Introduction

Routine use of full-field digital mammography (FFDM) for cancer screening reduces breast cancer mortality [1, 2]. Prior to the advent of digital breast tomosynthesis (DBT), no other breast cancer screening tool had a better combination of sensitivity and specificity [3]. The past 2 years have seen mounting evidence of the superiority of DBT used in combination with FFDM versus FFDM alone in screening and assessment [4–8]. The combined application allows 3D visual review of thin breast sections, unmasking cancers obscured by normal tissue located above and below breast lesions [9]. The most specific mammographic feature of malignancy is a spiculated soft tissue mass, with nearly 90 % of such lesions representing invasive cancer [10]. Five percent of invasive cancers present as areas of architectural distortion without any obvious mass [10]. Despite the advantages of 3D imaging to highlight and unmask benign and malignant signs, histopathological evaluation is frequently necessary [11]. In the literature, approximately 20 to 40 % of all suspicious lesions undergoing histopathological workup prove to be malignant [12–14]. Various types of small lesions, especially small architectural distortions, are often occult or equivocal on ultrasound and cannot be biopsied under ultrasound guidance [15, 16]. In these cases, lesions detected by FFDM or DBT can be biopsied under stereotactic or magnetic resonance (MR) guidance. Although stereotactic-guided vacuum-assisted biopsy (SVAB) is highly reliable and spares the majority of patients diagnostic open surgery [17–19], it is not well-suited for localization of architectural distortions [20]. SVAB is prone to targeting error because the user may fail to identify the same lesion in the pair of stereotactic images. This failure results in a miscalculation of the depth of the lesion, commonly referred to as the Z value. The magnitude and direction of this error are largely determined by the direction of the targeting error (Fig. 1a–d). This source of error is very problematic due to the high incidence of malignancy of architectural distortions [21, 22]. Tomosynthesis-guided vacuum-assisted biopsy (TVAB) appears to offer better lesion localization by 3D-visualization of low contrast masses and distortions [23, 24]. In this retrospective study, we evaluate the feasibility, clinical performance and effectiveness of TVAB in 148 consecutive patients admitted to our institute for SVAB.

Fig. 1
figure 1

(a) Example of an inaccuracy in SVAB lesion targeting that is unrecognizable on one or both stereotactic views, but manifests as an extreme error in the calculated Z value (depth) (patient not part of study group). Breast tomosynthesis showed a small spiculated mass with subtle architectural distortion in the upper outer quadrant of the right breast. The lesion was occult on ultrasound. (b) a + b. The lesion appears to be easily localizable in the stereotactic images of SVAB. The SVAB targeting appears to be accurate. (c) Histopathology reveals a B2 lesion. Breast tomosynthesis shows the still existing spiculated mass (black arrow) and a clip artefact anteriorly (white arrow). (d) MRI revealed a miscalculation by the SVAB of at least one quadrant. Instead of targeting the lesion in the upper outer quadrant of the right breast, SVAB had targeted an area in the upper inner quadrant (white arrow). MRI demonstrated the lesion still existing in the upper outer quadrant of the right breast (black arrow). Histopathology of the MR-guided biopsy showed the lesion to be an invasive carcinoma

Full size image

Materials and methods

Patients

One hundred and forty-eight consecutive patients (mean age 56.6 years, range 32–85 years) admitted to the Breast Centre between 15 December 2012 and 15 October 2013 (closing date for inclusion) for SVAB diagnostic workup were included in this retrospective study. TVAB was performed in all 148 patients. All TVAB procedures were performed by four radiologists experienced in minimally invasive breast biopsies (PC, PK, MS, CW) and trained by Hologic (Hologic, Inc., Bedford, MA). All 148 patients gave their written informed consent.

All clinical data and results of minimally invasive breast biopsies (MIBB) performed in Switzerland have to be archived in the MIBB (minimally invasive breast biospy) database (www.mibb.ch/). To compare the results of SVAB with the results of the ongoing TVAB minimally invasive biopsies, it was decided to use the complete data set of one year. 2011 was the last year with a complete data set that was evaluated before the implementation of TVAB.

In 2011, 86 patients underwent SVAB at the Breast Centre. All SVAB procedures were performed by the same four radiologists who would conduct the TVAB biopsies in 2012–2013.

Ethics

The local cantonal ethics committee has approved this study for full ethics waiver due to its retrospective and anonymised nature.

Tomosynthesis-guided vacuum-assisted biopsy

The Breast Center has offered DBT exams using the Selenia® Dimensions® 3D mammography system (Selenia®, Hologic, Inc., Bedford, MA) since December 2008, and performed TVAB procedures with an Affirm™ Breast Biopsy Guidance system (Affirm™, Hologic, Inc., Bedford, MA) mounted on the Selenia device since December 2012.

Overview of the TVAB system

In order to execute TVAB procedures, the Affirm™ system is mounted on the Selenia® Dimensions® device’s (FFDM/DBT system) C-arm and locked into place. The Selenia® Dimensions® device provides power and target information to the Affirm™ breast biopsy guidance system, and displays the acquired images. The Affirm™ system is connected to a vacuum device and provides motorized movement to the calculated target in the X and Y axes. The Z axis (depth) is adjusted manually.

The x-ray tube arm on the FFDM/DBT system moves separately from the compression arm to allow acquisition of images for use in target calculation.

Using DBT imaging, 3D tomosynthesis images can be acquired to determine the 3D (X-Y-Z) Cartesian coordinates of the area of concern (AOC).

Reconstructed tomosynthesis 3D volume images of the breast are displayed on the preview screen of the acquisition workstation. The user is able to scroll within the tomosynthesis 3D volume and view the tomosynthesis slices at the AOC, thus determining the Z coordinate (the distance from the AOC to the breast support platform). The user then positions a targeting cursor directly on the AOC within the selected tomosynthesis slice to determine the X and Y coordinates. This identifies the X-Y-Z coordinates representing the AOC within the tomosynthesis 3D breast volume. These X-Y-Z coordinates are the same coordinates that are computed using the stereotactic paired images in a conventional stereotactic procedure.

The acquisition workstation sends the X-Y-Z coordinates of the targeted AOC to the Biopsy Guidance Module (BGM) mounted on the gantry C-arm. The user guides the needle of the Affirm™ device into the breast to reach the AOC as defined by the previously identified X-Y-Z coordinates. The LCD display of the BGM provides numeric feedback on the location of the aperture and needle tip of the biopsy device in relationship to the targeted AOC. In addition, the user interface of the acquisition workstation displays a graphic representation of the aperture and needle tip of the biopsy device in relationship to the targeted AOC.

Clinical aspects of the TVAB procedure

TVAB were performed with an Eviva-Suros 9-gauge handpiece, needle diameter 9 gauge, needle aperture 2 cm, sample length 1.3 cm. Whenever possible, 12 samples are removed. To facilitate biopsy, DBT and stereotactic images with both craniocaudal and mediolateral projections were obtained from each target breast and stereotactic images compared with the DBT (Fig. 2a–c).

Fig. 2
figure 2

(a) Situation following successful TVAB of an architectural distortion in a study patient. The MIBB operator scrolled through the data set until the lesion center (crosshairs) appeared. (b) The TVAB system offers the depth of the lesion (crosshairs) and the distance to the adjacent skin (safe access path); it calculates the depth of the needle tip so that the lesion is positioned in the centre of the needle chamber. (c) After tissue sampling, the TVAB system shows the hematoma and the clip in the correct location (depth of lesion and hematoma correlate)

Full size image

All interventions were performed on patients in decubitus or upright position. Patients in upright position were offered 15 Etilefrin drops (Effortil®) to avoid vasovagal reactions. All patients were given local anaesthesia by injection of at least 10 to 15 ml Lidocaine 2 % (Streuli®) subcutaneously and deeply around the calculated biopsy cavity.

Stereotactic-guided vacuum-assisted biopsy

Up to the introduction of TVAB at the Breast Center in December 2012, the Breast Center performed SVAB employing upright stereotactic technology (Digital StereoLoc® II for Selenia®) using the Selenia® Dimensions® Mammography system (Selenia®, Hologic, Inc., Bedford, MA). The SVAB procedure needs two planes to localize the target lesion in 3D space and to guide the removal of tissue. It applies the principle of parallax to determine the depth or “Z-dimension” of the target lesion.

Clinical aspects of the SVAB procedure

SVAB were performed like TVAB with an Eviva-Suros 9-gauge handpiece, needle diameter 9 gauge, needle aperture 2 cm, sample length 1.3 cm. Whenever possible, 12 samples were removed.

To facilitate biopsy, stereotactic images with both craniocaudal and mediolateral projections were obtained. All SVAB interventions were performed as described above for TVAB procedures.

Data collection and analysis

Patients

A chart review was performed and included data on age and clinical indications for TVAB and SVAB.

Lesions

Lesions were characterized by size (maximum diameter in cm), type and visibility (according to the BIRADS lexicon) [25]: shape: oval, round or irregular; margins: circumscribed, obscured, microlobulated, indistinct, spiculated; density: high, equal, low or fat-containing. Size measurement of architectural distortions was based on a consensus of two readers.

TVAB and SVAB

All lesions had to be occult or at least equivocal on ultrasound. Any architectural distortion detectable on ultrasound was biopsied under ultrasound guidance and not included in this study. TVAB and SVAB examinations were evaluated for patient positioning, duration (procedure time between scout tomosynthesis for TVAB or scout stereo image for SVAB as start and proof of correct clip location by tomosynthesis or stereo image as end), accuracy of sampling in correlation to lesion size, and occurrence of complications such as pain, hematomas or infections. TVAB accuracy was checked by an additional DBT in mediolateral projection to assess removal of the target lesion.

Histopathology

Histopathology findings by TVAB and SVAB were graded by the B-classification and compared with the histopathology findings at surgery. The B-classification was produced by the UK National Coordinating Committee for Breast Screening Pathology and is endorsed by the European Commission working group on breast screening pathology [26]. It assigns the following grades: B1, normal tissue only; B2, benign lesion; B3, lesion of uncertain malignant potential; B4, suspicion of malignancy; B5, malignant. The categories take into account only the histological nature of the specimens [23]. All B5 lesions (TVAB n = 44; SVAB n = 30) were surgically excised. The 104 benign lesions of TVAB and the 56 of SVAB were not excised and did not undergo any form of regimented follow-up.

Statistics

Student t tests were used to compare biopsy time and biopsy type distribution for TVAB versus SVAB.

Results

Patients and feasibility

One hundred and forty-eight consecutive patients (mean age 56.6 years, range 32-85 years) admitted to the Breast Centre between 15 December 2012 and 15 October 2013 (closing date for inclusion) for SVAB diagnostic workup were included. All 148 patients could be biopsied with the TVAB system.

In 2011, 86 patients (mean 56.2 years, range 40-80 years) underwent SVAB at the Breast Centre. In one of the 86 SVAB procedures, the targeted microcalcifications could not be gained due to repeated miscalculations of the target depth.

Lesions

Twenty-four architectural distortions and 124 grouped microcalcifications were biopsied by TVAB. Calcified lesions ranged in size from 0.3 cm to 6 cm (mean size 0.9 cm), and architectural distortions ranged from 0.5 cm to 2 cm (mean size 0.9 cm).

82 grouped microcalcifications, four masses and no distortions were biopsied by SVAB. Calcified lesions ranged in size from 0.4 cm to 3.5 cm (mean size 1.5 cm), masses from 0.8 cm to 2 cm (mean size 1.3 cm).

Forty-four of 148 TVAB histologies were malignant (30 %), while 30 of 86 SVAB (35 %) were. Thirty-one of 124 TVAB biopsied microcalcifications were malignant (25 %), while 29 of 82 SVAB (35 %) were.

TVAB interventions

The mean duration of the 148 TVAB interventions (procedure time between scout tomosynthesis and proof of correct clip location by tomosynthesis) was 15.4 minutes (range 7–28 min), and therefore significantly faster than SVAB with 23 minutes (range 11–46 min), p < 0.0001. For radial architectural distortions, TVAB lasted a mean of 16 minutes (range 11–28 min). One patient suffered moderate to severe pain during the procedure despite extended local anaesthesia. This resulted in the longest intervention at 28 minutes. All other patients experienced no or only slight pain during the procedure. The patient with the longest procedure time was also the only patient with signs of a postoperative mastitis (treated with antibiotics). No infection of the biopsy site itself was detected.

SVAB interventions

The mean duration of the 86 SVAB interventions (procedure time between scout stereotactic image and proof of correct clip location by tomosynthesis) was 23 minutes (range 11–46 min). Despite multiple adjustments of the needle during one SVAB procedure, the targeted microcalcifications could not be gained. Nonetheless, being an equivocal target on ultrasound, the biopsy could be finally conducted sonographically. In three patients, the needle had to be adjusted at least once. After SVAB procedures, one large hematoma and two prolonged bleedings occurred. No infection of the biopsy site itself was detected. The reports are incomplete regarding pain experienced during the interventions, and evaluation was regarded to be infeasible.

TVAB histopathology

Twenty-four architectural distortions and 124 grouped microcalcifications were biopsied. The histopathological findings according to the B classification were: 44 B5, 40 B3, and 64 B2 lesions (Table 1). Architectural distortions were exclusively biopsied by TVAB, p < 0.0001. Thirteen of the 24 patents (54 %) with architectural distortions had breast carcinomas (11 invasive carcinomas, two DCIS), and 11 had a radial scar/ complex sclerosing lesion (CSL). One patient with AD had a CSL diagnosed on TVAB histology, but as only 25 % of the lesion was biopsied, the patient proceeded to magnetic resonance imaging (MRI). Open biopsy eventually identified invasive cancer.

Table 1 Histopathology of lesions biopsied by TVAB
Full size table

In one patient with a complex sclerosing lesion (2 cm in diameter), 25 % of the lesion could be removed. The histopathological finding of fibrosis was equivocal and MRI eventually revealed a highly suspicious 4 mm lesion on the periphery of the radial architectural distortion; it proved to be an invasive carcinoma after MRI-guided biopsy and surgery.

SVAB histopathology

Eighty-two grouped microcalcifications, four masses and no distortions were biopsied. The histopathological findings according to the B classification were: 30 B5, 23 B3, and 33 B2 lesions (Table 2). Three of the four patients with masses were B2 histologically (fibroadenoma), and one was B5 (DCIS).

Table 2 Histopathology of lesions biopsied by SVAB
Full size table

TVAB histopathology after surgery

Surgery was performed on all patients with B5 lesions; in all cases but the single architectural distortion described above, it confirmed the histopathological diagnosis based on the TVAB. Surgical confirmation was not performed on the 104 B2 and B3 lesions.

SVAB histopathology after surgery

Surgery was performed on all patients with B5 lesions; in all cases it confirmed the histopathological diagnosis based on the SVAB.

Discussion

The introduction of tomosynthesis-based imaging resulted in a significant increase in cancer detection rates, specifically in the detection of small invasive, node-negative cancers [5]. The diagnosis and management of these lesions and distortions, which are frequently small, presents the greatest test of radiological skill [27]. The majority of breast abnormalities are currently biopsied under ultrasound guidance; however, certain microcalcifications and small parenchymal deformities are not demonstrable and thus require stereotactic guidance [27]. Stereotactic-guided vacuum-assisted biopsy is an accurate technique for sampling breast microcalcifications [18], but is currently considered to be less suitable for detecting lesions very close to the skin, low contrast lesions and architectural distortions [20, 23, 24]. SVAB is also prone to inaccuracy in lesion targeting, because the user may fail to identify the same lesion in the pair of stereotactic images. This failure results in a miscalculation of the depth of the lesion, commonly referred to as the Z value. The literature contains little information on the accuracy of sonographic detection and localization of architectural distortions, and the conclusions offered are sometimes contradictory. Shatty et al [15] and Lee et al. [16] report that architectural distortions were visible in 50 to 68 % of ultrasounds. Finlay et al. [28], in contrast, identified 21 of 22 architectural distortions using ultrasound. In our experience, architectural distortions detected by DBT are often invisible or equivocal on FFDM and/or ultrasound. The SVAB data of 2011 in the MIBB archive support our observations. At that time, larger distortions or more accentuated ones were biopsied by SVAB if detected, but there were none in 2011. Thus, it may be assumed that many of the small distortions detected by DBT in 2012 and 2013 were just missed by FFDM or not traceable on stereotaxy in 2011. But this is speculative, because our data on biopsied architectural distortions by ultrasound and MRI by our own team and the teams referring patients to our centre are incomplete.

The corner stone of this study is that TVAB is able to biopsy even very small architectural distortions with a high accuracy. Fifty-four percent of biopsied distortions by TVAB were invasive cancer or DCIS, and 46 % were radial scars. In just one case did undersampling occurred. In correlation, in their DBT study, Freer et al. found a group of 36 ultrasound and mammography occult architectural distortions [24]. Obvious reasons for the inability of ultrasound to detect architectural distortions are large breasts, surrounding mastopathy-altered tissue, or the presence of multiple associated lesions. The use of ultrasound guidance to localize and biopsy architectural distortions is often unsatisfactory and time-consuming for both examiner and patient. Prior to the implementation of TVAB at our centre, patients with ultrasound-negative radial architectural distortions had to be transferred to cost-intensive and time-consuming MRI for further evaluation and possible MRI-guided vacuum-assisted biopsy. With DBT we can now perform a 3D visual review of thin breast sections unmasking distortions and lesions obscured by normal tissue, thus allowing detection and exact localization of the target lesion with consequent accurate planning of the intervention [9]. Three-dimensional imaging of a lesion by DBT allows the physician to scroll through the slices to precisely locate the lesion at the correct depth and mark it on one view for accurate planning of the intervention. Unlike TVAB, planning of SVAB is liable to varying types of systemic failure, because the planner has to mark the same target point on two geometrically different images. This is especially difficult for diffuse lesions and distortions, which can lead to extreme miscalculations (Fig. 1a–c). Furthermore, TVAB has the advantage of enabling accurate calculation of the distances between target location and skin in mediolateral and craniocaudal compression, facilitating planning of the easiest and safest access path to the lesion so as to avoid complications such as skin injuries and pain. The faster operation speed of TVAB and the need for fewer control images shorten the procedure time. In correlation to our findings, Schrading et al. found a mean time to complete TVAB of 13 minutes versus 29 minutes for SVAB [23]. This improves patient compliance and results in fewer movement artefacts. One patient in our study with a large architectural distortion had a false-negative finding of a benign complex sclerosing lesion because the malignant part of the lesion was situated on the periphery. These undersamplings can occur during all image-guided breast biopsy methods [29]. This can be attributed to the fact that malignancies in a radial scar, which itself represents a benign pathology, appear on a mammogram as an architectural distortion and cannot be differentiated mammographically as being benign or malignant. Unfortunately a benign radial scar can become malignant, the malignant part often being located, as in our patient, on the periphery [29]. In patients with architectural distortions larger than 1.5 cm or in whom postoperative tomosynthesis shows relevant residual tissue, MRI needs to be performed. The mean size of architectural distortions in our study was 0.9 cm and in 23 of 24 patients they could be almost completely removed with TVAB. Until studies show that TVAB is able to completely and accurately remove suspicious lesions, at a minimum all detected B4 and B5 lesions need to be operatively evaluated. The major limitation of this study is its lack of complete follow-up data for the lesions interpreted as benign at biopsy (B2 and B3 lesions), which could result in an underestimation of false-negative results. At that time, the state had no population based screening program. Patient follow-up was performed by the patients’ gynaecologists. As far as we could follow the patients, we have not found any missed cancer in both study groups. However, all our findings correlate well with the results of the DBT and TVAB studies by Freer et al. and by Schrading et al. [23, 24].

Conclusions

TVAB is able to biopsy very small architectural distortions with high accuracy. TVAB is easily feasible and appears to have the same high degree of clinical performance for diagnosing breast microcalcifications as SVAB, as far as can be known without long-term follow-up. The increased number of biopsied distortions by TVAB is presumably due to increased use of tomosynthesis and its diagnostic potential.