To evaluate the feasibility of magnetic resonance imaging (MRI)-guided preoperative needle localization (PNL) of breast lesions previously sampled by MRI-guided vacuum-assisted core needle biopsy (VACNB) without marker placement.
To evaluate the feasibility of magnetic resonance imaging (MRI)-guided preoperative needle localization (PNL) of breast lesions previously sampled by MRI-guided vacuum-assisted core needle biopsy (VACNB) without marker placement.
We reviewed 15 women with 16 breast lesions undergoing MRI-guided VACNB without marker placement who subsequently underwent MRI-guided PNL, both on an open 0.5T magnet using freehand techniques. Mammograms and specimen radiographs were rated for lesion visibility; MRI images were rated for lesion visibility and hematoma formation. Imaging findings were correlated with pathology.
The average prebiopsy lesion size was 16 mm (range 4–50 mm) with 13/16 lesions located in mammographically dense breasts. Eight hematomas formed during VACNB (average size 13 mm, range 8–19 mm). PNL was performed for VACNB pathologies of cancer (5), high-risk lesions (5), or benign but discordant findings (6) at 2–78 days following VACNB. PNL targeted the lesion (2), hematoma (4), or surrounding breast architecture (10). Wire placement was successful in all 16 lesions. Final pathology showed six cancers, five high-risk lesions, and five benign findings.
MRI-guided PNL is successful in removing lesions that have previously undergone VACNB without marker placement by targeting the residual lesion, hematoma, or surrounding breast architecture, even in mammographically dense breasts. J. Magn. Reson. Imaging 2010. © 2010 Wiley-Liss, Inc.
IN STEREOTACTIC CORE NEEDLE BIOPSY, radio-opaque markers are placed to identify the biopsy site for subsequent needle localization or follow-up (1–4). Magnetic resonance imaging (MRI)-guided vacuum-assisted core needle biopsy (VACNB) samples suspicious lesions detectable only by MRI, or most clearly seen on MRI, and has similarly evolved to include marker placement. Some facilities use this marker for preoperative needle localization under x-ray guidance to localize the MRI site for subsequent excisional biopsy (5–7).
In our facility, lesions undergoing preoperative needle localization (PNL) after MRI-guided VACNB are all localized by MRI guidance for surgery. Unlike x-ray mammography, MRI provides cross-sectional images that reveal unique 3D morphologic and architectural features of breast lesions within their surrounding mix of fibroglandular and fatty breast parenchyma. We hypothesized that the ability to target these unenhanced features, confirmed by subsequent contrast enhancement, would enable accurate localization of sites that had previously undergone MRI-guided VACNB without marker placement.
In this series we report our unique experience with PNL of lesions or their surrounding breast architecture after MRI-guided VACNB. Our techniques may be helpful for facilities performing preoperative MRI-guided localization in patients without post-MRI-guided VACNB marker placement or in patients with post-deployment marker migration.
The breast interventions are performed using a freehand technique (8–10). One unique aspect of our approach is that we usually begin our initial needle placement before giving intravenous gadolinium contrast material. We then give the contrast material after the needle is in the breast. This approach reserves peak enhancement at the target for the most critical part of the procedure, the final needle placement.
We use careful comparison of the breast architecture on unenhanced T1- (and occasionally T2-) weighted MR images with prior imaging to guide initial needle placement. This method serendipitously allowed us to evaluate the accuracy of breast needle placement without contrast. We present these results since they may be relevant to cases where the previous VACNB removed all enhancement, or in patients with nonenhancing lesions requiring excision.
The current study was performed with the approval of our Institutional Panel on Human Subjects. We retrospectively reviewed 117 MRI-guided PNLs performed at our facility between July 2003 and September 2004. Fifteen women (mean age 49, range 36–68) with 16 MRI-detected lesions fulfilled the inclusion criteria. To be included patients had to have undergone both MRI-guided VACNB and subsequent MRI-guided PNL during the study period. All women underwent diagnostic MRI, MRI-guided VACNB, MRI-guided PNL, and craniocaudal and mediolateral digital mammograms immediately after MRI-guided PNL at our facility. For 14/16 lesions x-ray specimen radiography was also performed.
Diagnostic MRI was performed on a 1.5T closed magnet (GE Medical Systems, Milwaukee, WI) with a dedicated breast coil (MRI Devices, Waukesha, WI) using sequences previously described (11,12). These include noncontrast T1-weighted axial whole-body imaging, sagittal T2-weighted imaging, precontrast high-resolution 3DSSMT imaging, interleaved spiral dynamic scanning during and after a 1 mmol/kg intravenous bolus of gadolinium DTPA with a 20-cc saline flush, postcontrast high-resolution centric-encoded 3DSSMT imaging, and repeated spiral dynamic imaging during the washout phase (12). Images were reviewed on a GE-Advantage Windows Workstation (Functool, software v. 2.6.0, GEMS, Milwaukee, WI).
The freehand method of MRI-guided breast interventions (8–10) differs from other facilities that use a grid and contrast enhancement prior to needle placement (6,7,13–18). This method uses noncontrast T1-weighted scans to identify unenhanced targets in the noncompressed breast without a grid (8,9). We reformat the diagnostic 3DSSMT contrast-enhanced MRI scan to show the enhancing target in the axial plane (Fig. 1). This is then compared to the noncontrast axial T1-weighted images to identify the lesion on the T1-weighted image, which is usually low in T1 signal intensity compared to fat (Fig. 2). If the unenhanced lesion is obscured by dense fibroglandular tissue, then its location is determined by the appearance of the surrounding breast architecture. An important advantage of the freehand technique is that no specialized grid devices are needed, allowing for lesion localization throughout the breast, including lesions near the chest wall, in the axillary region, and anterior in the breast (9,10).
MRI-guided VACNB is performed in a vertically open 0.5T magnet (General Electric) with an open breast coil (MRI Devices) using a method that has previously been described (9) and modified for the 9-gauge ATEC Breast Biopsy and Excision System (Suros Surgical Systems, Indianapolis, IN). All pulse sequences used were specifically designed for breast imaging on the open 0.5T magnet (9). The ATEC system contains a vacuum-assisted biopsy probe, a plastic coaxial introducer sheath, a sharp nonferrous trocar, and a plastic obturator. The VACNB biopsy method is similar to those previously reported in closed magnets (5,6) with a few important differences. First, because the open magnet allows the physician direct access to the patient, the entire procedure including targeting, needle adjustment, and VACNB is done at one setting without sliding the patient in and out of the magnet bore. Second, the entire procedure is done in the noncompressed breast without the use of a grid. Third, the skin entry site selection, skin preparation and anesthesia, and initial sheath insertion are all done before contrast enhancement. The final steps of final sheath positioning, determination of sampling direction, and biopsy are done after intravenous administration of gadolinium and water-specific T1-weighted imaging.
For VACNB, we first estimate the location of the lesion within the breast and place a fiducial marker on the overlying skin. We then perform an initial noncontrast T1-weighted spin echo sequence to identify the lesion or surrounding breast architecture via correlation with the prior diagnostic MR images. This allows us to choose a suitable skin entry site prior to contrast enhancement and mark it by measuring the distance from the fiducial in the superior–inferior and anterior–posterior direction. We sterilize the site with commercially available povidone-iodine, achieve local skin anesthesia using 1% lidocaine buffered with 8% sodium bicarbonate at a 10:1 ratio, and make a small skin incision with an MRI-compatible scalpel.
Using sterile technique and the freehand method, we insert the sharp metal trocar/sheath parallel to the scan plane so that the trocar is proximal to the lesion. To prevent air from entering the breast, we attach suction to a 3-way stopcock on the sheath. To image, we replace the metal trocar with a plastic obturator and perform a noncontrast T1-weighted axial spin echo sequence (Fig. 3). If the obturator is short of the lesion, we replace the trocar, advance incrementally to the lesion, and reimage. The obturator tip should optimally be in or adjacent to the lesion for accurate sampling. Once the obturator tip is in position we enhance the lesion using an intravenous 0.1 mmol/kg GdDTPA bolus followed by a saline flush and perform axial and sagittal 3-point Dixon spoiled gradient-recalled echo (SPGR) scans for confirmation of correct needle placement (Fig. 4) (8,19). If there is any discrepancy in targeting that will not be manageable by appropriately directing the biopsy instrument, the sheath is repositioned and confirmatory scanning is repeated. We anesthetize the deep breast tissue with Marcaine (0.75% Bupivicaine) using a long 22G Chiba needle inserted through the sheath, place the biopsy probe into the sheath, and obtain between three and nine 9G samples. We then remove the probe and achieve hemostasis using direct pressure. At the time of this study, VACNB was a very new technique and marker clips were not yet placed. The average time for the procedure is between 60 and 90 minutes.
In all cases involving VACNB a radiologist correlates the pathology to the MRI findings. We recommend excisional biopsy for cancer, high-risk lesions (lobular carcinoma in situ, atypical ductal hyperplasia, atypical lobular hyperplasia, other atypia of any kind, papillary lesion, radial scar), and discordant pathology findings.
In our facility we perform PNLs in the open MRI magnet with a 20 gauge MRI-compatible needle (E-Z-EM, Westbury, NY) and methylene blue dye using a previously described method (9,19). A similar freehand technique was also described on a 3.0T closed magnet by Meeuwis et al (20). Briefly, we use noncontrast axial T1-weighted scans to target the residual unenhanced lesion or its estimated location in the surrounding breast tissue using the same freehand method used for VACNB described above (Fig. 5). Following skin sterilization and anesthesia, an MRI-compatible localization needle is advanced to the lesion with its axis parallel to the scan plane (Fig. 6). Once the needle is adjacent to the lesion we enhance the lesion as for VACNB and perform 3-point Dixon SPGR scans for confirmation of correct needle placement. If the lesion is not visible we determine if the architecture looks correct. If it appears correct we continue with wire localization under the assumption that the lesion no longer enhances, the lesion was mostly removed by VACNB, or the needle artifact is obscuring it. We do not abort the procedure unless the target is invisible and the architecture suggests incorrect localization. After we confirm correct needle placement we inject 0.2 mL sterile methylene blue dye mixed with 1% lidocaine, deploy the hookwire, and remove the needle. Axial and sagittal 3-point Dixon SPGR scans are obtained to show the relationship of the hookwire tip to the lesion. Images are annotated as necessary to provide any localizing information we feel may be beneficial to the surgical team, such as the exact location of the primary lesion in relation to the wire (eg, at wire stiffener, tip, etc.).
In addition to MR images we obtain craniocaudal and mediolateral mammograms with the MRI-guided wire in place using analog (GE DMR, General Electric Medical Systems) (10 lesions, until March 2004) or digital technique (Senograph 2000D, General Electric Medical Systems) (six lesions, after March 2004) and a skin marker at the entry site. This facilitates subsequent surgical planning by our breast surgeons who are more familiar with radiographs than MRI. Since some MRI findings have no mammographic correlate, we estimate the expected mammographic location of the lesion along the wire based on the location of the enhancing finding as seen on the MRI. We evaluate the breast tissue architecture and any mammographic findings on the postwire localization mammograms. We annotate the hardcopy mammogram films showing the estimated location of the lesion based on the MRI and the location of the hookwire tip, and send the annotated mammograms and MRI films to the operating room.
We obtain x-ray specimen radiographs in all cases unless the surgeon specifies otherwise and compare them with the postlocalization mammograms. Similar to mammographic or ultrasound-guided specimen radiograph results, we notify the surgeon in the operating room, reporting inclusion of the hookwire, hookwire tip, excised masses or calcifications (if any), presence of radiologic findings at or near the specimen edge (if any), and whether the tissue containing the lesion has been removed. Since it is not possible to obtain MR images of the surgical specimen with contrast enhancement, and since most of these lesions are not visible using x-ray techniques, the surgical team is reminded that histologic evaluation of the specimen will be necessary before knowing with certainty if the primary target has been fully excised.
For this study the diagnostic MRI was reviewed by two radiologists and each lesion was measured and categorized according to American College of Radiology (ACR) BI-RADS terminology for lesion morphology and enhancement kinetics (21). Two radiologists reviewed the VACNB MRI images to determine if there was hematoma formation during the procedure and the size of the hematoma if present. Of note, patients were not routinely imaged immediately after VACNB to determine the presence or absence of a hematoma, and the MRI findings during the procedure were reviewed to make the determination if a hematoma had formed. MRI needle localization images were reviewed for residual lesion visibility, residual hematoma, and to determine if targeting was based on enhancement, architecture, or hematoma.
Available mammograms were reviewed to determine breast density according to ACR BI-RADS criteria, findings in the expected location of the lesion seen on MRI, and inclusion of the hookwire, hookwire tip, and any mammographic findings on specimen radiographs. If the mammograms were not available, the findings were recorded from the mammographic reports (two cases). Two radiologists compared the pathology from the VACNB, pathology from the needle localization biopsy, and imaging findings.
Imaging follow-up for patients varied according to their physicians' preferences, ie, mammography or MRI. At the time of these early MRI-guided interventions, no specific imaging follow-up was recommended by the radiologist. Thus, the patient records were reviewed to determine what imaging was done and the patient disposition at the end of the study.
Fifteen women with 16 lesions fulfilled the inclusion criteria of the study. The mean age was 49 years (range 36–68 years). Reasons for MRI referral in the 16 lesions included evaluation for contralateral breast cancer (4/16, 25%), staging for ipsilateral breast cancer (4/16, 25%), high-risk screening (2/16, 13%), abnormal mammogram (4/16, 25%), or abnormal MRI at another facility (2/16, 13%). The 16 lesions had an average size of 16 mm (range 4–50 mm) on MRI. They were composed of foci (1/16, 6%), masses (6/16, 38%), and nonmass-like areas of enhancement (9/16, 56%). Kinetic enhancement curves showed initial rapid enhancement in 15 of the 16 lesions, with late persistent enhancement in 5/16 (31%), late plateau in 7/16 (44%), washout in 2/16 (13%), or no late enhancement curves (due to technical problems) in 1/16 (6%) lesions; the sixteenth lesion (1/16, 6%) showed gradual initial enhancement with a late plateau phase.
Of the 16 lesions, 14 had mammographic studies that were available for review. In these 14 cases, breast density was estimated by mammogram review. Breast density was determined by the mammography report in the remaining two. The mammographic density of the breast tissue near the target lesion was scattered fibroglandular and fatty (2/16, 13%), heterogeneously dense (1/16, 6%), or dense (13/16, 81%).
On the unenhanced T1 axial MR images, initial VACNB probe placement was directed by visualization of the lesion in five cases (31%) and surrounding breast architecture in the remaining 11 cases (69%). Eight patients (50%) demonstrated hematoma formation during the procedure with an average size of 13 mm (range 8–19 mm), as measured during or immediately after the MRI-guided biopsy.
Pathology from VACNB showed five cancers including two ductal carcinomas in situ (DCIS), two invasive ductal carcinomas (IDC) with (one) or without (one) DCIS, and one invasive tubular carcinoma (ITC) with DCIS. There were five high-risk lesions including three papillary lesions (one with atypia), one flat epithelial atypia, and one atypical ductal hyperplasia (ADH); and six benign findings including two nonproliferative fibrocystic change (NPFCC), one adenosis with lactational change, one fibrocystic change (FCC), one fibroadipose tissue, and one benign adipose tissue. The six benign findings were recommended for excisional biopsy because of discordance with suspicious MRI findings.
PNL took place 2 to 78 days following VACNB, with an average of 31 days. As with all nonenhanced freehand needle localizations, we placed the needle tip in the expected location of the enhancing target. We accomplished this by using unenhanced findings surrounding the target, including the lesion itself, the surrounding breast architecture or the hematoma, if present, and corresponded to the expected location of the target. On the noncontrast T1-weighted axial images, initial needle placement was directed by visualization of the lesion in two cases (12%), hematoma at the site of prior biopsy in four cases (25%), and surrounding breast architecture in the remaining 10 cases (63%). Blood can dissect along the core needle biopsy track or may accumulate in the most pliable locations of the breast tissue. Thus, hematomas may form at locations other than that of the target. Therefore, the hematomas were used only as guides to the general location of the target, and only if the surrounding architecture corroborated the hematoma's formation at the target's expected location. Fifty percent (4/8) of the hematomas had resolved by the time of needle localization. Following needle placement and intravenous administration of gadolinium, all lesions demonstrated residual enhancing targets. Wire placement was successful in all 16 lesions, with accurate wire placement demonstrated on axial and sagittal images.
All 16 lesions had post-MRI needle localization mammograms, with 14 mammograms available for review and two had mammographic reports for review. Fourteen postlocalization mammograms demonstrated only normal glandular tissue at the expected location of the MRI finding; one demonstrated pleomorphic calcifications, and the remaining one demonstrated a mass with pleomorphic calcifications. In these last two cases, MRI localization was performed in the first case for a specific enhancing lesion within multiple clusters of calcifications and in the second case to bracket DCIS surrounding an invasive ductal cancer. In the latter case the extent of abnormality suggesting DCIS on the MRI was greater than the extent suggested by the calcifications on the mammogram.
Following surgical excision, 14 of the 16 lesions had specimen radiographs demonstrating normal glandular tissue in 11, pleomorphic calcifications in two, and a mass with calcifications in one lesion. In all cases, the hookwire, hookwire tip, and mammographic findings at the expected location of the MRI findings were included in the specimens.
Surgical pathology showed six cancers (three IDC/DCIS, one DCIS, one ITC with DCIS, one IDC and atypical lobular hyperplasia (ALH)), five high-risk lesions (three papillary lesions with ADH or atypia, one focal atypia in NPFCC, one ALH), and five benign findings (two PFCC, one adenosis with lactational change, one fibroadenoma, one lipoma with NPFCC) (Table 1).
|Needle localization results||Max. diameter (mm)||Distance of wire tip to lesion (mm)a||Type of imaging follow-up||Follow-up (months)||Disposition at last follow-up|
|Invasive tubular cancer with DCIS||Invasive tubular cancer with DCIS||6||2||No imaging follow-up||0||Returned to outside institution|
|Invasive ductal cancer||Invasive ductal cancer/DCIS||9||3||MRI (ipsilateral)||27||Benign ipsilateral MRI|
|Invasive ductal cancer/ DCIS||Invasive ductal cancer/DCIS||14||2||MRI (ipsilateral)||31||Ipsilateral MRI-guided VACNB showed recurrent invasive ductal cancer/DCIS: mastectomy|
|DCIS||DCIS||25||2 and 4b||Mammogram||12||Returned to outside institution|
|DCIS (left)*||Invasive ductal cancer/DCIS (Left)*||50||5||N/A||N/A||Mastectomy within a month|
|VACNB high risk|
|Flat epithelial atypia||ALH||4||4||MRI (ipsilateral)||8 and 32||Ipsilateral MRI-guided PNL 2x repeated after new areas of enhancement on surveillance MRI, yielding first ALH and then fat necrosis|
|Focal ADH within PFCC||PFCC||7||2||No imaging follow-up||0||Returned to outside institution|
|Papillary lesion||Papillary lesion + ADH||8||3||No imaging follow-up||0||Returned to outside institution|
|Papillary lesion||Papillary lesion + ADH||11||4||Mammogram||14||Benign mammogram|
|Papillary lesion with atypia||Papillary lesion + focal atypia||24||2||Mammogram||47||Benign mammogram|
|Adenosis and lactational change (right)**||Adenosis and lactational change (Right)**||6||4||Mammogram and MRI (ipsilateral)||50 MRI||Benign ipsilateral mammogram and MRI (prior mastectomy left breast*)|
|NPFCC||Fibroadenoma||6||3||Mammogram||52||Benign ipsilateral mammogram; Contralateral MRI-detected phyllodes tumor: mastectomy|
|NPFCC with florid apocrine change||Lipoma + NPFCC||6||4||Mammogram||53||Benign ipsilateral mammogram; Contralateral MRI-guided VACNB: benign NPFCC|
|Fibroadipose tissue||Focal atypia in NPFCC||56||2||Mammogram and MRI (ipsilateral)||52 Mammogram||Benign ipsilateral mammogram and MRI|
|Benign adipose tissue||Invasive ductal cancer and ALH||9||3||No imaging follow-up||0||Returned to outside institution|
Correlation of the pathology from VACNB and PNL showed that all cancers found at VACNB were confirmed by needle localization. One DCIS at VACNB showed IDC/DCIS on needle localization; this patient had a palpable IDC diagnosed by ultrasound-guided core biopsy that was included in the bracketed needle localization specimen (Table 1). Two of five VACNB high-risk pathologies correlated with needle localization pathology (one ALH, one papillary lesion with atypia), two papillary lesions at VACNB had ADH at needle localization, and the fifth case was downstaged from ADH to PFCC. In the six benign VACNB pathologies, needle localization showed invasive cancer in one lesion that we suspected we had missed at VACNB; atypia in a case where we felt that the VACNB pathology of benign fibroadipose tissue was discordant with imaging findings; and one fibroadenoma. There was concordance in the remaining three cases (Table 1).
Six of the 15 patients (40%) had cancer at PNL. Three of the six patients returned to their outside institutions for follow-up. One patient had a mastectomy within a month of PNL. The other two patients underwent follow-up MRI of the ipsilateral breast. One of these two patients had new MRI enhancement at the biopsy site 31 months after initial excision of IDC/DCIS. She had MRI-guided VACNB of the new enhancement, showing recurrent IDC with DCIS, and went on to mastectomy. The other patient had benign MRI studies at 27 months after PNL.
Ten patients had high-risk (n = 6) or benign findings (n = 4) at PNL (this includes one patient who had cancer on the contralateral side).
Of the six high-risk patients, two returned to their outside institutions for follow-up, two had mammograms only with benign findings at 14 and 47 months after PNL, and one had a benign mammogram and MRI at 52 and 56 months, respectively. The last high-risk patient with ALH had new abnormal enhancement on MRI at 8 months after PNL. Surgical excision of the new enhancement showed ALH. At 32 months, MRI showed abnormal enhancement at the biopsy site and subsequent MRI-guided PNL showed scarring and fat necrosis.
All of the four patients with benign findings at PNL had benign mammograms of the ipsilateral breast at an average of 45 months. One patient also had a benign MRI of the ipsilateral breast at 50 months. Two patients had no MRI follow-up of the ipsilateral breast but had abnormal MRIs of the contralateral breast. In the first of these two patients, MRI found a mass at 12 months which was excised by MRI-guided PNL. The mass was a phyllodes tumor and the patient underwent mastectomy. In the second patient, MRI found abnormal enhancement at 7 months which was sampled by MRI-guided VACNB showing NPFCC.
Our study shows that initial unenhanced targeting for PNL under MRI guidance without a marker in a VACNB site is both feasible and successful in removing the lesion, even in dense breast tissue. Breast architecture in axial or sagittal views can guide needle localization if the lesion is not seen on the unenhanced scans.
This report discusses the formation of hematomas and timing of their resolution after MRI-guided VACNB. Previously, Smith et al (22) reported on the feasibility of injecting a patient's own blood at the lesion site to direct excisional biopsy under intraoperative ultrasound guidance. The same research group described their experience using intraoperative ultrasound to localize the naturally occurring hematoma after vacuum-assisted breast biopsy. The hematoma and lesion were excised and surgical margins were compared to excisional biopsy specimens obtained after conventional needle localization. In this retrospective study the hematoma-directed localization procedure showed fewer positive margins compared to the conventional needle localization procedure (23). In our series, it was shown that although at least 50% (8/16) of cases formed small hematomas during or immediately following VACNB ranging from 8–19 mm in size, only four of these hematomas persisted by the time of needle localization. The residual hematomas did not pose a problem during PNL, sometimes serving as guides to the general location of the target. On the other hand, 50% of the hematomas resolved by the time of needle localization (which occurred at most 78 days following VACNB). Thus, a hematoma cannot be expected to serve as a reliable “marker” for needle localization. Furthermore, hematomas may form from transected blood vessels anywhere along the course of instrumentation or may track along the biopsy path to accumulate in the most pliable tissue (ie, adipose tissue), which may be far from the target.
Our study confirms the work of others that suggests that correlation of VACNB pathology and MRI findings is important to avoid missed cancers (5–7). Similar to other imaging-guided vacuum assisted biopsy methods, VACNB did not accurately sample one cancer despite our best efforts. Imaging/pathologic correlation and an honest estimation of whether the lesion was sampled at VACNB allowed us to proceed to needle localization for the missed cancer. On the other hand, the finding of invasive cancer after VACNB showing DCIS is not surprising, as this is well documented in the stereotactic core biopsy literature (24,25). Similarly, the finding of PFCC after VACNB showing ADH has also been described in the stereotactic core biopsy literature (26).
Our core biopsy method differs from others working in an open, vertically oriented magnet (27) in that our technique uses vacuum assistance and targeting of unenhanced areas. In comparison to both closed and open magnet biopsy methods using initial contrast enhancement for targeting, our method of PNL using initial unenhanced targets could be of use if postbiopsy marker placement was not performed and excision was required. A second scenario in which our technique may be helpful is if the target does not enhance and does not contain a marker, as might be the case if the finding was completely removed by VACNB without marker placement, or if the abnormal MRI finding was an architectural distortion that enhanced identically to the rest of the breast. Third, our method would be valuable if the marker has migrated from its original location away from the biopsy site, a situation that has been described to occur in about 10%–20% of cases (28,29). In all of the three aforementioned scenarios the surrounding breast architecture could be targeted for excision using our method of PNL.
Limitations of our study include a limited assessment of whether MRI-detected lesions are removed at the time of surgery. Since the contrast enhancement persists for only a short time, specimen MRI is useless. Our main method of determining lesion removal has been use of postlocalization mammogram and specimen radiography with identification of breast tissue similar to that seen along the wire on the postlocalization mammogram, and confirmation of removal of the hookwire and hookwire tip on the specimen radiograph.
Our use of post-MRI-guided needle localization mammograms is an important addition to the procedure. First, post-MRI-guided needle localization mammograms demonstrate the course of the wire through the breast and show the final location of the hookwire tip, thereby guiding the surgeon to the optimal location for the initial surgical incision on images where the wire is projected over the full thickness of the breast. The mammograms are marked for the surgeon based on the location of the target lesion with respect to the stiffener and hookwire tip on the post-MR procedure images. Second, the mammogram shows the appearance of the breast tissue along the path of the hookwire and in the expected location of the MRI finding. In a series by Erguvan-Dogan et al (30), 82% of MRI-guided needle localized lesions were seen by x-ray specimen mammography and all invasive ductal cancers were identified as masses on the specimen radiograph. Unlike in Erguvan-Dogan et al's series, the majority of lesions in our study demonstrated only normal glandular tissue on the specimen radiograph. However, the breast tissue surrounding the wire in the expected location of the MRI-detected lesion was identified on all specimen radiographs. In addition, two pleiomorphic calcification clusters and one mass with calcifications seen near the MRI wire also were removed. Correlation of the specimen radiographs with the tissue around the wire could only be accomplished because post-MRI-guided needle localization mammograms were obtained. Therefore, obtaining post-MRI needle localization mammograms and x-ray specimen radiography was essential to make sure the tissue around the wire is removed.
Follow-up MRI is an important step in ensuring that MRI-detected lesions are adequately sampled or removed. Although we recommend a diagnostic MRI 6 months following an MRI-guided VACNB or PNL, compliance with follow-up MRI results is variable. The reasons for follow-up MRI in only one-third of our patients are uncertain. Possible explanations include high MRI costs, lack of physician referral for follow-up MRI, and lack of patient motivation to continue MRI. Compliance with post-MRI biopsy imaging is a potential area for further research which may improve our ability to assess adequacy of MRI-guided procedures.
In conclusion, freehand MRI-guided PNL in an open magnet is a feasible and successful technique after MRI-guided vacuum-assisted core biopsy without marker placement. The use of x-ray mammography after MRI-guided preoperative needle localization is useful to confirm that the tissue containing the MRI-detected finding has been removed.