Research On Artifact Removal Method Applied To Microwave Near-field Imaging

Breast cancer has endangered women’s health since years.Therefore,patients detected by the early detection of breast cancer can receive timely treatment,which is important to improve the prognosis of patients.Due to the advantages of less radiation and low cost,microwave near-field imaging become one of the major research directions,attracting attention from the researchers.Microwave confocal imaging relies on the scattering effect of ultra-wide band pulse signals propagating in different tissues caused by the difference of electrical characteristics between tumor and normal tissues.The information describing the location,shape of tumor can be extracted from the analysis of the scattered signals.However,since most of the energy in the scattered signals is occupied by the clutter signals of tissues such as skin and fat,the tumor response is obscure.Microwave confocal imaging of the scattered signals will produce artifacts and cannot present the tumor.Aiming at this problem,this thesis uses several signal processing methods to remove artifacts from the signals,extracts the tumor response,and then uses microwave confocal imaging to verify the effect of artifact removal method.In some previous studies of artifact removal methods,the model merely considered skin,fat,and tumor,rather than the complex tissue such as gland and others.Therefore,two different two-dimensional simulation model constructed by MRI to fit the reality situation are illustrated.In view of the dispersive effect of the human tissues,this thesis demonstrates the electrical characteristics of normal and tumor tissues at different frequency and the ColeCole model,with the study of the precondition provided by ultra-wideband pulse signals for microwave confocal imaging.In the process of artifact removal,the entropy-based time window algorithm is used to determine the time window which helps the Wiener filter to construct the filtered signal.Due to the least square’s regression solution method is vulnerable to multicollinearity influence,the partial least square’s regression method is more effective to get the filtered signal.Through the spectrum analysis of the signal,it is found that the frequency of tumor response is relatively low,and the artifact signal mainly appears at the early stage of the signal.Therefore,the advantages of wavelet transform are stimulated for performing noise reduction of the filtered signal and obtaining expected tumor response.Because several artifacts in the image of the monostatic is nonnegligible,scattered signals are required to be grouped in the multistatic.The similarity degree in each group is calculated to evaluating the qualification whether the grouped signal deserves to process by artifact removal methods.In the monostatic,the expected tumor response is processed by using microwave confocal imaging to produce image,compared with HAR and the ideal.It was found that in 30 acquisition positions,the results of the proposed method make little difference with HAR.Whereas,in the case of the results of HAR unable to depict the tumor in 12 acquisition positions,the results of the proposed illustrate the tumor precisely.In the multistatic,the proposed method will be compared with the ideal.Benefit by the signal grouping,the proposed method removes artifacts and present an image of the tumor accurately,with a small gap compared with the ideal removal method.