Measurement Of Dielectric Properties And Automatic Identification Of Human Tumor Tissues Based On Open-ended Coaxial Probe

The dielectric properties of biological tissues are inherent physical properties,which are characteristics of the absorption and coupling of electromagnetic energy by biological tissues under the action of electromagnetic fields.For human tissue,its dielectric properties can provide basic data for related research such as the identification of benign and malignant tissues during surgery,magnetic resonance electrical tomography(MR-EPT),and so on.Therefore,it is of great significance to measure the dielectric properties of human tissues and establish a database of human tissue dielectric properties.However,current reports on the dielectric properties of human tissues are not comprehensive.Studies show that there is a significant difference in the dielectric properties between normal human tissues and tumor tissues,which means that the tumor tissues can be distinguished through the measurement of dielectric properties,which may become a new technical methods for early cancer detection.For example,in the detection of surgical resection margins during surgical resection of cancer,the resection margins of tumor tissue need to be removed to prevent recurrence of the operation.However,when using the open-end coaxial probe method to measure the dielectric characteristics of human tissues,multiple measurements are required to obtain Mean value to determine the tissue type has the disadvantages of low repeatability and high measurement error.Therefore,we apply machine learning algorithms to the measurement of the dielectric properties of lung tissues,so as to realize the automatic identification of normal tissues and tumor tissues,which is of great significance for the implementation of tumor tissue edge detection technology during surgery.In this paper,the dielectric properties of metastatic and non-metastatic lymph nodes(LNs)in lung cancer were measured using the open-ended coaxial probe method.The results show that there is a significant difference in the dielectric characteristics between metastatic and non-metastatic LNs at wideband frequencies,and the difference is statistically significant at six single frequencies.In addition,the measured data conformed to the double-pole Cole-Cole equation,and a good fitting effect was obtained.Therefore,this study can not only provide basic data for other related studies,but also provide a new possibility for identifying intraoperative LN metastasis.Based on the differences in the dielectric characteristics of normal human lung tissue and tumor tissue,this paper uses the LSTM network as the basic unit to build a Sequence Analyze Model(SAM)to automatically identify human tumor tissue.The obtained classification results are compared with Support Vector Machine(SVM)and Multilayer Perceptron(MLP).SAM finally obtained a classification accuracy rate of 88.05%,which is significantly higher than 83.19%of SVM and 76.55%of MLP.This shows that compared with the SVM method and the MLP method,the SAM proposed in this paper has the best classification effect on organizational categories.Based on the above research,some improvements in measurement methods are proposed.1)Using Visual Studio 2012 software to call the dynamic link library instructions,to achieve direct control of the network analyzer,combined with the calculation process of the dielectric characteristics,developed an organization dielectric characteristics calculation software.2)Using deionized water and 0.9%NaCl solution as the calibration solution,the relationship between measurement error and solution correlation was explored.The results show that the stronger the correlation between the dielectric properties of the calibration solution and the measured solution,the more accurate the measurement result.These have provided experience and reference for future measurement and research.