Techniques for Improving Ultrasound Visualization of Biopsy Markers in Axillary Lymph Nodes

INTRODUCTION

When it was determined that the false-negative rate of sentinel lymph node surgery in patients with node-positive breast cancer (T0-4, N1-2, and M0 [TNM classification]) could be reduced by identifying and resecting the marked positive axillary lymph node,^[1] pre-operative localization of the marked node became the practice at many medical facilities. Pre- operative localization of the marked node can be challenging, particularly after positive treatment response to neoadjuvant therapy. At this time, the positive axillary lymph node often becomes mammographically or sonographically occult, leaving only the biopsy marker in or near the site of the node.

Biopsy markers are made of either titanium, stainless steel, an alloy of titanium, and nickel, or ceramic and typically range from 1 mm (smallest dimension) to 8 mm (greatest dimension) [Table 1]. A myriad of marker conformations is commercially available, each describing visibility on ultrasound (US), mammograms, and magnetic resonance imaging (MRI), as well as minimal soft-tissue migration of the marker once it has been deployed.^[2-4]

Despite mammographic conspicuity of all metallic biopsy markers, US is the imaging modality of choice in the axilla. Aside from using ionizing radiation, a mammographic approach can be technically challenging in the axilla, particularly when the marked positive node must be included between the mammographic plates. For many patients, this can be an uncomfortable position, making it difficult to perform mammographic- or tomosynthesis-guided localization. If the marked positive lymph node is located very far posteriorly, a mammographic technique may be prohibitive. For these reasons, breast radiologists prefer US- guided localizations of the marked lymph node. However, confidently identifying the marked positive lymph node by US after neoadjuvant chemotherapy can be challenging, if not impossible.

Two main factors contribute to the difficulties in detecting the marked lymph node after neoadjuvant chemotherapy. The first factor is poor marker conspicuity due to surrounding echogenic Cooper’s ligaments, dense fibroglandular tissue, or shadowing tissue interfaces related to the heterogeneity of the breast parenchyma. The other factors are suspected extrusion or migration of the marker away from the positive lymph node as the lymph node normalizes during neoadjuvant chemotherapy. Both factors uniquely contribute to the complexity of US detection of the positive lymph node at the time of pre-operative localization.

Very little, if any, literature is available that describes ways to optimize sonographic visualization of the marker in an axillary lymph node, particularly after neoadjuvant chemotherapy. The purpose of this manuscript is to optimize sonographic scanning parameters for marker conspicuity in a phantom and cadaver model. A reflection on more than a year of clinically implementing these optimization trends is provided. Finally, a few helpful strategies are suggested, which might increase the likelihood of successfully identifying the marked lymph node by US in clinical practice.

MATERIALS AND METHODS

This study was approved by our Institutional Review Board and was compliant with the Health Insurance Portability and Accountability Act (HIPAA). The ex vivo section of this study involved (1) designing a reproducible phantom model that mimics normal mammary and axillary soft tissue and (2) sonographically evaluating biopsy markers used in our breast radiology practice, first in the phantom model and then in the cadaver model. The biopsy markers or clips used in this investigation were either expired or opened but never used; these included O-, cylinder-, top hat-, coil-, and U-shaped markers [Table 1]. The O-shaped and U-shaped markers are associated with resorbable packing or embedding material;^[5] the coil-shaped clip with biodegradable hydrogel polymer; and the top hat-shaped marker with resorbable netting. Such surrounding material was removed, and the metallic clips were isolated for the ex vivo studies to simulate what generally remains in patients after months of neoadjuvant chemotherapy.

Eight clinically available US scanning parameters, frequency, dynamic range, cross-beam setting, beam steering, speckle reduction processing, harmonic imaging, colorization, and speed of sound, were adjusted to optimize conspicuity of the clip on a GE LOGIQ E9 US unit (GE Healthcare Inc., Wauwatosa, WI USA). The broad-spectrum linear matrix array transducer (ML6–15) is the probe most often used for breast USs, and this was the probe used for the cadaver study. For the phantom study, the 9L probe was also used. Time gain compensation, focusing, and output power were variably adjusted as in the clinical setting.

Phantom study

The phantom study began with the design and development of a phantom model that simulates mammary parenchyma on US. Grapefruit pulp provided heterogeneity suggestive of background breast parenchyma that would make finding a biopsy marker challenging [Figure 1]. Grapefruit pulp is difficult to maintain in form and difficult to preserve. Therefore, the recipe for a standard gel phantom was modified to include talcum powder or flour of varying amounts. The phantom was created with layers of molded gel (some with and without talcum powder or flour) and variably thick layers of paper towels/tissue [Figure 2]. A further modification which wrapped the gel pieces with cellophane and then stacked them randomly in a mold resulted in similar imaging features.

Close

Figure 1:

Grapefruit simulating mammary tissue on ultrasound (US). Pulp from a grapefruit was collected and placed in a container, and US gel was placed on top (a). The US image (b) shows heterogeneity and shadowing that mimics the tissue interfaces in mammary tissue.

Export to PPT

Close

Figure 2:

Construction of a phantom that simulates normal mammary parenchyma on ultrasound (US). Using a standard recipe for making gel phantoms, the material was poured into a cellophane-covered mold with shallow half-egg shapes (a). These gel molds were layered (b) along with other layers of phantom material mixed with either talcum powder containing Ca3(PO4)2 or ZnO or mixed with 1 g flour (c). Layers of paper towels of variable thicknesses (d, layers) were layered in between the gel layers. US shows heterogeneous echotexture (e,f) that simulates mammary parenchyma.

Export to PPT

The O-, cylinder-, top hat-, coil-, and U-shaped markers were placed into one of the phantoms and scanned with an ML6–15 MHz probe using default settings that simulate the clinical environment.

RESULTS

In the phantom study, the O- and U-shaped markers were essentially inconspicuous [Figure 3] with the O-shaped marker marginally more conspicuous than the U-shaped one. For the cadaver study, the adjustments to the eight scanning parameters of the default settings are summarized in Table 2, where additional figures are referenced.

Close

Figure 3:

Biopsy markers in a phantom. The O-shaped (a) and U-shaped (e) markers were less conspicuous than the cylinder (b), top hat (c), and coil-shaped (d) markers on ultrasound (US) using default scanning parameters for breast US and an ML6-18 MHz probe.

Export to PPT

Close

Figure 4:

Optimization of frequency in a cadaver model. Ultrasound (US) transducers are specified by their center frequencies but operate within a range of frequencies or bandwidth. Breast US typically uses a few transducers to provide frequencies in the 7 MHz–18 MHz range. Although higher frequencies offer higher resolution, the depth of penetration through soft tissues is lower, making higher frequencies better for superficial imaging and lower frequency better for imaging deeper soft tissues. The biopsy marker is an echogenic focus in each image (arrow). The default biopsy mode uses 11 MHz (a). The adjustments were 9 MHz (b), 13 MHz (c), and 15 MHz (d). The preferred frequencies for this case were 13 MHz and 15 MHz, all higher than the default. This could be related to the depth of the target for this case. As a rough approximation, the depth of penetration can be estimated by (60 cm MHz)/f, where f is in MHz. Using this rough calculation in this particular case, a 60 MHz center frequency transducer might be helpful for this depth of penetration. However, this rough calculation is affected by the attenuation coefficient, an indicator of acoustic loss, which can be quite variable depending on the insonated soft tissues.^[12] In this case, increasing the frequency narrows the width of the ultrasound beam which makes it more and more difficult to get around the marker and thus makes the shadow prominent.

Export to PPT

Close

Figure 5:

Optimization of dynamic range in a cadaver model. The dynamic range describes the range in decibels (dB) between the largest and smallest signal at any given processing step. It is roughly analogous to how wide or narrow the window is on a radiograph or computed tomography. A wide dynamic range has more shades of gray, less contrast, and makes the image appear smoother.^[13] A narrow dynamic range has fewer shades of gray, more contrast, and makes the image appear grainy. The biopsy marker is an echogenic focus in each image (arrow). The default biopsy mode is 72 dB (a). The adjustments were 41 dB (b), 48 dB (c), and 69 dB (d). The preferred dynamic range was 69 dB. The increased contrast associated with the narrower dynamic range seemed to make it harder to distinguish the hyperechoic marker from adjacent hyperechoic Cooper’s ligaments and parenchymal interfaces.

Export to PPT

Close

Figure 6:

Optimization of cross-beam in a cadaver model. Cross-beam, also called spatial compound imaging, averages three or more echoes from different steering angles into a single image.^[14,15] This is accomplished when the probe sends multiple interrogation beams at various angles through the same tissue instead of at a single angle as in conventional B-mode imaging. Spatial compound imaging or cross-beam techniques have been shown to eliminate noise, graininess, and refractive shadows,^[14,16] creating improved edge detail and contrast. The use of cross-beam slows down the frame rate (due to more lines of sight) and is variably helpful clinically for either very superficial or very deep structures in the breast. The biopsy marker is an echogenic focus in each image (arrow). The default biopsy mode is cross-beam low mean (a), which means that separate sets of pulse or echo data are acquired for a low number of angles and that the averages of the detected echo signals from each angle are then assigned to the compound image pixel values. The adjustment to cross-beam medium hybrid (b) was preferred for this case. This means that a medium number of angles were used to acquire the data, and a hybrid calculation between mean and max (determined by the vendor) was used to determine the compound image pixel values.

Export to PPT

Close

Figure 7:

Optimization using steering in a cadaver model. Beam steering is a form of compound imaging, in which the beam is directed perpendicular to a biopsy needle or other object of interest, such as a marker. The result is that the object appears brighter. The biopsy marker is an echogenic focus in each image (arrow). The default biopsy mode is without beam steering (a). The preferred adjustment is with beam steering (b). In this case, it is presumed that beam steering is perpendicular to the marker.

Export to PPT

Close

Figure 8:

Optimization of speckle reduction index in a cadaver model. Speckle is an intrinsic artifact which degrades ultrasound image quality. Speckle reduction imaging (SRI) uses image processing techniques to remove speckle.^[17,18] SRI will generally make the image appear smoother. Low SRI removes a small amount of artifact, while high SRI may look over-processed. Higher SRI is associated with a slower frame rate. The biopsy marker is an echogenic focus in each image (arrow). The default biopsy mode has a speckle reduction set at 4. The adjustments are SRI = 0 (B), SRI = 3 (C), and SRI = 5 (D). The most preferred of the three was SRI = 3.

Export to PPT

Close

Figure 9:

Optimization of harmonic imaging in a cadaver model. Harmonic imaging processes the second harmonic signal generated during insonation.^[13,19] Advantages of harmonic imaging include increased signal-to-noise ratio, which makes it occasionally helpful in the breast to characterize near-field, small, isoechoic masses, and far-field entities. The biopsy marker is an echogenic focus in each image (arrow). Harmonic imaging is disabled in the default biopsy mode setting (a). There was no preference between turning harmonic imaging off (a) or on (b) for this case. Harmonic imaging also improves resolution (2× imaging frequency), which can be another observation here that improving spatial resolution did not really help – it may be because we are improving the resolution of everything by an equal amount.

Export to PPT

DISCUSSION

Identifying marked axillary lymph nodes for radioactive I-125 seed localization after neoadjuvant therapy can be challenging for several reasons. These include altered morphology (normalization) of the treated lymph node, suspected extrusion of the marker from the lymph node, and poor sonographic conspicuity of the markers. This study reviewed modifications of several widely available scanning parameters during US evaluation of marked axillary lymph nodes to improve marker conspicuity and confidence of marker detection. Although the availability of cadavers with axillary lymph nodes only allowed for limited evaluations, it did provide insight and the realization that some of the expected optimization trends were occasionally not the ones preferred. While the preference of higher frequencies, for example, made theoretical sense based on physics, other adjustments were more counter intuitive. For instance, in this study, there was no preference when using harmonic imaging. This seems to support studies which have shown that harmonic imaging provided limited value in needle visualization in the breast.^[6,7] There is also an interplay between dynamic range, spatial compound imaging, speckle reduction imaging, and harmonic imaging that need to be explored for each case, given the variations in depth of the marked lymph node, the type of marker, and the conspicuity of the positive lymph node after neoadjuvant chemotherapy. For the cadaver model, sonographic optimization techniques included increasing transducer frequency, slightly decreasing dynamic range, setting cross-beam to medium hybrid, turning on beam steering, and setting speckle reduction imaging to mid-range. Different vendors have slight variations that instantiate the parameters tested, and these will need to be optimized accordingly.

A limitation of this study is that any material surrounding the metallic marker was removed. For some markers, the surrounding material may be the very feature that makes it sonographically conspicuous. The biodegradability of materials surrounding the metallic components varies, so markers identified indirectly by said materials may or may not be detectable months after neoadjuvant chemotherapy.

After more than a year of implementing these US techniques optimized for marker detection in the clinical setting, successful detection seems to be a balance between techniques that increase contrast and those that minimize graininess. Vocal fremitus, elastography, Doppler, and power reduction provided no additional improvement in marker detection. However, a few strategies seem to improve the likelihood of detecting or confirming the marked positive node by US, and it starts from the time of the fine-needle aspiration and marker placement [Table 3].

When US is unsuccessful in identifying a marked axillary node, alternative imaging guidance such as CT or tomosynthesis can be considered.^[8-10] A limited selection of biopsy markers with similar imaging characteristics was used for the cadaver model. There are a variety of biopsy markers on the market, and a large number have been reviewed as options for localization of axillary lymph nodes.^[11] Hygroscopic markers such as the HydroMARK coil-shaped markers (Mammotome HydroMARK, Norderstedt, Germany) seemed hopeful for improving sonographic detection, but aside from challenging detectability over time and migration or extrusion of the marker, there were initially, at our institution, concerns of false positives at the time of surgical pathology. In addition, non-wire localization techniques continue to explore their roles as biopsy markers to potentially obviate the localization step; these techniques include non-radioactive signaling sources such as iron particles (Magseed, Sentimag, Leica Biosystems, Wetzlar, Germany), radar (Savi Scout, Cianna Medical, Aliso Viejo, CA, USA), and radiofrequency tagging (LOCalizer, Faxitron, Tucson, AZ, USA).^[11]

In the meantime, non-ionizing features, patient comfort, and costs make US the preferred imaging modality for identifying a marked axillary lymph node. There are several currently available US scanning parameters that can be explored and tweaked to improve the conspicuity of the marked node, and it may take extra time to determine the optimal setting for each case. In addition, there are also clinical strategies that can be taken to improve the chances of detecting the marked positive lymph node.

CONCLUSION

Ultrasound scanning parameters can be adjusted to improve the conspicuity of biopsy markers. Overall, optimization requires a balance between techniques that increase contrast (dynamic range, harmonic imaging, and steering) and those that minimize graininess (spatial compound imaging, speckle reduction imaging, and steering). Additional scanning and procedural strategies can improve the confidence of sonographic detection of biopsy markers associated with the intended target.