Biomedical Engineering Reference
In-Depth Information
Frequency Domain Breast Lesion Classification
Using Ultra-Wideband Lesion Response
Arash Maskooki, Cheong Boon Soh, Aye Chan, Erry Gunawan
and Kay Soon Low
Abstract
Shape of an object has been shown to affect the backscattered ultra-
wideband (UWB) pulse in frequency domain. Frequency-domain response of the
backscattered signal is the sum of damping harmonics where each harmonic is asso-
ciated with a specific scatterer and the damping factor is a function of the shape of
the corresponding scatterer. Backscattered signals from benign and malignant breast
tumors are obtained through numerical analysis. The damping factors were extracted
from frequency response of the backscattered signal and neural networks were used
to classify breast lesions into benign and malignant tumors. Different architectures
of neural networks were investigated and cascaded Kolmogorov network was found
to be most accurate classifier. Results show that the cascaded Kolmogorov network
classifier can increase the accuracy of diagnosis up to 75.2 %, which is higher than
existing methods.
Keywords
Ultra-wideband
·
Breast cancer detection
·
Lesion classification
·
Neural
networks
·
Numerical lesion model
·
Kolmogorov architecture
·
Finite-difference
time-domain (FDTD)
·
Geometrical theory of diffraction (GTD)
·
Frequency domain
analysis
·
Scattering
·
Diffraction
Introduction
Breast cancer is the second most prevalent type of cancer and the second leading
cause of death by cancer in women in the USA according to the Center for Disease
Control and Prevention (CDC) [
1
]. Lifetime risk of breast cancer among women in
the USA is very close to one in eight and it is estimated that it has claimed 40,676
lives in the USA in 2009.
C. B. Soh (
)
·
A. Maskooki
·
A. Chan
·
E. Gunawan
·
K. S. Low
School of Electrical and Electronic Engineering,
Nanyang Technological University, 50 Nanyang Drive,
Singapore, 637553 Republic of Singapore
e-mail: ecbsoh@ntu.edu.sg
Search WWH ::
Custom Search