Real-Time Differentiation Between Benign and Malignant Breast Tumors and Other Tissues Using Dielectric Properties

Introduction

Breast cancer stands as the most commonly diagnosed cancer and the second leading cause of cancer-related deaths among women [1]. While mortality rates from breast cancer are on the decline in high-income countries [2], both incidence and mortality rates are on the rise in the developing world [1]. In China alone, over 1.6 million individuals are diagnosed with breast cancer each year, and approximately 1.2 million succumb to the disease annually, representing 12.2% of all new breast cancer diagnoses and 9.6% of all breast cancer deaths globally [3]. Despite the prevalence of breast cancer, most clinical breast changes in women are benign [3] with only 3% to 6% of women with clinical symptoms being diagnosed with breast cancer [4]. It is estimated that around 50% of women will experience a benign breast lesion at some point in their lives [5], with these changes being more prevalent among women of childbearing age, particularly between the ages of 30 and 50 [6]. Distinguishing between benign and malignant breast lesions typically involves a thorough medical history review, along with clinical, imaging, and histological examinations. Diagnostic procedures, ranging from fine needle biopsies to minimally invasive breast surgeries, are essential for removing and histologically examining the affected tissue. This process is crucial for determining the nature of the lesion, deciding if further surgical intervention is necessary, and discussing treatment plans based on histological findings from paraffin-sectioned tissues during subsequent consultations. Should histological analysis reveal a malignant lesion, secondary surgery may be required.

Furthermore, given that breast cancer is predominantly a locoregional disease spreading via the lymphatic system, axillary lymph node dissection (ALND) has been a cornerstone of treatment for the last century [7]. However, to mitigate the risks of lymphedema, pain, and nerve damage associated with ALND, sentinel lymph node biopsy has emerged as the standard approach for clinically negative lymph nodes. The American College of Surgeons Oncology Group’s randomized clinical trial Z0011 (ACOSOG Z0011) demonstrated that for patients with limited metastatic breast cancer undergoing breast conservation and systemic therapy, sentinel lymph node dissection alone did not compromise survival compared with ALND [8]. The evaluation of axillary sentinel lymph node metastases is critical in deciding the extent of ALND that is needed. Imaging techniques such as mammography, computed tomography (CT), magnetic resonance imaging (MRI), and ultrasonography are routinely employed to assess the characteristics of axillary lymph node metastasis preoperatively [9]. Among these, ultrasonography stands out as the preferred method for axillary lymph node evaluation and image-guided lymph node interventional therapy [10]. In clinical practice, surgeons typically remove the axillary sentinel lymph nodes first, conduct a rapid frozen section biopsy, and then decide on proceeding with ALND based on the biopsy’s findings regarding metastatic lymph nodes. However, the frozen section biopsy process can take upwards of half an hour, with examination times increasing with the number of lymph nodes, thereby elevating anesthesia risks and diminishing surgical efficiency.

Given these challenges, there is a critical need for techniques that can provide real-time differentiation between benign and malignant tissues during the initial surgery. Such techniques would not only minimize the necessity for subsequent operations and reduce associated anesthesia risk, but would also significantly aid surgeons in formulating treatment strategies and conserving medical resources.

Dielectric properties, including dielectric relative permittivity (ɛ) and electrical conductivity (σ), are fundamental characteristics of human tissues [11], influenced by their water, protein, and ion content [12]. Bayford et al [13] have documented the conductivity and relative permittivity of both malignant and normal human tissues – including colon, kidney, liver, lung, mammary gland, and muscle tissue – across frequencies ranging from 50 to 900 MHz. Their findings revealed that differences in both conductivity and relative permittivity were frequency-dependent. Leveraging these distinctions, various diagnostic and therapeutic approaches have been devised, including microwave imaging [14] and simulated annealing probabilistic neural networks [15]. Reflecting on the documented disparities in dielectric properties across gastric cancer, colorectal cancer, breast cancer, adjacent tissues, and different thyroid nodules [16–18], Yu et al [19] observed significant increases in the dielectric properties (both permittivity and conductivity) of ex vivo intrathoracic lymph nodes in lung cancer at specific frequencies (64, 128, 298, 433, and 915 MHz, and 2.45 GHz). Additionally, Yu and Sun assessed the dielectric properties of 43 lymph nodes over a frequency span of 50 MHz to 20 GHz, albeit in specimens sourced from porcine mesenteric lymph nodes [19].

Despite these findings, the application of dielectric properties in distinguishing between benign and malignant breast tissues and lymph nodes remains underexplored. Given the complexity and variability of breast cancer, there is a need for further research to determine whether these dielectric measurements can serve as a rapid and accurate diagnostic tool during surgery. The present study aims to explore the potential of dielectric properties (including permittivity and conductivity) in distinguishing between benign and malignant breast tissues and lymph nodes by measuring these properties across different tissue types. The primary objective is to evaluate the effectiveness of these dielectric properties at various frequency ranges, particularly in their applicability for real-time intraoperative diagnosis. Based on existing literature, this study expands the frequency range to 10 MHz to 4 GHz and combines in vivo and ex vivo experiments to provide a more comprehensive analysis of tissue dielectric properties. Additionally, we have validated the potential of radio frequency technology for in vivo detection in a nude mouse model for the first time, laying the foundation for further research and clinical applications in this field. The study seeks to determine whether these measurements can serve as a rapid and accurate diagnostic tool during surgery, thereby reducing the need for secondary operations and enhancing surgical efficiency.

Material and Methods

MATERIAL AND MEASUREMENT SYSTEM:

The measurements were performed in the operating rooms of Zhujiang Hospital, Southern Medical University, Guangzhou, China, with a measurement system consisting of a laptop computer, a network analyzer (AV3680A), and an open-ended coaxial probe (OECP) with a characteristic impedance of 50 Ω (UT-41-50). All tested tissues were removed from the patients by experienced surgeons at Zhujiang Hospital, Southern Medical University, and the ex vivo tissues were analyzed within 10 minutes.

Suspected benign breast tumor tissues were removed by ultrasound-guided EnCor vacuum-assisted minimally invasive rotary mastectomy. Typical tissues with hard and flexible texture were selected by touch for measurement. The lymph nodes and cancer specimens were removed by modified radical mastectomy with fluorescence endoscopy, blood and adipose tissues on the surface were removed, and the internal tissues were tested by cutting open the lymph nodes. Breast tissues 3 cm from the cancer boundary were also analyzed. The temperature of all specimens and the room temperature were measured with a digital thermometer.

The whole experimental protocol was reviewed and approved by the ethics committee of Zhujiang Hospital, Southern Medical University (Approval No.: 2020-KY-072-02).

ANIMAL EXPERIMENTS:

Twenty 4-week-old nude mice were selected for this study due to their immunodeficient status, which makes them ideal for human tumor cell transplantation without immune rejection. Nude mice lack a thymus and therefore cannot produce T cells, allowing for the growth of human cancer cells and making them a widely used model in cancer research. The mice were numbered and randomly divided into 4 feeding cages for isolation and adaptive feeding for 2 weeks. The MCF-7 human breast cancer cell line was chosen for its well-documented use in breast cancer research. MCF-7 cells are estrogen receptor-positive, which is representative of a common subtype of breast cancer. The cells were cultured in Dulbecco’s modified Eagle medium (DMEM), a high-glucose medium containing 10% calf serum + 1% penicillin and streptomycin. After stable passage, a cell suspension with a density of approximately 6–8 million cells/mL was prepared with normal saline. In each feeding cage, 4 nude mice were randomly injected with 0.2 mL cell suspension as the experimental group, and the other nude mice were injected with 0.2 mL normal saline as the control group, fed with a normal diet, and provided with drinking water with a barrier system.

Two weeks later, some of the nude mice had developed tumors measuring approximately 2 mm on the back. The mice continued to be fed until the diameter of the transplanted tumors was measured to be 5–10 mm. Among the 20 experimental nude mice, 16 underwent cell suspension transplantation, of which 13 demonstrated tumorigenesis.

The tumor-bearing nude mice were anesthetized with isoflurane by an animal gas anesthesia machine, and the dielectric properties of the tumor-bearing and normal skin of the nude mice were measured by OECP radio frequency waves. The above operations were completed in the animal experiment center of Zhujiang Hospital, Southern Medical University according to a protocol approved by the animal ethics review committee (Approval No.: LACE-2021-029) (Figure 1).

CALIBRATION METHOD:

The calibration procedure was conducted with an open circuit, a short circuit, samples of deionized water, a 0.9% NaCl solution, methanol, and absolute ethanol before use with the specimens [20,21]. Additionally, the temperatures of the 4 liquids were also measured and recorded. Calibration steps, such as the open circuit and short circuit, are essential to establish reference points for the measurement system. An open circuit, where the probe is suspended in air, represents a condition of no electrical continuity, while a short circuit, achieved by placing the probe against a piece of tin foil, represents complete electrical continuity. These extremes help ensure that the system is correctly calibrated to measure the full range of possible dielectric properties. Additionally, using liquid samples like deionized water and saline solution helps account for variations in the dielectric properties of different substances, further refining the accuracy of the measurements. After polishing and removing the oxide layer on the probe tip with sandpaper, the probe was suspended in the air to obtain an open circuit, then placed close to the surface of a piece of tin paper to obtain a short circuit [22]. To avoid the influence of air gaps caused by manual operation on the error of the open-end coaxial method [23], any small bubbles that formed at the tip of the probe when it was inserted into the liquid were removed quickly. To ensure that the probe tip was in full contact with the measured object, an appropriate amount of pressure was applied without allowing the probe to be inserted into the tested object. Similarly, it was very important to select flat detection points to obtain correct data and to wait several additional seconds after fully contacting the measured objects until a stable Smith chart was shown on the network analyzer and laptop screen for each measurement.

DATA ANALYSIS:

The relative permittivity (ɛr’) and conductivity (σ) of tissues were obtained using the above-described methods. Generally, the complex permittivity of a tissue is represented by the following formula [24]:

where ɛ_r* is the complex permittivity of each sample, which depends on the frequency; ɛ_r’ is the relative permittivity of the tissue which represents the biological material under electromagnetic excitation; and ɛ_r is the loss factor, which indicates the loss of electromagnetic energy in the sample material due to direct-current (DC) conductivity and the various polarization mechanisms of other substances in the sample. In practice, dielectric parameters are typically described using ɛ_r’ and σ, where σ=ωɛ_r,, ɛ₀ and ω is the angular frequency (2πf).

DOUBLE-POLE COLE-COLE FITTING:

To improve data accessibility, the Cole-Cole model [24] was applied to the data over frequencies from 1 MHz to 4 GHz. The theoretical calculation of the tissue complex permittivity value was obtained by the Cole-Cole model. Data for the dielectric properties were then simplified into 8 parameters using the following expression of the double-pole Cole-Cole model:

where ɛ∞, Δɛ1, Δɛ2, τ1, τ2, α1, α2, and σs are the Cole-Cole model parameters estimated from the experimental data. These coefficients were computed using the mean of the various breast tissues and lymph nodes using MATLAB software (The Math Works, Inc., Natick, Massachusetts, United States).

We chose the Cole-Cole model because of its widespread application in analyzing the dielectric properties of tissues, as well as its superior ability to describe relaxation phenomena across different frequencies. Compared with other models, the Cole-Cole model offers greater flexibility in fitting the frequency dependence of complex permittivity, making it particularly well-suited for dielectric studies of biological tissues.

HISTOPATHOLOGICAL EXAMINATION:

All the tumor and adjacent tissues removed from radical mastectomy were embedded in paraffin and sectioned by the Pathology Department of Zhujiang Hospital, Southern Medical University, and these sections were used for diagnosis by experienced pathologists. For the sentinel lymph nodes removed by endoscopic axillary sentinel lymph node biopsy, frozen pathological biopsy was performed during the operation, and the examination time was recorded from receipt of the specimens by the pathology department to the issuance of the report. Routine paraffin pathological biopsy was performed to verify the results after the operation.

STATISTICAL ANALYSIS:

Statistical analyses were performed using the SPSS 20.0 software package (IBM, Chicago, IL). For comparing dielectric properties between different tissue types, nonparametric tests were employed due to the non-normal distribution of the data. The Mann-Whitney U test was used for pairwise comparisons of dielectric permittivity and conductivity between the different groups (eg, metastatic vs non-metastatic lymph nodes, normal vs diseased breast tissues). Statistical significance was determined using a 2-tailed P value, with a threshold set at P<0.05 for general significance and P<0.01 for more stringent comparisons. All statistical results are reported with corresponding P values, indicating the level of significance in the observed differences.

Results

MEASUREMENT INFORMATION:

The present study encompassed 97 female patients, aged between 20 and 79 years. Among them, 53 underwent either endoscopic modified radical mastectomy or breast-conserving surgery, while 47 were treated with ultrasound-guided EnCor minimally invasive rotary-cut surgery (notably, 3 patients received both treatments). All surgical procedures were conducted at the Breast Department of Zhujiang Hospital, Southern Medical University, in 2021. This resulted in the collection of 179 breast tissue specimens and 190 sentinel lymph nodes. The breast tissue specimens were categorized into adenoses, fibroadenomas, fibrocystic lesions, and tissues located 3 cm from the tumor boundary. Meanwhile, the lymph node specimens were classified into metastatic and non-metastatic lymph nodes (Figure 2).

Examples of the pathology slides for the different tissue types (both normal and diseased) that were encountered in the pathologist reports are shown in Figure 3.

The measurement process was completed within 10 minutes of specimen resection. The temperature of the operating rooms during the measurement was 21.5±1.3°C, the average temperature of the breast tissue was 23.3±2.0°C, and that of the lymph nodes was 20.6±1.4°C.

COLE-COLE FITTING RESULTS:

The fitting coefficient used is shown in Figure 4, representing the fitting curves for non-metastatic lymph nodes, metastatic lymph nodes, and the different types of breast tissues. The data measured for both lymph nodes and the various breast tissues fit closely to the double-pole Cole-Cole model.

COMPARISON OF THE DIELECTRIC PROPERTIES OF BREAST NODULES AND LYMPH NODES:

Reflecting on the literature, it is noted that data stability at low frequencies is suboptimal [22]; hence, this study commenced measurements and calculations of the dielectric permittivity and conductivity of tissue samples starting from 10 MHz. Within the frequency range of 10 to 100 MHz, the dielectric constants for the 7 distinct breast tissue types exhibited greater complexity. Significant differences in dielectric permittivity among these tissues were observed at frequencies of 20, 50, 70, and 100 MHz (P<0.05). However, in pairwise comparisons, the frequencies at which significant differences occurred varied between each pair of tissues. At 20, 50, 70, and 100 MHz, notable differences were identified between normal tissue and cancer, adenosis, and fibroadenoma (P<0.05), as well as between adenosis and fibroadenoma vs cancer (P<0.05). Fibrocystic lesions and fibromas, as well as cancer and adenosis, exhibited significant disparities at 70 MHz and 100 MHz (P<0.05). Only at 20 MHz and 100 MHz were differences observed between normal tissue and fibrocystic lesions, and between fibroma and cancer, respectively (P<0.05) (Table 1).

The changes in the dielectric permittivity of the lymph nodes and various breast tissues under different frequency values are shown in Figure 4. In the frequency range of 10 MHz to 100 MHz, a significant difference in dielectric permittivity was observed between metastatic and non-metastatic lymph nodes (P<0.01). Beyond 100 MHz, the disparity in dielectric properties between these 2 lymph node types began to diminish as the frequency increased.

Contrary to dielectric permittivity, the electrical conductivity of tissues exhibited an increase in frequency, ranging from 10 MHz to 4000 MHz, as shown in Figure 4. Significant differences in conductivity were observed between metastatic and non-metastatic lymph nodes, as well as between normal breast tissues and other diseased breast tissues (P<0.05). However, among the various breast lesions, no significant differences in conductivity were detected (P>0.05).

The findings indicate that under stimulation frequencies of 20 to 100 MHz, the dielectric properties of tissues – particularly dielectric permittivity – more accurately reflected the pathological states of the sentinel lymph nodes and different breast tissue types, proving to be more indicative than electrical conductivity.

COMPARISON OF IN VIVO RESULTS:

At frequencies of 100 MHz and 200 MHz, notable differences were observed in the dielectric permittivities between tumor-bearing and normal skin in nude mice (P=0.044). Additionally, conductivity exhibited significant variations with increasing frequency (P=0.029). These findings suggest that alterations in tissue dielectric properties occur in vivo and can be detected using radio frequency technology, as illustrated in Figure 5.

Discussion

LIMITATIONS:

In this study, although we successfully measured the dielectric properties of breast tumors and lymph nodes and achieved significant results, there are still some important limitations that need to be considered. First, we did not fully account for the heterogeneity of blood perfusion within breast tissue and tumors, which is crucial for treatment effectiveness [37]. The traditional Pennes bioheat equation assumes that blood perfusion is homogeneous or isotropic, which may limit the accuracy of thermal energy deposition modeling within the tumor [38]. In reality, blood perfusion within tumors is highly heterogeneous, with regions of high, moderate, and low perfusion, and areas of no perfusion or necrosis [39]. These differences in thermal energy distribution can significantly impact treatment outcomes. The heterogeneity of blood perfusion is closely related to tumor hypoxia, with hypoxic regions typically exhibiting lower blood perfusion and potentially responding less effectively to hyperthermia or other treatments [40]. Moreover, our study did not take into account the role of healthy tissue regeneration at the tumor margins in regulating thermal damage. Under quasi-static thermal conditions, healthy tissue continuously regenerates due to the sustained supply of oxygen via arterial blood, which helps balance the accumulation of thermal damage [41]. At the interface between the tumor and healthy tissue, the regeneration of healthy cells not only repairs damaged tissue but also triggers an immune response to limit the further accumulation of thermal damage.

Moreover, while this study identified differences in dielectric properties among various types of breast tumors and lymph nodes across different pathological states – offering a potential tool to assist clinicians in determining the scope of surgical resection and lymph node dissection—a reliable computerized intelligent algorithm is essential for translating specimen data into real-time, intuitive results. Achieving this requires extensive data collection, analysis, and classification, alongside relevant software development and research, necessitating multidisciplinary collaboration.

Conclusions

This study validated the potential of dielectric properties in distinguishing between benign and malignant tissues by measuring the dielectric characteristics of breast tumors and lymph nodes within the frequency range of 10 MHz to 100 MHz. Our results demonstrate that dielectric permittivity and conductivity within this specific frequency range can significantly differentiate tissues with varying pathological states, providing a theoretical foundation for real-time intraoperative diagnosis. The application of these dielectric measurements during surgery could reduce the need for secondary operations, enhance surgical efficiency, and ultimately improve patient outcomes.