Machine learning aided diagnosis of hepatic malignancies through in vivo dielectric measurements with microwaves

Open-Ended Coaxial Probe for Effective Reconstruction of Biopsy-Excised Tissues' Dielectric Properties

Dielectric characterization is extremely promising in medical contexts because it offers insights into the electromagnetic properties of biological tissues for the diagnosis of tumor diseases. This study introduces a promising approach to improve accuracy in the dielectric characterization of millimeter-sized biopsies based on the use of a customized electromagnetic characterization system by adopting a coated open-ended coaxial probe. Our approach aims to accelerate biopsy analysis without sample manipulation. Through comprehensive numerical simulations and experiments, we evaluated the effectiveness of a metal-coating system in comparison to a dielectric coating with the aim for replicating a real scenario: the use of a needle biopsy core with the tissue inside. The numerical analyses highlighted a substantial improvement in the reconstruction of the dielectric properties, particularly in managing the electric field distribution and mitigating fringing field effects. Experimental validation using bovine liver samples revealed highly accurate measurements, particularly in the real part of the permittivity, showing errors lower than 1% compared to the existing literature data. These results represent a significant advancement for the dielectric characterization of biopsy specimens in a rapid, precise, and non-invasive manner. This study underscores the robustness and reliability of our innovative approach, demonstrating the convergence of numerical analyses and empirical validation.

Dielectric Characterization of Breast Biopsied Tissues as Pre-Pathological Aid in Early Cancer Detection: A Blinded Feasibility Study

Dielectric characterization has significant potential in several medical applications, providing valuable insights into the electromagnetic properties of biological tissues for disease diagnosis, treatment planning, and monitoring of therapeutic interventions. This work presents the use of a custom-designed electromagnetic characterization system, based on an open-ended coaxial probe, for discriminating between benign and malignant breast tissues in a clinical setting. The probe's development involved a well-balanced compromise between physical feasibility and its combined use with a reconstruction algorithm known as the virtual transmission line model (VTLM). Immediately following the biopsy procedure, the dielectric properties of the breast tissues were reconstructed, enabling tissue discrimination based on a rule-of-thumb using the obtained dielectric parameters. A comparative analysis was then performed by analyzing the outcomes of the dielectric investigation with respect to conventional histological results. The experimental procedure took place at Complejo Hospitalario Universitario de Toledo-Hospital Virgen de la Salud, Spain, where excised breast tissues were collected and subsequently analyzed using the dielectric characterization system. A comprehensive statistical evaluation of the probe's performance was carried out, obtaining a sensitivity, specificity, and accuracy of 81.6%, 61.5%, and 73.4%, respectively, compared to conventional histological assessment, considered as the gold standard in this investigation.

Machine Learning-Based Classification of Abnormal Liver Tissues Using Relative Permittivity

The search for non-invasive, fast, and low-cost diagnostic tools has gained significant traction among many researchers worldwide. Dielectric properties calculated from microwave signals offer unique insights into biological tissue. Material properties, such as relative permittivity (εr) and conductivity (σ), can vary significantly between healthy and unhealthy tissue types at a given frequency. Understanding this difference in properties is key for identifying the disease state. The frequency-dependent nature of the dielectric measurements results in large datasets, which can be postprocessed using artificial intelligence (AI) methods. In this work, the dielectric properties of liver tissues in three mouse models of liver disease are characterized using dielectric spectroscopy. The measurements are grouped into four categories based on the diets or disease state of the mice, i.e., healthy mice, mice with non-alcoholic steatohepatitis (NASH) induced by choline-deficient high-fat diet, mice with NASH induced by western diet, and mice with liver fibrosis. Multi-class classification machine learning (ML) models are then explored to differentiate the liver tissue groups based on dielectric measurements. The results show that the support vector machine (SVM) model was able to differentiate the tissue groups with an accuracy up to 90%. This technology pipeline, thus, shows great potential for developing the next generation non-invasive diagnostic tools.

Utilization of dielectric properties for assessment of liver ischemia-reperfusion injury in vivo and during machine perfusion

There is a shortage of donor livers and patients consequently die on waiting lists worldwide. Livers are discarded if they are clinically judged to have a high risk of non-function following transplantation. With the aim of extending the pool of available donor livers, we assessed the condition of porcine livers by monitoring the microwave dielectric properties. A total of 21 livers were divided into three groups: control with no injury (CON), biliary injury by hepatic artery occlusion (AHEP), and overall hepatic injury by static cold storage (SCS). All were monitored for four hours in vivo, followed by ex vivo plurithermic machine perfusion (PMP). Permittivity data was modeled with a two-pole Cole–Cole equation, and dielectric properties from one-hour intervals were analyzed during in vivo and normothermic machine perfusion (NMP). A clear increasing trend in the conductivity was observed in vivo in the AHEP livers compared to the control livers. After four hours of NMP, separations in the conductivity were observed between the three groups. Our results indicate that dielectric relaxation spectroscopy (DRS) can be used to detect and differentiate liver injuries, opening for a standardized and reliable point of evaluation for livers prior to transplantation.

Characterization of Open-Ended Coaxial Probe Sensing Depth with Respect to Aperture Size for Dielectric Property Measurement of Heterogeneous Tissues

The open-ended coaxial probe (OECP) method is frequently used for the microwave dielectric property (DP) characterization of high permittivity and conductivity materials due to inherent advantages including minimal sample preparation requirements and broadband measurement capabilities. However, the OECP method is known to suffer from high measurement error. One well-known contributor to the high error rates is tissue heterogeneity, which can potentially be managed through the selection of a probe with a proper sensing depth (SD). The SD of the OECP is dependent on many factors including sample DPs and probe aperture diameter. Although the effects of sample DPs on SD have been investigated to some extent in the literature, the probe aperture diameters, particularly small diameters, have not been fully explored. To this end, the SDs of probes with three different apertures (0.5, 0.9 and 2.2 mm-diameters) were analyzed in this study. Probes' SDs were first investigated with simulations using a double-layered sample configuration (skin tissue and olive oil). Next, experiments were performed using a commercial OECP with a 2.2 mm aperture diameter. The SD was categorized based on 5%, 20% and 80% DP change. Among these threshold values, a 5% DP change was selected as the benchmark for SD categorization. The findings suggest that probes with a smaller aperture size and correspondingly smaller SD should be utilized when measuring the DPs of thin and multilayered samples, such as healthy and diseased skin tissues, to increase the measurement accuracy.

Classification of rat mammary carcinoma with large scale in vivo microwave measurements

Mammary carcinoma, breast cancer, is the most commonly diagnosed cancer type among women. Therefore, potential new technologies for the diagnosis and treatment of the disease are being investigated. One promising technique is microwave applications designed to exploit the inherent dielectric property discrepancy between the malignant and normal tissues. In theory, the anomalies can be characterized by simply measuring the dielectric properties. However, the current measurement technique is error-prone and a single measurement is not accurate enough to detect anomalies with high confidence. This work proposes to classify the rat mammary carcinoma, based on collected large-scale in vivo S[Formula: see text] measurements and corresponding tissue dielectric properties with a circular diffraction antenna. The tissues were classified with high accuracy in a reproducible way by leveraging a learning-based linear classifier. Moreover, the most discriminative S[Formula: see text] measurement was identified, and to our surprise, using the discriminative measurement along with a linear classifier an 86.92% accuracy was achieved. These findings suggest that a narrow band microwave circuitry can support the antenna enabling a low-cost automated microwave diagnostic system.

High-resolution probe design for measuring the dielectric properties of human tissues

Background

In order to use the microwave to measure the dielectric constant of the human body and improve the measurement resolution, a small near-field probe working at 915 MHz is designed in this paper.

Method

Based on the electric small loop antenna model loaded by the spiral resonator (SR), a small near-field probe was designed. The probe model is designed and optimized by the HFSS (high frequency structure simulator) software. The human tissues were tested by the manufactured probe and the relationship between the S11 parameters of the probe and the human tissues was analyzed.

Results and conclusions

A probe with small size was designed and fabricated, with the overall size of 10.0 mm × 12.0 mm × 0.8 mm. The probe has a good performance with a 30.7 dB return loss, a 20 MHz bandwidth at the resonance point, and a distance resolution of 10 mm. Due to the small size and good resolution of the probe, it can be used in the measurement of human tissues.

Classification of Metastatic and Non-Metastatic Thoracic Lymph Nodes in Lung Cancer Patients Based on Dielectric Properties Using Adaptive Probabilistic Neural Networks

Objective

Dielectric properties can be used in normal and malignant tissue identification, which requires an effective classifier because of the high throughput nature of the data. With easy training and fast convergence, probabilistic neural networks (PNNs) are widely applied in pattern classification problems. This study aims to propose a classifier to identify metastatic and non-metastatic thoracic lymph nodes in lung cancer patients based on dielectric properties.

Methods

The dielectric properties (permittivity and conductivity) of lymph nodes were measured using an open-ended coaxial probe. The Synthetic Minority Oversampling Technique method was adopted to modify the dataset. Feature parameters were scored to select the appropriate feature vector using a Statistical Dependency algorithm. The dataset was classified using adaptive PNNs with an optimized smooth factor using the simulated annealing PNN (SA-PNN). The results were compared with traditional Probabilistic, Support Vector Machines, k-Nearest Neighbor and the Classify functions in MATLAB.

Results

The conductivity frequencies of 3959, 3958, 3960, 3978, 3510, 3889, 3888, and 3976 MHz were selected as the feature vectors for 219 lymph nodes (178 non-metastatic and 41 metastatic). Compared with the other methods, SA-PNN achieved the highest classification accuracy (92.92%) and the corresponding specificity and sensitivity were 94.72% and 91.11%, respectively.

Conclusions

Compared with the other methods, the SA-PNN proposed in the present study achieved a higher classification accuracy, which provides a new scheme for classification of metastatic and non-metastatic thoracic lymph nodes in lung cancer patients based on dielectric properties.

Towards Accurate Microwave Characterization of Tissues: Sensing Depth Analysis of Open-Ended Coaxial Probes with Ex Vivo Rat Breast and Skin Tissues

Dielectric properties of biological materials are commonly characterized with open-ended coaxial probes due to the broadband and non-destructive measurement capabilities. Recently, potential diagnostics applications of the technique have been investigated. Although the technique can successfully classify the tissues with different dielectric properties, the classification accuracy can be improved for tissues with similar dielectric properties. Increase in classification accuracy can be achieved by addressing the error sources. One well-known error source contributing to low measurement accuracy is tissue heterogeneity. To mitigate this error source, there is a need define the probe sensing depth. Such knowledge can enable application-specific probe selection or design. The sensing depth can also be used as an input to the classification algorithms which can potentially improve the tissue classification accuracy. Towards this goal, this work investigates the sensing depth of a commercially available 2.2 mm aperture diameter probe with double-layered configurations using ex vivo rat breast and skin tissues. It was concluded that the dielectric property contrast between the heterogeneous tissue components has an effect on the sensing depth. Also, a membrane layer (between 0.4-0.8 mm thickness) on the rat wet skin tissue and breast tissue will potentially affect the dielectric property measurement results by 52% to 84%.

Towards Non-Invasive Diagnosis of Skin Cancer: Sensing Depth Investigation of Open-Ended Coaxial Probes

Dielectric properties of biological tissues are traditionally measured with open-ended coaxial probes. Despite being commercially available for laboratory use, the technique suffers from high measurement error. This prevents the practical applications of the open-ended coaxial probes. One such application is the utilization of the technique for skin cancer detection. To enable a diagnostic tool, there is a need to address the error sources. Among others, tissue heterogeneity is a major contributor to measurement error. The effect of tissue heterogeneity on measurement accuracy can be decreased by quantifying the probe sensing depth. To this end, this work (1) investigates the sensing depth of the 2.2 mm-diameter open-ended coaxial probe for skin mimicking material and (2) offers a simple experimental setup and protocol for sensing depth characterization of open-ended coaxial probes. The sensing depth characterized through simulation and experiments using two double-layered configurations composed to mimic the skin tissue heterogeneity. Three thresholds in percent increase of dielectric property measurements were chosen to determine the sensing depth. Based on the experiment results, it was concluded that the sensing depth was effected by the dielectric property contrast between the layers. That is, high contrast results in rapid change whereas low contrast results in a slower change in measured dielectric properties. It was also concluded that the sensing depth was independent of frequency between 0.5 to 6 GHz and was mostly determined by the material located immediately at the aperture of the probe.

[Discrimination of lung cancer and adjacent normal tissues based on permittivity by optimized probabilistic neural network]

OBJECTIVE

To propose a probabilistic neural network classification method optimized by simulated annealing algorithm (SA-PNN) to discriminate lung cancer and adjacent normal tissues based on permittivity.

METHODS

The permittivity of lung tumors and the adjacent normal tissues was measured by an open-ended coaxial probe, and the statistical dependency (SD) algorithm was used for frequency screening.The permittivity associated with the selected frequency points was taken as the characteristic variable, and SA-PNN was used to discriminate lung cancer and the adjacent normal tissues.

RESULTS

Three frequency points, namely 984 MHz, 2724 MHz and 2723 MHz, were selected by SD algorithm.SA-PNN was used to discriminate 200 samples with the permittivity at the 3 frequency points as the characteristic variable.After 10-fold cross-validation, the final discrimination accuracy was 92.50%, the sensitivity was 90.65%, and the specificity was 94.62%.

CONCLUSIONS

Compared with the traditional probabilistic neural network, BP neural network, RBF neural network and the classification discriminant analysis function (Classify) in MATLAB, the proposed SA-PNN has higher accuracy, sensitivity and specificity for discriminating lung cancer and the adjacent normal tissues based on permittivity.

In Vivo Dielectric Properties of Healthy and Benign Rat Mammary Tissues from 500 MHz to 18 GHz

This work investigates the in vivo dielectric properties of healthy and benign rat mammary tissues in an attempt to expand the dielectric property knowledge of animal models. The outcomes of this study can enable testing of microwave medical technologies on animal models and interpretation of tissue alteration-dependent in vivo dielectric properties of mammary tissues. Towards this end, in vivo dielectric properties of healthy rat mammary tissues and chemically induced benign rat mammary tumors including low-grade adenosis, sclerosing adenosis, and adenosis were collected with open-ended coaxial probes from 500 MHz to 18 GHz. The in vivo measurements revealed that the dielectric properties of benign rat mammary tumors are higher than the healthy rat mammary tissues by 9.3% to 35.5% and 19.6% to 48.7% for relative permittivity and conductivity, respectively. Furthermore, to our surprise, we found that the grade of the benign tissue affects the dielectric properties for this study. Finally, a comparison with ex vivo healthy human mammary tissue dielectric properties revealed that the healthy rat mammary tissues best replicate the dielectric properties of healthy medium density human samples.

Multiclass Classification of Hepatic Anomalies with Dielectric Properties: From Phantom Materials to Rat Hepatic Tissues

Open-ended coaxial probes can be used as tissue characterization devices. However, the technique suffers from a high error rate. To improve this technology, there is a need to decrease the measurement error which is reported to be more than 30% for an in vivo measurement setting. This work investigates the machine learning (ML) algorithms’ ability to decrease the measurement error of open-ended coaxial probe techniques to enable tissue characterization devices. To explore the potential of this technique as a tissue characterization device, performances of multiclass ML algorithms on collected in vivo rat hepatic tissue and phantom dielectric property data were evaluated. Phantoms were used for investigating the potential of proliferating the data set due to difficulty of in vivo data collection from tissues. The dielectric property measurements were collected from 16 rats with hepatic anomalies, 8 rats with healthy hepatic tissues, and in house phantoms. Three ML algorithms, k-nearest neighbors (kNN), logistic regression (LR), and random forests (RF) were used to classify the collected data. The best performance for the classification of hepatic tissues was obtained with 76% accuracy using the LR algorithm. The LR algorithm performed classification with over 98% accuracy within the phantom data and the model generalized to in vivo dielectric property data with 48% accuracy. These findings indicate first, linear models, such as logistic regression, perform better on dielectric property data sets. Second, ML models fitted to the data collected from phantom materials can partly generalize to in vivo dielectric property data due to the discrepancy between dielectric property variability.

Microwave dielectric property based classification of renal calculi: Application of a kNN algorithm

The proper management of renal lithiasis presents a challenge, with the recurrence rate of the disease being as high as 46%. To prevent recurrence, the first step is the accurate categorization of the discarded renal calculi. Currently, the discarded renal calculi type is determined with the X-ray powder diffraction method which requires a cumbersome sample preparation. This work presents a new approach that can enable fast and accurate classification of discarded renal calculi with minimal sample preparation requirements. To do so, first, the measurements of the dielectric properties of naturally formed renal calculi are collected with the open-ended contact probe technique between 500 MHz and 6 GHz with 100 MHz intervals. Cole-Cole parameters are fitted to the measured dielectric properties with the generalized Newton-Raphson method. The renal calculi types are classified based on their Cole-Cole parameters as calcium oxalate, cystine, or struvite. The classification is performed using k-nearest neighbors (kNN) machine learning algorithm with the 10 nearest neighbors, where accuracy as high as 98.17% is achieved.

Radio-Frequency and Microwave Techniques for Non-Invasive Measurement of Blood Glucose Levels

This paper reviews non-invasive blood glucose measurements via dielectric spectroscopy at microwave frequencies presented in the literature. The intent is to clarify the key challenges that must be overcome if this approach is to work, to suggest some possible ways towards addressing these challenges and to contribute towards prevention of unnecessary ‘reinvention of the wheel’.

[Support vector machine?assisted diagnosis of human malignant gastric tissues based on dielectric properties]

OBJECTIVE

To achieve differential diagnosis of normal and malignant gastric tissues based on discrepancies in their dielectric properties using support vector machine.

METHODS

The dielectric properties of normal and malignant gastric tissues at the frequency ranging from 42.58 to 500 MHz were measured by coaxial probe method, and the Cole?Cole model was used to fit the measured data. Receiver?operating characteristic (ROC) curve analysis was used to evaluate the discrimination capability with respect to permittivity, conductivity, and Cole?Cole fitting parameters. Support vector machine was used for discriminating normal and malignant gastric tissues, and the discrimination accuracy was calculated using k?fold cross?

RESULTS

The area under the ROC curve was above 0.8 for permittivity at the 5 frequencies at the lower end of the measured frequency range. The combination of the support vector machine with the permittivity at all these 5 frequencies combined achieved the highest discrimination accuracy of 84.38% with a MATLAB runtime of 3.40 s.

CONCLUSION

The support vector machine?assisted diagnosis is feasible for human malignant gastric tissues based on the dielectric properties.