Ultra-wide Band Imaging Algorithm For The Early Breast Cancer Detection

In recent years, the incidence of breast cancer ranks the first in female non-skin malignancies,and becomes one of the common high-mortality diseases, early detection is very helpful to fast andeffective treatment and the reduction of patients’mortality.In this thesis, conventional means of breast cancer detection, such as X-ray photography,magnetic resonance imaging(MRI), laser imaging technology are introduced briefly with theiradvantages and limitations. The ultra-wide band(UWB) microwave imaging is considered to be agreen breast tumor detection method, tumor extracting mainly depends on the electromagneticcharacteristics of biological tissue under the UWB microwave. The diseased tissue would besignificantly different from the healthy breast tissue which with lower moisture content and thedifference of their major electrical parameters are more than five times. Compared with commondetection methods, UWB microwave imaging detection technology has higher resolution ratio,lower radiation and be more convenient, etc.A physical model of semicircular breast is established in the dissertation. The Perfect MatchedLayer(PML) absorb boundary condition and Finite Difference Time Domain(FDTD) method areused to analog field intensity distribution in certain region. Two imaging algorithms are mainlystudied through the scattered field intensity information that acquired before. One is ConfocalMicrowave Imaging(CMI) algorithm. The simulation results tell us that the tumors with differentsizes or locations can be detected effectively by this method even their radium are smaller than5mm.It proves the feasibility of CMI. Then, the impacts of imaging results which coming from thevariation of breast tissue’s value and the signal pulse width is discussed. The results show that thisalgorithm maintains effective for imaging when the tissue values vary and requires the signal pulsewidth be within a certain range. The other algorithm is Support Vector Machine which be regardedas a new learning algorithm on the basis of Statistic. The problem of tumor’s parametersreconstruction and location estimation can be solved by Support Vector Regression principle. It isfound that this method can achieve accurate imaging results rapidly. Then, localization results ofdifferent sampling points numbers are compared. The result shows that the increasing number ofsampling points of receiving antennas will improve the accuracy.