Background
The response of matter to electromagnetic fields is characterized by the material’s dielectric properties: permittivity and conductivity. The conductivity is a measure of the ease with which free electric charges can migrate through the material; the permittivity reflects the extent to which bound charge distributions can be distorted through polarization by an external field.
Differences in dielectric properties between human tissues and specifically between benign and malignant tissue have been studied and are well established [
1‐
3]. Tissue dielectric properties are determined by concentration and mobility of intra and extracellular components, cell size, structure and arrangement, amongst other characteristics. Specifically, the tissue dielectric properties have been extensively studied in breast tissue [
4‐
7], where differences between tissue types, and specifically between normal and malignant were observed over a broad range of frequencies. These properties have been successfully used to differentiate between normal and malignant breast tissue [
8,
9] and for intraoperative margin assessment during lumpectomies [
10]. Normal breast tissue is heterogeneous, being composed of three different types of tissue (Adipose, Glandular, and Connective), In all the studies to date, the evaluated dielectric properties were based on measurements and comparisons performed on scales larger than the intrinsic scale of tissue heterogeneity within breast tissue. Therefore, the differences observed represent average values of the dielectric properties. Results were generally reported on properties of “normal” and “cancer” types. In some reports [
5,
7] an attempt has been made to partition between the three intrinsic tissue types of the normal breast. There has been no reporting on the specific properties of glandular and connective tissue types. Also, cancer of the breast has three major types: Ductal Carcinoma
in Situ (DCIS), Invasive Ductal Cancer (IDC), and Invasive Lobular Cancer (ILC). The dielectric properties of these types, to date, have not been characterized separately. Additionally, specifically of importance in breast biopsy procedures, the dielectric characteristics of abnormalities in the breast that may progress to cancer, such as: non-malignant proliferative, non-malignant proliferative with atypia, and LCIS, have not yet been characterized.
The burden of breast cancer is high. Approximately 230,480 American women are diagnosed annually, and 39,520 women die from this disease [
11]. Global cancer statistics show that breast cancer is the most frequently diagnosed cancer and the leading cause of cancer death among females, accounting for 23 % of total cancer cases and 14 % of cancer deaths [
12]. The majority of breast cancers are diagnosed as a result of an abnormal mammogram or ultrasound, and in selected populations abnormal MRI findings. Some lesions are found by the patient or her physician as a palpable mass. Not all abnormal findings diagnosed by the methods mentioned represent cancer. To determine whether a mass in the breast is a suspicious mass or not the BI-RADS System was developed [
13]. All patients with a BI-RADS category of 4 and 5 should undergo a biopsy. Those with category 3 should be followed more frequently. A clinically suspicious mass should also be biopsied, regardless of imaging findings, as about 15 % of such lesions can be mammographically occult [
14].
Screening mammography is the most common way to diagnose early breast cancer but carries a high rate of recalls (16.3 % at first and 9.3 % at subsequent mammography) [
15]. Biopsy is further recommended in 0.6–1 % of all screened women [
16,
17]. Millions of women are screened each year, therefore these figures represent a high number of breast biopsies performed each year, emphasizing the need for accurate biopsy.
During the breast biopsy procedure part or all of a suspicious breast tissue growth is sampled and examined for the presence of cancer, most often in a minimally invasive procedure. Current biopsy techniques have several limitations: First, patients diagnosed with Atypical Ductal Hyperplasia (ADH) are routinely sent for an open surgical biopsy, following which 10–25 % of these patients’ diagnosis will be ‘upgraded’ DCIS, which requires a further lumpectomy or mastectomy [
18‐
21]. Second, patients with a diagnosis of DCIS in biopsy will undergo lumpectomy, typically without a sentinel lymph node biopsy. Following the Lumpectomy, about 20 % of these patients’ diagnosis will be upgraded to invasive cancer, requiring a further surgery for node biopsy. Third, studies have shown that up to 10 % of patients endure repeat biopsies [
22,
23]. These repeat biopsies reveal carcinoma in up to 25 % of cases [
24]. Forth, published data show a 1–7 % false negative rate with current breast core biopsy techniques [
25].
The inaccuracies in the biopsy procedure result mainly from the uncertainty in the exact location from which the biopsy sample is taken relative to the image, and from the fact that the features presented on imaging may not be the most abnormal tissue present. Having an in-situ, at the needle tip, tissue characterization ability when performing biopsy procedures has the potential to increase the accuracy of the procedure.
In the presented study, we use a miniature, 0.8 mm in diameter, RF sensor to characterize the dielectric properties of different breast tissue types and abnormities. Based on these characteristics, we tested the potential of this type of sensor in differentiating between normal, abnormal, and malignant breast tissue. The sensor has a coaxial opening that results in a fringing electrical field close to the sensor surface. The sensor is manufactured using flexible printed circuit board technology and can be potentially placed on various devices having different geometries, such as open surgical tools as well as minimally invasive ones, like core biopsy needles. A similar device has been already been used for evaluating freshly excised radical prostatectomy specimens [
26].
Discussion
Our results quantify dielectric properties of 10 different tissue types present within the breast, and show that they all have different dielectric properties. These includes the types related to normal breast tissue (Adipose, Glandular, and Connective), and types associated with various abnormalities, including the differentiation between the dielectric properties of the 3 types of cancer types: IDC, ILC, and DCIS. Prior studies [
4‐
7] have demonstrated that, generally, there are differences between normal and cancer tissue within the breast, but have not provided the level of differentiation presented in this work. Additionally, the data on the dielectric properties of pre-malignant types is new. These identified differences served as a basis for constructing a score variable that demonstrated correlation with the degree of abnormality, including pre-malignant phase, of the breast tissue. Additionally, using the dielectric properties and the score variable good differentiation between normal and malignant, or non-malignant abnormal, tissue was established. The configuration of the sensor as a 0.8 mm circular sensor can be thought of as a basic sensing unit for use in breast biopsy procedures, with an array of sensors aligned along the biopsy needle.
The goal of the initial biopsy is to obtain sufficient diagnostic material using the least invasive approach and to avoid surgical excision of benign lesions. CNB offers a definitive histologic diagnosis, avoids inadequate samples and may permit the distinction between invasive versus
in situ cancer. In most centers, image guided CNB has replaced wire localization and surgical excision as the most common initial biopsy method for nonpalpable abnormalities [
33‐
35]. The accuracy of CNB was shown in a series of 952 consecutive breast CNBs performed at one institution (342 without image guidance, 241 with ultrasound guidance, and 369 using a stereotactic vacuum assisted biopsy (VAB)) [
36]. The false-negative rate with 11-gauge VAB was 3 %, compared to 13, 5, and 22 % for non-image guided, surgeon-performed ultrasound-guided, and 14-gauge VAB, respectively. In most of these false negative patients (5–22 %) the reason for the false negative result was due to sampling error, meaning that the biopsy was not taken from the lesion as planned.
The false negative rate can be reduced by using very large needles [
37]. An alternative approach could be to keep the needle size relatively small, but add local sensors located at the needle’s tissue collection region that provide
in-situ information on the tissue about to be sampled. The basic units for these sensors can be sub-millimeter, circular near-field radio-frequency sensing units, as those we have used in this work for characterizing breast tissue properties. These sensors may provide real-time measurements of tissue dielectric properties at the locations of tissue to be biopsied. As the power transmitted by the sensor is very low and the RF radiation is non-ionizing, these type of sensors are well suited for in-vivo use.
As per the current standard of care, imaging will be used to direct the needle to the general location of the suspected abnormality. The in-site sensors will provide indication of the tissue type at the immediate vicinity of the needle and in contact with the sensor, as the sensor is effectively a surface characterization sensor. The penetration depth the 0.8 mm sensor in no more than 0.1 mm. By scanning/moving the needle around the suspected region, the most suspicious tissue abnormality can be identified, and biopsied. The spatial resolution of a sensing device designed using the sensors as the basic building blocks is dictated by the sensor size and by the ability to arrange sensors close together. Arranging 0.8 mm sensors in an array will preserve this resolution, as the sensors can share the same ground plane (the outer conductor of the coaxial aperture).
For a potential set-up for use in a biopsy device configuration, approximately 10 0.8 mm circular sensors will be arranged in a 1D array on the biopsy needle, in a location overlapping the biopsy sampling cavity. As each measurement takes approximately 2.5 msec, a reading from all 10 sensors will take ~ 0.025 s. The full measurement cycle, including displaying the results to the user, will take approximately 0.2 s, thus providing real-time tissue characterization as the needle is progressed through the tissue. Therefore, when the sensors will be integrated with the biopsy needle, it is anticipated that the duration of the biopsy procedure will not be extended.
DCIS is presented mostly as microcalcifications on mammography and diagnosed by stereotactic mammography guided biopsy (mammotomy). The ability of the sensor to distinguish between DCIS and normal breast tissue elements seems promising for using this type of sensor also in mammotomies.
Patients with ADH that were diagnosed on a CNB will be found to have in up to 25 % DCIS or even IDC present at open biopsy. Therefore patients with ADH on CNB are routinely sent for an open surgical biopsy [
18‐
21]. A more accurate CNB can reduce the number of unnecessary open biopsies in these patients. The ability of the sensor to differentiate between normal breast tissue, ADH and DCIS, can potentially improve the accuracy of CNB and reduce the number of open biopsies.
The diagnostic capability of the sensor for differentiating between malignant and normal tissue is high, with a Sensitivity of 90.5 % and Specificity of 90.1 %. This is also reflected by the ROC curves, with an area under the curve of 0.95. The ability of the sensor to differentiate any abnormal tissue from normal tissue is also high, Sensitivity 88.6 % and Specificity 91.7 %. With the current false-negative rates of CNB, this level of sensitivity has the potential of reducing the false negative rates in biopsy procedures to below 1 %.
The detection sensitivity of the sensor is dependent on the feature size of the malignant tissue. A 90.2 % Sensitivity features of at least 0.8 m in size, down to 66.8 % sensitivity for features smaller than 0.5 mm. It is anticipated that most biopsied malignancies will be of at least 1 mm in size, as, based on final pathology, most all of the malignant lesions are found to be larger than 3 mm.
There are some limitations to this type of sensor design. The manufacturing process of the sensor has to account for potential chemical modifications (over time) of the senor face, which can affect the reflected signals. Arranging sensors in a tightly packed array provides an additional challenge in isolating the electrical response of the individual sensors from their neighbors. The tissue has to be in direct contact with the sensor. Therefore, a practical device will need good attachment to substrate, as any (air) voids between the sensor and the tissue will skew the impedance measurement results. The very low penetration depth may limit the scope of applications in which this type of sensor will provide benefit. The sensor is gold-plated, with gold known not to be durable with regards to mechanical abrasion. In a biopsy procedure each sensor would be used for a limited duration, typically no more than a few minutes. Also, breast tissue is soft, and therefore it is not expected that the sensor structure will be mechanically effected during use.