-guided carbon marking (9). In the residual three patients, two patients had a scheduled mastectomy due to the extensive extent of an index cancer and one had breast conserving surgery because of the lesion localized 1 cm just inferior to the index cancer within the same quadrant.

In the 10 patients without a US correlate, the lesions in six patients could not be localized. Of the four that underwent a US-guided localization, US-guided needle localization was used in one and carbon marking in three. Although lesions are not defined on US with certainty, localizations were performed in five lesions of four patients for subtle sonographic lesions likely consistent with the additional suspicious MR lesions. As the lesions without a US correlate could not be localized, careful histological examinations were requested at the area of any additional MR lesions.

Of the 149 total patients for whom MRI was used for preoperative evaluation of breast cancer, 69 additional suspicious breast lesions were detected by MRI in 62 patients (42%). Of those 69 lesions, 26 (38%) were confirmed as cancers by histology. By MRI, 48 of 69 lesions (70%) revealed a mass, and 21 (30%) demonstrated nonmass-like enhancements and the average diameter of the 69 additional enhancing lesions in these 62 patients was 1.1 cm (range, 0.4-5.0 cm).

In addition, of these 62 patients with 69 lesions, 57 (92%) had a single additional lesion, four (6%) had two lesions, and one (2%) had four lesions. Among 69 lesions, additional cancers were diagnosed in 26 (38%) in 23 patients. Contralateral enhancing lesions were detected in ten patients (16%), of which confirmed cancers in three patients (5%).

Of the 62 patients with additional lesions, 31 patients with 38 lesions underwent targeted US. Histological findings of these 38 lesions in 31 patients performed with targeted US revealed cancers in 18 (47%) lesions, high-risk lesions in two (5%), benign lesions in 15 (39%), and unidentified lesions in three (8%) (Table 1). The latter three lesions that were not confirmed by histological examination were removed by a mastectomy in one patient and breast conservative surgery with a wide extent in two patients.

Of the 38 lesions examined with targeted US, 27 (71%) had a US correlate, which included 15 (56%, 15/27) cancers (Fig. 1), two (7%, 2/27) high-risk lesions (lobular carcinoma in situ and atypical ductal hyperplasia in each one) and ten (37%, 10/27) benign lesions (Fig. 2). Of the 11 lesions without a US correlate, three (27%, 3/11) proved to be cancerous (Fig. 3) and five (45%, 5/11) were benign as determined by mastectomy or breast conserving surgery with additional excision (Fig. 4), however, three (27%, 3/11) were not confirmed by histology. The cancer rate was statistically higher in lesions with a US correlate than in ones without a US correlate (p = 0.028) (Table 2). Of 15 cancers with a US correlate US, ten (67%) were confirmed as invasive ductal carcinoma, and five (33%) as ductal carcinoma in situ. Of the three cancers without a US correlate, two (67%) were ductal carcinoma in situ, and one (33%) as an invasive ductal carcinoma. The difference between invasive and in situ carcinoma was not statistically significant (p = 0.528). Of US correlate lesions identified by histology, the high-risk lesions included lobular carcinoma in situ and atypical ductal hyperplasia for each one. The benign lesions were diagnosed as papilloma in three, fibroadenoma, stromal fibrosis, fibrocystic disease in two each, and columnar cell change in one. Imaging findings and pathological results of US correlate lesions are summarized in Table 3.

Of 31 patients who underwent targeted US, the surgical plan was altered in 27 patients (87%) and not in four patients. Targeted US justified a change in treatment for 22 patients (81%) and misled five patients (19%) into an unnecessary surgical excision.

Of 21 patients with US correlates, the surgical treatment plan for 19 patients was altered preoperatively with the consent of the patient. These 19 patients included 14 with additional cancers, two with high-risk lesions and three with benign lesions (one fibroadenoma and two intraductal papillomas). Two patients with high-risk lesions underwent conserving surgery and a wider excision. Even though the US-guided core biopsy for three patients confirmed benign lesions, these patients underwent conserving surgery and a wider excision, because the surgeon did not want to defer the operation because of a personal concern of the patient or delayed pathological results with the consensus of the patient. Of the ten patients without US correlates, surgical treatment was changed in eight patients. However, three patients for whom cancers were confirmed subsequently underwent mastectomy at the discretion of surgeon (two), and conservative surgery with a wide excision during the operation (one). Five patients underwent unnecessary wide excision due to the difficulty in localizing the additional lesions.

In 13 (76%) of 17 patients (14 with US correlates and 3 without US correlates) in whom MRI detected additional cancers, the surgery was converted to mastectomy in ten or additional contralateral conserving surgery in three.