|Home | About | Journals | Submit | Contact Us | Français|
To determine the spatial localization errors of magnetic resonance imaging (MRI) guided core biopsy for breast lesions using the hand-held vacuum assisted core biopsy device in phantoms and patients.
Biopsies were done using a 10-gauge hand-held vacuum-assisted core biopsy system (Vacora, Bard, Arizona) on a 1.5 T MRI scanner (Philips Achieva, Best, The Netherlands). A standardized biopsy localization protocol was followed by trained operators for multi-planar planning of the biopsy on a separate workstation. Biopsy localization errors were determined as the distance from needle tip to center of the target in 3 dimensions.
Twenty MRI-guided biopsies of phantoms were performed by 3 different operators. The biopsy target mean size was 6.8 ± 0.6 mm. The overall mean three dimensional (3D) biopsy targeting error was 4.4 ± 2.9 mm. Thirty two MRI breast biopsies performed in 22 patients were reviewed. The lesion mean size was 10.5 ± 9.4 mm. The overall mean 3D localization error was 5.7 ± 3.0 mm. No significant differences between phantom and patients biopsy errors were found (p-value >0.5).
MRI guided hand-held vacuum assisted core biopsy device shows good targeting accuracy and should allow localization of lesions to within approximately 5–6 mm.
Breast magnetic resonance imaging (MRI) is reported to have very high sensitivity (>90%) but only moderate specificity for breast cancer evaluation (1–3). In a large multi-center study, the specificity for both invasive breast cancer and ductal carcinoma in situ (DCIS) was only 67% (4). For high risk populations, the pre-test probability of breast cancer is sufficiently high that its use has been recommended for breast cancer surveillance (5). Nevertheless, a common clinical scenario is detection of small lesions (less than 10 mm) with MRI that cannot be further characterized or that are not identified on the corresponding mammogram or with second look ultrasound (6).
As a result, MRI guided biopsy must frequently be used to provide histological determination of breast lesions that are otherwise not characterized as benign or malignant. Prior patient studies using various MRI guided biopsy devices correlated biopsy outcome with clinical follow-up for benign biopsies, or surgical results for positive biopsies (7). To our knowledge, no studies have previously measured the limits of accuracy for the MRI guided hand-held vacuum assisted core biopsy device. Knowledge of biopsy precision would help to establish (a) the minimum lesion size to be attempted for patient studies and (b) understand factors that influence biopsy procedures.
This study consisted of two parts: 1) MRI guided biopsy on gel-based biopsy phantoms with MRI visible randomly positioned embedded targets and 2) MRI guided biopsy of breast lesions in patients. For both phantom and patient studies, a standardized approach (discussed below) was used by the operators for multi-planar reconstruction of magnetic resonance images for target/lesion localization and distance measurement The study was approved by our institutional review board.
MRI guided biopsy was performed for 20 targets embedded in 6 different gel based biopsy phantoms (Invivo, Orlando, FL). Three different operators with varying level of experience performed the biopsies (median, 5 biopsies per operator). At least 6 samples were acquired from each target. Biopsy targets were randomly assigned to the operators by an independent observer.
Retrospective reviews of MRI guided biopsies that were performed for 32 breast lesions detected in 22 patients by four radiologists (median, 9 biopsies per operator). All patients were referred for breast biopsy based on a prior breast MRI examination that detected a suspicious lesion that could not be localized on mammogram or ultrasound. Institutional review board approval was obtained for this retrospective analysis.
A 10 gauge hand-held vacuum-assisted core biopsy gun (Vacora, Bard, Arizona) and localization system were used. All biopsies were done using a 1.5 T system (Philips Achieva, Best, The Netherlands) with a biopsy compatible 7 channel breast coil (in-Vivo, Orlando, FL). The patients were scanned in the prone position and both patients and biopsy phantoms were positioned for a lateral approach with mild compression to minimize the shift in breast tissue during the procedure which would decrease the localization errors. Initial localizing scans were obtained followed by a sagittal non-fat suppressed T1 weighted scan on the targeted side (bilateral in cases of bilateral biopsies). For patient studies, high spatial resolution three dimensional (3D) T1 weighted spoiled gradient echo (GRE) scans with fat-suppression were acquired in the transverse plane (TR: 7.08; TE: 3.56; flip Angle: 10; slice thickness: 2 mm; FOV: 35 × 35 cm; and matrix: 512 × 512) both before and after intra-venous administration of Gadobenate Dimeglumine (Gd-BOPTA, Multihance; Bracco Imaging SpA, Milan, Italy) in the rate of 2 ml/sec and a dose of 0.1 mMol/Kg. On the other hand for phantom studies the scans were acquired once. Subtraction images were created as well. A written biopsy protocol was followed by each operator for multi-planar planning of the biopsy on a separate workstation (Ultravisual, Emageon, Alabama), (8) including lesion targeting with 3-plane coordinates and depth assessment. Operators were not allowed to visually inspect the phantom during the procedure. A 9 gauge introducer was placed into the phantom by the operator allowing for the 20 mm throw of the biopsy gun needle during sampling. The introducer stylet was replaced by a plastic cannula, through which the 10 gauge core biopsy system was advanced to the target. A line was drawn using a sterile marker on the cannular to mark the desired depth. A minimum of 6 samples were acquired from each lesion. The needle was removed after obtaining each sample to evacuate the tissue sample. The phantom was rescanned after each biopsy.
For the biopsy system, we measured all parts of the localization system as well as the hand-held vacuum assisted core biopsy gun and its needles. The localization system consists of a compression grid plate and a biopsy block (fits into any of the grid holes). Measurements included: a) thickness of inner biopsy block and outer grid margins, b) block thickness, c) length and diameter of the biopsy chamber, d) length of throw of the biopsy needle when firing the biopsy gun.
For phantom studies, review of the post biopsy images was performed to determine the biopsy errors and the distance between the center of biopsy track and the center of the target, in anterior to posterior (AP), right to left (RL) and superior to inferior (SI) directions. For further confirmation, the biopsy chamber was inspected to determine if target was within the chamber. The phantom was directly inspected to determine if the biopsy track was seen passing through the target.
For patient studies, we reviewed MRI guided breast biopsy studies in our hospital over the period between July, 2007 and March 2008. The needle localization errors were measured as the distance between the localization needle tip and the center of the lesion in AP, RL and SI directions. Adjustment was made for the RL (depth) localization due to the throw of the biopsy needle in order to report distances to the center of the biopsy chamber by subtracting the measured RL error from 10 (10 – RL error) to account for the 10 mm intended RL distance when localizing. For patients, needle localization errors were measured prior to firing of the biopsy gun to avoid obscuration of the lesion by the associated hematoma or washout out of the enhancing lesion. Pathology results were reviewed to confirm positive biopsies.
The overall error for both phantom and patients studies was calculated using the following formula:
Mean and standard deviations were calculated for all measurements. Distance to the centers of the targets/lesions are presented as mean and standard deviation as well as range (minimum -maximum). All statistical analysis were performed using Excel software (Microsoft, Redmond WA).
The biopsy system consisted of a compression grid plate used to stabilize the breast with a biopsy block used to stabilize the biopsy needle. The biopsy block had 9 holes (3 × 3 matrix) with an inter-hole spacing (center to center) of 5 mm. The thickness of the grid lines was 1 mm. The thickness (depth) of the biopsy block was 20 mm. The biopsy gun specimen chamber length was 20 mm and its width was 3 mm. The length of the needle throw was 20 mm.
Twenty MRI-guided biopsies of randomly positioned embedded targets were performed (Figures 1, ,2).2). Target mean size was 6.8 ± 0.6 mm (range, 5.4–7.8 mm) and target mean depth was 41.2 ± 15.4 (range, 15–65 mm). The mean biopsy error to the target center was 0.95 ± 1.1 mm (range, 0–3.9 mm) in AP direction, 1.3 ± 1.6 mm (range, 0–4.7 mm) in SI direction, and 3.8 ± 3.3 mm (range, 0–10 mm) in RL direction. The overall mean 3D error was 4.4 ± 2.9 mm (range, 0–10 mm).
In 19 biopsies, biopsy tracks were visualized through the embedded targets. Fragments of the embedded targets were obtained in the specimen chamber in 19/20 (95%) of biopsy attempts. In 1/20 (5%) attempt, the introducer was repositioned by the operator when the target was located directly underneath the grid bar due to planning error. For this lesion the biopsy error, was 2.3, 3.8 and 0 mm in AP, SI and RL directions respectively.
Thirty-two MRI-guided biopsies of breast lesions in 22 patients were evaluated. The mean patient age was 53 ± 10.7 years. Lesion mean size was 10.5 ± 9.4 mm (range, 4–50 mm) and lesion mean depth was 29.5 ± 13 (range, 9–46 mm). The mean localization errors were similar to those observed in phantoms (p >0.5, unpaired t-test): 1 ± 1.1 mm (range 0–4.5 mm) in AP direction, 1.4 ± 1.8 mm (range 0–5 mm) in SI direction and 4.1 ± 4.1 mm (range 0–7.8 mm) in the RL direction. The overall mean 3D error for patients was 5.7 ± 3.0 mm (range 0–10.8 mm).
With the advances in MRI technology and its improved ability to detect small lesions (e.g. less than or equal to 5 mm), the need for developing MRI guided biopsy system for these lesions has increased. Knowledge of biopsy error is important when considering the minimum lesion size targeted for MRI guided biopsy. The biopsy system chosen in this study was a hand-held vacuum assisted core biopsy device with relatively large biopsy core (10 gauge). The mean errors were similar in the phantom study compared to the patient study (4.4 vs. 5.7 mm, respectively). Overall, biopsy errors were larger in the right-left (biopsy depth) direction in both phantom and patient studies due to less depth constraint on the operator using the grid biopsy system. The biopsy device we used had a long biopsy sample chamber (20 mm). In practice, this long chamber would reduce the impact of localization errors in the depth direction.
One of the major difficulties that are commonly faced when using a grid system in lesion localization is having the lesion coordinates corresponding to one of obstructing grid/block lines instead a block hole. For these cases, angled biopsy guides are available. However in most cases, the operators were able to slightly angle the biopsy gun using a fixed hole guide while choosing the biopsy hole closest to the lesion coordinates. For the system we used, the thickness of the biopsy grid lines was 1 mm. The maximum localization errors that actually observed were on the order of 5 mm in the AP and SI directions.
The overall accuracy was 95% in phantoms and 94% in patients. In 2 patient studies, there was inaccurate lesion localization (Figure 6). One error was due to incorrect biopsy hole choice by the operator. The second error was due to incorrect consideration of the 20 mm biopsy throw. These errors can be avoided by establishment of a consistent biopsy protocol when multiple operators are involved. In our case, a biopsy template/checklist is now filled out for each lesion that should minimize operator variability.
Prior studies for a hand-held MRI guided biopsy system were performed to determine accuracy and cancer yield, although none was performed to quantify 3D biopsy error as in the current study. Based on correlation of MRI guided biopsy results and surgical pathology, the previously reported accuracy of MRI guided biopsy ranged from 95–100% and cancer yield ranged from 5–61% (Table 1). Our results were consistent with prior studies in this regard, with biopsy accuracy of 94% and cancer yield of 12.5%.
Biopsy “success” is more likely for larger lesions and is easier to demonstrate if the biopsied tissue shows malignancy. Liberman et al. (9) performed vacuum assisted MRI guided biopsy on 98 patients with median lesion size of 10 mm (4–85 mm) (table 1). They divided their sample into two groups; the first group consisted of patients with MRI guided biopsy results of malignancy and the second group was the benign biopsy results patients. For the first group they matched the MRI guided biopsy results with the surgical pathology results, accuracy of MRI guided biopsy was 77.4 % (24/31). For the second group a follow up MRI exam was performed (1 to 11 month post-biopsy) to confirm the benign diagnosis; in this group, one out of 52 patients developed a new mass after 11 months that was proved to be in situ and infiltrating ductal carcinoma.
Lehman et al. (10) evaluated MRI biopsy of 38 lesions by comparing biopsy results to surgical results. They divided their lesions into two groups; the first group included masses and foci (mean 9 mm range 2.5–19 mm) and the second group included non mass like enhancement (mean 23.6 and range 6–70 mm). They reported an overall 100% accuracy and 40% cancer yield. Orel et al. (11) evaluated 85 lesions reporting 61% cancer yield in lesions with 17 mm mean diameter (range 5–100 mm). Ghate et al. (12) performed MRI guided biopsy of 20 lesions with mean lesion diameter of 8 mm (range, 4–20 mm). They reported 95% accuracy and 5% cancer yielding comparing biopsy results with surgical results or follow up for benign lesions both by biopsy results and MRI characteristics.
Limitations of this study include measurement errors due to difficulties accompanying susceptibility artifacts produced by the localization needle. To reduce this problem, we withdrew the metallic biopsy needle prior to imaging and used a plastic cannula both for hemostasis in patients and to decrease susceptibility artifacts. Also, in patient studies we measured the localization error (prior to discharging the biopsy gun) rather than the biopsy error due to the frequent occurrence of hematoma following biopsy.
In conclusion, there was good agreement between phantom studies and patient studies regarding the degree of biopsy error associated with a MRI guided hand-held vacuum assisted core biopsy device. Use of a standard biopsy protocol should allow biopsy of lesions as small as 5–6 mm. The procedure may relatively be independent of the operator when a consistent procedure plan is used.
Supported in part by NIH R01CA125258, P50CA103175 and by the intramural research program of the NIH/Clinical Center. Dr. El Khouli is supported by the Egyptian government.