Innovation for using dielectric properties to distinguish lung tumor from normal lung tissues and preliminary exploration for the relevance with metabolic features
Original Article

Innovation for using dielectric properties to distinguish lung tumor from normal lung tissues and preliminary exploration for the relevance with metabolic features

Lijuan Wang1# ORCID logo, Shaobin Li2#, Hu Zhou3, Huajie Li4, Nengke Lin3, Qiang Huang5, Hongfeng Yu6, Zhizhi Wang2, Zhiming Chen2, Yin Zhang1 ORCID logo, Di Lu2

1Department of Nuclear Medicine, Nanfang Hospital, Southern Medical University, Guangzhou, China; 2Department of Thoracic Surgery, Nanfang Hospital, Southern Medical University, Guangzhou, China; 3The First School of Clinical Medicine, Southern Medical University, Guangzhou, China; 4The Second School of Clinical Medicine, Southern Medical University, Guangzhou, China; 5Equipment Department, The First Affiliated Hospital of Guangzhou Medical University, Guangzhou, China; 6Department of Medical Equipment, Hangzhou Fuyang Traditional Chinese Medicine Orthopedic Hospital, Hangzhou, China

Contributions: (I) Conception and design: Y Zhang, D Lu; (II) Administrative support: Y Zhang, D Lu; (III) Provision of study materials or patients: S Li, Z Wang, Z Chen, D Lu; (IV) Collection and assembly of data: L Wang, S Li, H Zhou, H Li, N Lin, Q Huang, H Yu, Z Wang, Z Chen; (V) Data analysis and interpretation: L Wang, S Li, H Zhou, H Li, N Lin, Q Huang, H Yu, Z Wang, Z Chen; (VI) Manuscript writing: All authors; (VII) Final approval of manuscript: All authors.

#These authors contributed equally to this work.

Correspondence to: Di Lu, MD. Department of Thoracic Surgery, Nanfang Hospital, Southern Medical University, 1838 Guangzhou North Road, Guangzhou 510515, China. Email: ludi1989@smu.edu.cn; Yin Zhang, MD. Department of Nuclear Medicine, Nanfang Hospital, Southern Medical University, 1838 Guangzhou North Road, Guangzhou 510515, China. Email: lzyxytwzy@163.com.

Background: Our previous studies successfully established deep learning methods to differentiate metastatic lymph nodes from lymphadenitis in patients with lung cancer using dielectric properties. In this study, we aimed to develop a simpler, more interpretable method to differentiate lung tumor tissue from normal lung tissue. Additionally, we explored the correlation between [18F] fluorodeoxyglucose (FDG) positron emission tomography (PET)/computed tomography (CT) parameters and the dielectric properties.

Methods: Dielectric properties of lung tumors were measured using an open-ended coaxial probe across a frequency range of 1–4,000 MHz after excision from patients. Normal lung tissues were also measured for comparison. Based on the data trends, the permittivity curves were fitted to logarithmic functions, while the conductivity curves were fitted to exponential functions. The intercepts and slopes of these permittivity and conductivity functions were then used to distinguish lung tumors from normal lung tissues. These parameters were also used to assess the correlation between dielectric properties and PET parameters.

Results: Totally 21 lung tumor tissues and 19 normal lung tissues from 21 patients were included in this study. The permittivity and the conductivity values were acquired at frequency range from 1 to 4,000 MHz and then fitted into logarithmic function [y = slopep × ln(x) + interceptp] and exponential function [y = interceptc × e(slopec×x)], respectively. The goodness of fit for all the conductivity function (R2>0.95) and most of the permittivity function (R2>0.75) were well. The absolute value of the functions parameters of lung tumor were higher than that of normal lung tissues (P<0.05 for all). Receiver operating characteristic (ROC) analyses indicated that the interceptc, slopep and interceptp could effectively distinguish lung tumor from normal lung tissues [all area under the curve (AUC) values >0.79]. Both the slopep and interceptp were statistically correlated with PET parameters, such as standard uptake value (SUV)max, SUVmean, SUVmin, SUVpeak and standard deviation of SUV (SUVSD) (P<0.01 for all). Differences of variable coefficient of SUV (CV_SUV) between the high- and low-slopep and interceptp groups were both statistically significant (t=2.22 and 2.654, P=0.04 and 0.02, respectively). There were no statistical correlations between slopec, interceptc and PET parameters.

Conclusions: The fitted functions of dielectric properties, including permittivity and conductivity, were able to distinguish lung tumors from normal lung tissues. Further study should be carried out to explore the ability of dielectric properties for differentiating malignant lung tumor from benign lung disease. The metabolic activity and heterogeneity of the tumors, as reflected by [18F] FDG PET/CT, were associated with the permittivity of tumor tissues but not the conductivity. Deeper study could be carried out to explore the mechanism for this correlation.

Keywords: Dielectric properties; permittivity; conductivity; lung tumor; [18F] fluorodeoxyglucose positron emission tomography/computed tomography ([18F] FDG PET/CT)

Submitted May 12, 2025. Accepted for publication Aug 08, 2025. Published online Oct 17, 2025.

doi: 10.21037/jtd-2025-934

Highlight box

Key findings

• We developed a novel method by using dielectric properties for distinguishing lung tumors from normal lung tissues. We also discovered the correlation between dielectric properties and fluorodeoxyglucose positron emission tomography/computed tomography features.

What is known and what is new?

• Our previous studies revealed the feasibility of using dielectric properties for distinguishing lymph node metastasis from lymphnoditis in patients with lung cancer.

• We further discovered that dielectric properties could also be used to distinguish lung tumors from normal lung tissues. And we developed a novel and simpler method by utilizing the dielectric properties.

What is the implication, and what should change now?

• Faster and more precise decision can be made during surgery using dielectric properties than traditional frozen pathology. Further studies should be carried out to explore the correlation between physiological and physical features of lung tumor.

Introduction

For a long time, research on tumor theranostics has primarily focused on biological aspects. However, in recent years, there has been increasing interest in biomechanics (1-3), bioelectronics (4), and other interdisciplinary approaches (4,5). It has become increasingly evident that a comprehensive understanding of tumor development, progression, and prognosis requires an integrated perspective that spans multiple dimensions.

The dielectric properties are measurements of the response to the electric field for certain substances (6,7). Our previous studies revealed that the dielectric properties contributed to the differential diagnosis between the metastatic lymph nodes and the lymphadenitis in patients with lung cancer (8,9). However, the computational complexity and limited explainability of existing analytical approaches restrict their clinical translation. Although the interpreting time was obviously shorter than frozen biopsy using our previous approaches, the accuracy could be potentially improved. A simpler approach is better suited for the time-sensitive environment during surgery and a more explainable model would likely be more widely accepted in clinical practice.

Tumor cells exhibit a high metabolic rate, changes in water content, and electrolyte imbalances, which lead to alterations in dielectric properties (10-12). But few studies have confirmed the association between tumor metabolism and dielectric properties (10). When reviewing data from our previous studies, we found that some patients had undergone [18F] fluorodeoxyglucose (FDG) positron emission tomography (PET)/computed tomography (CT) examinations. [18F] FDG PET/CT is widely used in tumor diagnosis, staging and response assessment, especially in lung cancer (13,14). In general, the [18F] FDG accumulation on PET/CT reflects the glucose metabolism activity. Associations between [18F] FDG PET parameters and dielectric properties can be evidences of the relationship between tumor metabolism and dielectric properties. Moreover, an overall evaluation of lung tumors, integrating pre-operative PET/CT characteristics and intra-operative dielectric properties, could offer a more precise therapeutic strategy for patients.

In this study, we aim to explore a simpler and more explainable approach to distinguish lung tumor from normal lung tissues using dielectric properties. Additionally, we investigate the relationship between dielectric properties and [18F] FDG PET parameters preliminarily. We present this article in accordance with the STARD reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-934/rc).

Methods

Patient data

Patients who underwent PET/CT before surgery and dielectric properties measurement during surgery were retrospectively analyzed. All the patient data included in this study were obtained from a previously approved study (8,9). The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the ethics committee of the Nanfang Hospital, Southern Medical University (No. NFEC-2017-070) and individual consent for this retrospective analysis was waived.

Dielectric properties measurement

The detailed method of dielectric properties measurement was described in our previous studies (8,9). The permittivity and the conductivity values were acquired at frequency that range from 1 to 4,000 MHz. Because of the heterogeneity of dielectric properties within tumor tissues, we employed a multi-point measurement approach to better capture the overall characteristics of the tumor tissues. Four random points were measured for each tumor tissue, and the mean value was calculated. The normal lung tissues were also measured far from the lung tumors in the same lung lobe. Normal lung tissue refers to the tissue that locate within the same lung lobe at a distance of 3 to 5 cm away from the tumor during surgery, and which shows no suspicious tumor lesions by observation and touch. Within a frequency range of 100 MHz, if the difference between the maximum and minimum values of conductivity exceeds 0.1, or the difference between the maximum and minimum values of permittivity exceeds 100, this was regarded as a significant fluctuation. Data with significant fluctuation were abandoned.

PET/CT parameters

[18F] FDG PET/CT images were obtained from the routine PET/CT examinations using Biograph mCTx (Siemens, Forchheim, Germany). The images were analyzed by a practiced PET physician in the post-processing workstation Syngo MMWP (Siemens). Then the regions of interest (ROIs) of lung tumors were delineated and the frequently-used PET parameters, which included standard uptake value (SUV)max, SUVmean, SUVmin, SUVpeak, standard deviation of SUV (SUVSD), metabolic tumor volume (MTV) and total lesion glycolysis (TLG) were recorded. ROIs of normal lung tissues were also drawn far from the lung cancer in the same lung lobe and PET parameters were recorded. The variable coefficient of SUV(CV_SUV) was calculated by the following formular: CV_SUV=(SUVSD/SUVmean)×100%. Target to background ratio (TBR) was calculated by the following formular:  TBR=(SUV of lung tumor/SUV of normal lung tissue)×100%.

Curve fitting

Previous studies (8,9) have shown that the permittivity decreases with increasing frequency, whereas the conductivity tends to rise with higher frequencies. After eliminating the data that have obvious fluctuations (i.e., the data with frequencies ranging from 1 to 50 MHz), the curve of permittivity tended to be logarithmic function and the curve of conductivity tended to be exponential function. In this study, we tried to explore the pattern of permittivity and conductivity values through curve fitting for each tissue. The curve fitting processes were accomplished by Python (version 3.13.2). For ease of understanding, the curve of permittivity was fitted into logarithmic function expressed by the formula of  y=slopep×ln(x)+interceptp. The curve of conductivity was fitted into exponential function expressed by the formula of y=interceptc×e(slopec×x). The slope term indicates the rate at which the curve increases or decreases and the intercept term represents the initial value of the curve. The goodness of fit (expressed by R2) was calculated for each function to evaluate how well the fitting curve fits the observed values.

Differences between lung tumors and normal lung tissues were evaluated statistically by using the slopes and intercepts of permittivity and conductivity functions. Besides, the difference values between lung tumors and normal lung tissues were calculated and then converted to a scatter plot. Subsequently we explored the pattern of scatter plots.

Due to the low [18F] FDG uptake of ground-glass nodules and partial solid nodules on PET imaging, these lesions would be considered outliers if included in the analysis. As a result, they were excluded from the analysis of the association between dielectric properties and PET parameters. We then investigated the correlation between PET/CT parameters and the slopes and intercepts of the permittivity and conductivity functions. Additionally, patients were divided into two groups based on the median of slopes and intercepts of these permittivity and conductivity functions, respectively. The differences in PET parameters between the two groups were then analyzed.

Statistical analysis

Patients’ characteristics were exhibited by amount and proportion. Paired-samples t-tests were used to assess the difference of conductivity and permittivity functions between lung tumors and corresponding normal lung tissues. Receiver operating characteristic (ROC) curve analyses were used to evaluate the ability of slopes and intercepts of conductive function and permittivity function to differentiate lung tumor from normal lung tissues. Independent sample t-tests were used to evaluate the differences of the parameters between adenocarcinoma and non-adenocarcinoma tissues.

PET parameters were exhibited by range and median. Relationships between PET parameters and the parameters of conductivity and permittivity functions were evaluated by Pearson correlation analyses. We further divided the patients to high-parameter of function group and low-parameter of function group and then assessed the differences of PET parameters between these two groups.

All statistical analyses were completed by IBM Statistical Package for the Social Sciences (SPSS) Statistics 20 (IBM Corp., Armonk, NY, USA). P value less than 0.05 was considered statistically significant.

Results

Totally 21 patients (21 lung tumor tissues and 19 normal lung tissues) from October 2016 to June 2018 were included in this study. The patient characteristics were revealed in Table 1.

Table 1

Patients characteristics

Characteristics Value
Gender
   Male 15 (71.43)
   Female 6 (28.57)
Age (years) 61±9.11 [46–81]
   ≥60 13 (61.90)
   <60 8 (38.10)
Height (cm) 165±8.09 [147–181]
Weight (kg) 61.5±7.62 [45–75]
Pathology
   Adenocarcinoma in situ 2 (9.52)
   Adenocarcinoma 10 (47.62)
   Squamous cell carcinoma 8 (38.10)
   Sarcoma 1 (4.76)
Lesion location
   Right upper lung 12 (57.14)
   Right lower lung 1 (4.76)
   Left upper lung 4 (19.05)
   Left lower lung 4 (19.05)
Long diameter (cm) 2.8±2.76 [0.8–13.4]
   ≥3 10 (47.62)
   <3 11 (52.38)
T stage†
   Tis 2 (9.52)
   T1 4 (19.05)
   T2 10 (47.62)
   T3 4 (19.05)
   T4 1 (4.76)

Data are presented as n (%) or median ± SD [range]. †, the patient with sarcoma was defined as T3 stage. SD, standard deviation; T, tumor.

Curve fitting results of permittivity and conductivity

By means of computer-assisted curve fitting, we finally constructed functions of permittivity and conductivity for each patient. The parameters of functional formulas were showed in Table 2. The permittivity and conductivity values of normal tissue for patient No. 15 exhibited significant fluctuation (Figure 1). So, this patient was not included in the following analyses. The intercepts of conductivity functions differed from each other. All the R2 value for the conductivity function were above 0.95. The slopes and intercepts of permittivity functions differed from each other too. All the R2 value for the permittivity function were high than 0.71.

Table 2

Curve fitting results of conductivity and permittivity for malignant and benign tissues for each patient

Patient No. Conductivity [y=interceptc×e(slopec×x)] Permittivity [ y=slopep×ln(x)+interceptp]
Normal lung tissue Lung tumor tissue Normal lung tissue Lung tumor tissue
Slope Intercept R² Slope Intercept R2 Slope Intercept R² Slope Intercept R2
1 3.9394×10−10 0.5686 0.9943 3.9629×10−10 0.7304 0.9956 −5.5487 161.0996 0.8184 −6.2501 188.6242 0.7757
2 Untested 3.4575×10−10 0.9521 0.9942 Untested −4.7265 160.5832 0.7796
3 4.2165×10−10 0.5192 0.9806 4.2637×10−10 0.6690 0.9882 −2.3786 89.3199 0.8358 −3.0139 115.8870 0.7927
4 3.6783×10−10 0.5733 0.9917 3.6509×10−10 0.6041 0.9892 −4.2596 129.2803 0.8359 −4.5722 141.1800 0.7376
5 3.7847×10−10 0.6406 0.9902 3.7460×10−10 0.7575 0.9933 −5.7041 166.9657 0.8214 −6.4040 190.7679 0.8007
6 3.7620×10−10 0.6736 0.9977 3.8437×10−10 0.8045 0.9972 −4.9211 152.7830 0.7979 −6.3061 191.8817 0.8000
7 4.0533×10−10 0.7460 0.9948 3.7336×10−10 0.7606 0.9963 −4.5611 152.8965 0.8310 −4.9693 158.3333 0.8020
8 3.9566×10−10 0.5051 0.9786 3.8172×10−10 0.8278 0.9887 −3.5151 113.7340 0.7051 −6.1879 191.9757 0.7101
9 3.5620×10−10 0.5623 0.9946 3.8423×10−10 0.7670 0.9983 −4.7917 140.2617 0.8176 −5.8017 180.6276 0.8028
10 3.7244×10−10 0.5734 0.9915 3.8239×10−10 0.7560 0.9933 −3.9682 124.2254 0.8753 −5.3315 167.5961 0.8029
11 3.6116×10−10 0.3603 0.9907 3.7912×10−10 0.8076 0.9962 −1.6827 60.7357 0.8357 −4.9201 161.5167 0.8166
12 3.9683×10−10 0.7630 0.9871 4.3033×10−10 0.6192 0.9771 −5.8770 180.0140 0.8254 −5.4187 164.9418 0.8235
13 3.9752×10−10 0.5831 0.9950 4.1146×10−10 0.7671 0.9961 −4.7375 145.0551 0.8707 −6.1419 188.8084 0.8476
14 3.8018×10−10 0.4895 0.9894 3.9583×10−10 0.7213 0.9909 −2.9446 97.7363 0.8767 −5.6503 174.8980 0.8141
15† 3.8210×10−10 0.6226 0.9314 4.7405×10−10 0.5305 0.9867 −4.9082 151.4103 0.6476 −5.0984 164.6277 0.8459
16 3.5627×10−10 0.6289 0.9584 3.8890×10−10 0.8110 0.9785 −3.2547 108.1309 0.8351 −5.2576 168.4498 0.8150
17 3.6753×10−10 0.5540 0.9948 3.8548×10−10 0.7535 0.9966 −4.0201 124.0719 0.7930 −5.6743 176.9445 0.8017
18 3.8681×10−10 0.4135 0.9740 3.9938×10−10 0.6928 0.9858 −2.2791 80.0751 0.7710 −5.7255 173.3488 0.8587
19 3.9250×10−10 0.6193 0.9735 3.8783×10−10 0.7011 0.9778 −3.5659 120.0613 0.7793 −3.9771 137.0555 0.8112
20 3.9554×10−10 0.6510 0.9940 3.9466×10−10 0.6718 0.9952 −6.5276 187.5023 0.7916 −6.6033 192.6734 0.7523
21 Untested 4.0986×10−10 0.6629 0.9957 Untested −7.1154 205.3351 0.7812

†, the permittivity values of normal lung tissue for patient No. 15 exhibited significant fluctuation.

Figure 1 PET/CT and dielectric properties for patient No. 15. (A) The lung tumor was marked by red arrows on [18F] FDG PET/CT imaging. (B) Curve of conductivity measurement of the lung tumor (orange curve) and normal lung tissue (blue curve). (C) Curve of permittivity measurement of the lung tumor tissue (orange curve) and normal lung tissue (blue curve). CT, computed tomography; FDG, fluorodeoxyglucose; PET, positron emission tomography.

The difference values scatter plots indicated that most of the difference values of permittivity decreased quickly as the frequency rises while the difference of conductivity gradually increased as the frequency rises, like the tendency of permittivity and conductivity themselves.

The relationship between dielectric properties and tissue nature

Most of the permittivity and conductivity values of malignant tissues were higher than those of benign tissues. The only exceptions are No. 7 and No. 12 patients (Figure 2). Paired-samples t-test revealed that the absolute value of slopes and intercepts of conductivity functions and permittivity functions for lung tumor tissues were higher than the normal lung tissues (P<0.05 for all, see Table 3).

Figure 2 PET/CT and dielectric properties for patient No. 7 and No. 12. (A-C) [18F] FDG PET/CT imaging, curve of conductivity measurement and curve of permittivity measurement for patient No. 7. (D-F) [18F] FDG PET/CT imaging, curve of conductivity measurement and curve of permittivity measurement for patient No. 12. The lung tumor was marked by red arrow. Orange curve and blue curve were representative of lung tumor tissues and normal lung tissues, respectively. CT, computed tomography; FDG, fluorodeoxyglucose; PET, positron emission tomography.

Table 3

Paired-samples t-test exhibiting the difference between dielectric properties and tissue nature

Object Lung tumor tissue
(mean ± SD)		 Normal lung tissue
(mean ± SD)		 t value P Cohen’s d
Slopec 3.956×10−10±2.532×10−10 3.835×10−10±1.748×10−10 −2.117 0.048 0.486
Interceptc 0.724±0.078 0.581±0.100 −4.454 <0.001 1.022
Slopep −5.456±0.919 −4.141±1.335 4.989 <0.001 1.176
Interceptp 170.306±21.541 129.664±34.651 −5.4 <0.001 1.273

SD, standard deviation.

ROC analyses indicated that not only the intercepts of conductivity functions but also the slopes and intercepts of permittivity functions could effectively distinguish lung tumor from normal lung tissues [all area under the curve (AUC) values >0.79, see Table 4 and Figure 3].

Table 4

Diagnostic accuracy and their precision of dielectric properties for distinguish lung tumors from normal lung tissues

Object AUC (95% CI) Standard error P
Slopec 0.634 (0.456–0.813) 0.091 0.157
Interceptc 0.867 (0.750–0.984) 0.06 <0.001
Slopep 0.793 (0.641–0.945) 0.078 0.003
Interceptp 0.852 (0.727–0.977) 0.064 <0.001

AUC, area under the curve; CI, confidence interval.

Figure 3 ROC curves analyses revealed the different AUCs of slopes and intercepts of conductivity function and permittivity function for differentiating lung tumor tissues from normal lung tissues. AUC, area under the curve; ROC, receiver operating characteristic.

When patients are categorized into adenocarcinoma and non-adenocarcinoma groups, statistically significant differences in slopes and intercepts of permittivity functions were observed between the two groups (t=2.71 and 2.43, P=0.02 and 0.03, respectively). But the difference of the interceptc between these two groups was not statistically significant (t=1.20, P=0.24).

The relationship between dielectric properties and PET parameters

The PET characteristics of the 21 lung tumor tissues were revealed in Table 5. Because of the low SUV of ground glass nodules and partial solid nodules for patient Nos. 4,5,9 and 16, these data were not included in following statistical analyses. Correlation analyses showed that there was no statistical correlation between slopec and interceptc and the PET parameters (all correlation coefficients less than 0.15 and all P>0.1). While both the slopep and interceptp were statistically correlated with some PET parameters, especially the SUVpeak (shown as the correlation coefficients heatmap, Figure 4A). The correlation coefficients of the slopes were a little higher than that of the intercepts. The representative scatter plot indicated that there were nearly linear correlations between the slopep and SUVpeak (Figure 4B). The MTV, TLG and TBR were not statistically correlated with the dielectric parameters (all correlation coefficients less than 0.30 and all P>0.1).

Table 5

PET characteristics for lung tumor tissues

Characteristics Range Median SD
SUVmax 0.80–27.08 15.51 7.85
SUVmean 0.51–18.90 9.48 5.01
SUVmin 0.37–10.83 6.23 3.15
SUVpeak 0.55–18.55 12.27 5.62
SUVSD 0.11–3.46 1.9 1.08
CV_SUV 0.16–0.30 0.23 0.03
MTV (cm3) 1.14–137.40 5.52 29.67
TLG 0.68–1,566.39 73.06 348.78
TBR_SUVmax 1.34–44.32 25.04 13.74
TBR_SUVmean 1.17–53.53 27.14 15.23

CV_SUV, coefficient of variation for SUV; MTV, metabolic tumor volume; PET, positron emission tomography; SD, standard deviation; SUV, standard uptake value; SUVSD, standard deviation of SUV; TBR, target to background ratio; TLG, total lesion glycolysis.

Figure 4 Correlation between permittivity function features and PET features. (A) Correlation coefficients heatmap revealed the correlation analyses results between slopep, interceptp and PET parameters. Blue color represented negative correlation, and red color represented positive correlation. The gradation of color represented the correlation coefficient, which was labeled inside the square by specific numerical value. Size of the circle also reflected the correlation coefficient. (B) The representative scatter plot indicated that there were linear correlations between the slopep and SUVpeak (P=0.02, r=−0.66). PET, positron emission tomography; SUV, standard uptake value; SUVSD, standard deviation of SUV.

Between the high- and low-slopec and interceptc groups, differences of all PET parameters were not statistically significant (all P≥0.27). So as the differences between high- and low-slopep group (all P≥0.07). While the differences of SUVmax, SUVmin, SUVmean, SUVpeak and CV_SUV between the high- and low-interceptp groups were statistically significant (0.05>P≥0.02, see Table 6), there were no statistically significant differences for other PET parameters (all P>0.10).

Table 6

Statistical P values of the differences between high- and low-function parameters groups

Group SUVmax SUVmin SUVmean SUVpeak SUVSD CV_SUV MTV TLG TBR_SUVmax
High- vs. low-slopec 0.71 0.73 0.99 0.73 0.72 0.98 0.27 0.38 0.85
High- vs. low-interceptc 0.71 0.73 0.99 0.73 0.72 0.98 0.27 0.38 0.85
High- vs. low-slopep 0.12 0.12 0.11 0.08 0.35 0.07 0.79 0.87 0.59
High- vs. low-interceptp 0.03Δ 0.04Δ 0.04Δ 0.02Δ 0.18 0.03Δ 0.89 0.94 0.18

Δ, statistical differences were discovered for these parameters. CV_SUV, coefficient of variation for SUV; MTV, metabolic tumor volume; SUV, standard uptake value; SUVSD, standard deviation of SUV; TBR, target to background ratio; TLG, total lesion glycolysis.

Discussion

In this study, we fitted the measured dielectric properties of lung tumors and normal lung tissues to corresponding exponential and logarithmic functions. We then analyzed the relationships between the parameters of these functions and the nature of the tissues, as well as the associations between these parameters and PET parameters. Our results showed that the slopes and intercepts of the permittivity and conductivity functions could effectively distinguish lung tumor tissues from normal lung tissues. Furthermore, statistical correlations were observed between the slopes and intercepts of the permittivity functions and some PET parameters, which are indicative of metabolic activity and heterogeneity.

For all patients except No. 7 patient and No. 12 patient, the permittivity and conductivity of lung tumor tissues were higher than that of normal lung tissues, which were consistent with previous studies (8,9,15). For No. 7 patient, there existed huge part of necrosis in the tumor, which was obvious on PET/CT images. Water content was increased obviously in the necrosis tissues. The data of malignant tissues may reflect the necrosis tissues and hence these data were lower than that of normal lung tissues (6,16). For another patient, we noted that the Ki-67 index of the tumor tissue was only 5%. This suggested that the proliferation rate of this tumor was relatively slow and it might be difficult to distinguish it from normal lung tissues (17). Measurement error could also be one of the potential reasons.

According to the literature, the dielectric properties of different tissues varied (16,18), and this serves as the theoretical foundation for differentiating lesion characteristics and delineating lesion boundaries. In the area of oncology, dielectric properties have been primarily used in detection and real-time intraoperative differentiation. In the case of skin cancer, the dielectric properties of malignant basal cell carcinoma of the skin were notably higher compared to that of healthy skin (19). Multiple studies have successfully developed intraoperative real-time monitoring techniques that capitalize on the dielectric properties to differentiate malignant tissues, normal tissues and metastatic lymph nodes (20,21). In gastrointestinal cancer surgeries, for example, the dielectric properties were proven to be effective in differentiating metastatic lymph nodes from non-metastatic lymph nodes, thereby enabling accurate identification of malignant tumor tissues (22). Similarly, in breast cancer research, significant differences in dielectric properties have been observed (21). These dielectric properties can be utilized to rapidly identify tissue types during surgical procedures. Dielectric properties hold the potential to be a surgical navigation tool, which could effectively reduce intraoperative miscuts and waiting time, thus enhancing the precision and efficiency of surgical operations.

Our study revealed that not only the slope of conductivity function, but also the slope and intercept of permittivity function could contribute to the differential diagnosis between lung tumor tissue and normal lung tissue. There were statistically significant differences of these three parameters between lung tumor tissue and normal lung tissue, among which the difference of the intercept of permittivity function is the greatest. Furthermore, AUC values of these three parameters for distinguishing lung tumor tissue from normal lung tissue were all relatively high, with the highest reaching 0.87. This indicated the well classified ability of this method, which was similar with the complicated methods used in the previous studies (8,9). FDG PET/CT was considered as a more precise tool than common CT, especially for delineating lesion boundaries. Combining intraoperative dielectric properties with preoperative FDG PET/CT could further improve the precision. Unfortunately, the sample size in this study is limited and there was a lack of data on benign lung tumors. This limitation restricts the comprehensiveness of our analysis.

Scatter plots of the differences between lung tumor tissue and normal lung tissue for permittivity and conductivity provided another beneficial information. Differences between lung tumor and normal lung tissues for permittivity tended to decrease quickly and finally verged to zero. Conversely, differences between lung tumor and normal lung tissues for conductivity tended to increase slowly and constantly. This suggested that different frequency should be used for permittivity and conductivity measurements. Lower frequency (50–100 MHz) was suitable for permittivity while higher frequency (4,000 MHz) was suitable for conductivity measurement to obtain the maximum differences between benign and malignant tissues. This implied that choosing permittivity as the basis for classification possessed better stability of parameter than conductivity. For these reasons, we believe that using the logarithmic permittivity function for differential analysis was better than using the exponential conductivity functions.

There were statistically significant correlations between slopes and intercepts of permittivity function and multiple PET parameters (such as SUVmax). This manifestation supported the hypothesis that the metabolic activity of tumors was related to tumor permittivity and more samples need to be included for further analysis. Beyond our expectation, the MTV and TLG of lung tumor, which represent the gross glycometabolism, were both not correlated with dielectric properties. Although no statistical correlation was shown between slopes of permittivity function and CV_SUV in the Pearson correlation analysis, the t-test revealed a significant difference in the CV_SUV values of tumors between the high-slopes of permittivity function group and the low-slopes of permittivity function group. This might suggest that the relationship between slopep and CV_SUV was not linear, and the sample size may also have an impact on the results of statistical analysis. Besides, because of CV_SUV was one of the markers indicating the metabolic heterogeneity, this also indicates that the metabolic heterogeneity of tumors was associated with permittivity.

Our study has several limitations. First, the small sample size may introduce selection bias and increase statistical uncertainty, potentially impacting the robustness of our conclusions. However, this pilot study provides preliminary evidence supporting the feasibility and necessity of further investigation. Second, within this limited sample, only a subset of patients underwent comprehensive immunohistochemistry and genetic testing. This resulted in the lack of further analyses. Additionally, follow-up data were available for only a portion of the patients, limiting our ability to assess the relationship between dielectric properties and clinical outcomes.

Conclusions

Using the fitted exponential and logarithmic functions of dielectric properties, including permittivity and conductivity, could distinguish lung tumor from normal lung tissues. Besides, pathological category may also be classified by conductivity which needs further verification. The metabolic activity and heterogeneity of tumors which were reflected by [18F] FDG PET/CT parameters were associated with the dielectric properties. The method used in this study is simpler and more explainable than the previous studies, making it a promising candidate for broader clinical application.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STARD reporting checklist. Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-934/rc

Peer Review File: Available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-934/prf

Funding: This work was supported by the Guangzhou Municipal Science and Technology Project, China (No. 202201011365 to Y.Z.; No. 2023A04J2397 to D.L.).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-934/coif). D.L. receives article processing charges from Guangzhou Municipal Science and Technology Project, China (No. 2023A04J2397). Y.Z. receives article processing charges from Guangzhou Municipal Science and Technology Project, China (No. 202201011365). The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the ethics committee of the Nanfang Hospital, Southern Medical University (No. NFEC-2017-070) and individual consent for this retrospective analysis was waived.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

  1. Zhang J, Li J, Hou Y, et al. Osr2 functions as a biomechanical checkpoint to aggravate CD8(+) T cell exhaustion in tumor. Cell 2024;187:3409-3426.e24. [Crossref] [PubMed]
  2. Baghdadi MB, Houtekamer RM, Perrin L, et al. PIEZO-dependent mechanosensing is essential for intestinal stem cell fate decision and maintenance. Science 2024;386:eadj7615. [Crossref] [PubMed]
  3. Fan W, Adebowale K, Váncza L, et al. Matrix viscoelasticity promotes liver cancer progression in the pre-cirrhotic liver. Nature 2024;626:635-42. [Crossref] [PubMed]
  4. Carnicer-Lombarte A, Malliaras GG, Barone DG. The Future of Biohybrid Regenerative Bioelectronics. Adv Mater 2025;37:e2408308. [Crossref] [PubMed]
  5. Zhou Y, Wei Y, Li L, et al. Optogenetics in medicine: innovations and therapeutic applications. Curr Opin Biotechnol 2025;92:103262. [Crossref] [PubMed]
  6. Gabriel C, Gabriel S, Corthout E. The dielectric properties of biological tissues: I. Literature survey. Phys Med Biol 1996;41:2231-49. [Crossref] [PubMed]
  7. Egot-Lemaire S, Pijanka J, Sulé-Suso J, et al. Dielectric spectroscopy of normal and malignant human lung cells at ultra-high frequencies. Phys Med Biol 2009;54:2341-57. [Crossref] [PubMed]
  8. Lu D, Peng J, Wang Z, et al. Dielectric property measurements for the rapid differentiation of thoracic lymph nodes using XGBoost in patients with non-small cell lung cancer: a self-control clinical trial. Transl Lung Cancer Res 2022;11:342-56. [Crossref] [PubMed]
  9. Lu D, Yu H, Wang Z, et al. Classification of Metastatic and Non-Metastatic Thoracic Lymph Nodes in Lung Cancer Patients Based on Dielectric Properties Using Adaptive Probabilistic Neural Networks. Front Oncol 2021;11:640804. [Crossref] [PubMed]
  10. Shetty S, U A, Kumar R, et al. Electrical conductivity spectra of hepatic tumors reflect hepatocellular carcinoma progression in mice. Biomed Phys Eng Express 2020; [Crossref]
  11. Schepps JL, Foster KR. The UHF and microwave dielectric properties of normal and tumour tissues: variation in dielectric properties with tissue water content. Phys Med Biol 1980;25:1149-59. [Crossref] [PubMed]
  12. Duché G, Sanderson JM. The Chemical Reactivity of Membrane Lipids. Chem Rev 2024;124:3284-330. [Crossref] [PubMed]
  13. Kandathil A, Subramaniam RM. FDG PET/CT for Primary Staging of Lung Cancer and Mesothelioma. Semin Nucl Med 2022;52:650-61. [Crossref] [PubMed]
  14. Rogasch JMM, Frost N, Bluemel S, et al. FDG-PET/CT for pretherapeutic lymph node staging in non-small cell lung cancer: A tailored approach to the ESTS/ESMO guideline workflow. Lung Cancer 2021;157:66-74. [Crossref] [PubMed]
  15. Evans AL, Buehler D, Schulte JJ, et al. Microwave dielectric properties of normal, fibroelastotic, and malignant human lung tissue. Biomed Phys Eng Express 2024; [Crossref]
  16. Farina L, Ruvio G, Shatwan R, et al. Histology-Validated Dielectric Characterisation of Lung Carcinoma Tissue for Microwave Thermal Ablation Applications. Cancers (Basel) 2023;15:3738. [Crossref] [PubMed]
  17. Vesselle H, Salskov A, Turcotte E, et al. Relationship between non-small cell lung cancer FDG uptake at PET, tumor histology, and Ki-67 proliferation index. J Thorac Oncol 2008;3:971-8. [Crossref] [PubMed]
  18. Sasaki K, Porter E, Rashed EA, et al. Measurement and image-based estimation of dielectric properties of biological tissues -past, present, and future. Phys Med Biol 2022; [Crossref]
  19. Naqvi SAR, Mobashsher AT, Mohammed B, et al. Benign and Malignant Skin Lesions: Dielectric Characterization, Modelling and Analysis in Frequency Band 1 to 14 GHz. IEEE Trans Biomed Eng 2023;70:628-39. [Crossref] [PubMed]
  20. Sarreshtehdari A, García-Sánchez T, Sánchez-Velázquez P, et al. Electrical Conductivity Measurement in Human Liver Tissue: Assessment on Normal vs. Tumor Tissue and under In Vivo vs. Ex Vivo Conditions. Biosensors (Basel) 2024;14:382. [Crossref] [PubMed]
  21. Chen Q, Rao X, Cao D, et al. Real-Time Differentiation Between Benign and Malignant Breast Tumors and Other Tissues Using Dielectric Properties. Med Sci Monit 2025;31:e947531. [Crossref] [PubMed]
  22. Rao X, Chen Q, Ding L, et al. Application of dielectric properties for identification of normal and malignant gastrointestinal tumors and lymph nodes ex vivo. Phys Eng Sci Med 2025;48:75-85. [Crossref] [PubMed]
Cite this article as: Wang L, Li S, Zhou H, Li H, Lin N, Huang Q, Yu H, Wang Z, Chen Z, Zhang Y, Lu D. Innovation for using dielectric properties to distinguish lung tumor from normal lung tissues and preliminary exploration for the relevance with metabolic features. J Thorac Dis 2025;17(10):7778-7789. doi: 10.21037/jtd-2025-934

Article Options

Download Citation

Share