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

Tuba Yilmaz, Mahmut Alp Kiliç, Melike Erdoǧan, Mehmet Çayören, Doruk Tunaoǧlu, Ismail Kurtoǧlu, Yusuf Yaslan, Hüseyin Çayören, Akif Enes Arkan, Serkan Teksöz, Gülden Cancan, Nuray Kepil, Sibel Erdamar, Murat Özcan, Ibrahim Akduman, Tunaya Kalkan

Research output: Contribution to journalReview articlepeer-review

39 Citations (Scopus)

Abstract

In the past decade, extensive research on dielectric properties of biological tissues led to characterization of dielectric property discrepancy between the malignant and healthy tissues. Such discrepancy enabled the development of microwave therapeutic and diagnostic technologies. Traditionally, dielectric property measurements of biological tissues is performed with the well-known contact probe (open-ended coaxial probe) technique. However, the technique suffers from limited accuracy and low loss resolution for permittivity and conductivity measurements, respectively. Therefore, despite the inherent dielectric property discrepancy, a rigorous measurement routine with open-ended coaxial probes is required for accurate differentiation of malignant and healthy tissues. In this paper, we propose to eliminate the need for multiple measurements with open-ended coaxial probe for malignant and healthy tissue differentiation by applying support vector machine (SVM) classification algorithm to the dielectric measurement data. To do so, first, in vivo malignant and healthy rat liver tissue dielectric property measurements are collected with open-ended coaxial probe technique between 500 MHz to 6 GHz. Cole-Cole functions are fitted to the measured dielectric properties and measurement data is verified with the literature. Malign tissue classification is realized by applying SVM to the open-ended coaxial probe measurements where as high as 99.2% accuracy (F1 Score) is obtained.

Original languageEnglish
Pages (from-to)5089-5102
Number of pages14
JournalPhysics in Medicine and Biology
Volume61
Issue number13
DOIs
Publication statusPublished - 20 Jun 2016

Bibliographical note

Publisher Copyright:
© 2016 Institute of Physics and Engineering in Medicine.

Keywords

  • Cole-Cole parameters
  • contact probe technique
  • dielectric properties
  • hepatic malignancies
  • support vector machine

Fingerprint

Dive into the research topics of 'Machine learning aided diagnosis of hepatic malignancies through in vivo dielectric measurements with microwaves'. Together they form a unique fingerprint.

Cite this