Microwave dielectric property based classification of malignancies
Microwave dielectric property based classification of malignancies
Dosyalar
Tarih
2019
Yazarlar
Yılmaz, Tuba
Süreli Yayın başlığı
Süreli Yayın ISSN
Cilt Başlığı
Yayınevi
İstanbul Teknik Üniversitesi
Özet
Broadband dielectric property measurements of bi-
ological tissues is mostly performed with open ended coaxial
probe technique due to a number of advantages such as flexible
sample shape and size. However, the technique is known to
suffer from high error rates; thus, envisioned applications of
the technique remains hampered by this problem. One way to
mitigate such error for medical applications is to perform tissue
classification with machine learning algorithms. In this work,
Cole-Cole parameters of rat liver dielectric properties are used
for training and testing of an in house Support Vector Machine
(SVM) algorithm to enable malignant tissue classification. Cole-
Cole parameters are fitted with Particle Swarm Optimization
(PSO) to a total of 700 dielectric property measurements collected
from 22 rats. The Cole-Cole parameters are fed to the SVM
algorithm and k-fold cross validation is used to prevent the
algorithm from memorizing the data. Hepatic malignancies are
classified with 96% accuracy where a better accuracy is obtained
in comparison to plain dielectric property measurement and also
an automated decision making mechanism is enabled.
Açıklama
Anahtar kelimeler
Component,
Formatting,
Style,
Styling,
Insert
Alıntı
Tuba Yilmaz, 'Microwave dielectric property based classification of anomalies', ICEAA 2019 International Conference on Electromagnetics in Advanced Applications, September 9-13, Granada, Spain.