Microwave imaging for early stage breast tumor detection and discrimination via complex natural resonances
- Publication Type:
- Thesis
- Issue Date:
- 2014
Open Access
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is open access.
In this thesis, a new microwave imaging technique for early stage breast cancer detection is developed to achieve two key aims: (i) to reconstruct the radar image of the suspicious region within the breast and (ii) to decide whether a suspicious region has malignant or benign tumors by differentiating their morphological features in terms of their complex natural resonances.
For our investigations we employ both numerical and chemical tissue mimicking breast phantoms. The breast phantom is illuminated by UWB pulses radiated from antenna elements arranged in a multistate configuration surrounding the breast. An efficient pre-processing technique is proposed to process the received pulses for the removal of early-time artifacts. To reduce the interferences of the background tissue clutter in inhomogeneous breast environment, a new time-of-arrival (TOA) auto-calibration is presented to estimate accurate TOA for confocal imaging. For determining the suspicious region within the breast, a novel and efficient data independent beam former known as Modified Weighted Delay and Sum (MWDAS) algorithm has been proposed. Once the suspicious region is localized by MWDAS method, the waveform of late-time backscattered field will be estimated using a proposed two-stage waveform estimation method. The accuracy of the waveform improves the extraction of complex natural resonances (CNR) that will be used to discriminate of whether a suspicious tissue is malignant or benign. Basing on radar target discrimination, we propose that the CNRs extracted from the late-time resonant tumor response can be closely related their morphological properties: spiculated lesion has CNR poles that differ from CNR poles of a smooth lesion. To validate our proposal, we perform FDTD simulations on 2D and 3D numerical breast phantoms that have been developed based on MRI-derived tissue dielectric properties. These simulations have revealed that the CNRs from malignant tumors have significant lower damping factors than the benign ones. These simulation results helped to reconfirm that it is possible to distinguish malignant and benign breast tumors based on their CNRs.
To validate the proposed method of tissue discrimination, we have developed an experimental UWB imaging prototype using novel UWB sensors and tissue mimicking chemical breast phantoms to carry out preliminary preclinical experiments. Three novel end-fire compact sized UWB antennas have been proposed. After thoroughly investigating their characteristics, a novel UWB horn antenna known as BAHA that offered superior UWB performance is chosen, fabricated and measured to confirm its characteristics. A prototype experimental imaging system that incorporates 32 BAHA antenna elements forming a hemispherical UWB array is fabricated and tested using a vector network analyzer. Tissues mimicking chemical phantoms with dielectric properties similar to human breasts have been manufactured to have both adipose-tissue dominated homogeneous phantom with a dielectric contrast of 4:1 and a low-adipose in-homogeneously dense phantom with dielectric contrast of 1.7:1. Experimental results obtained using the hemispherical array prototype and phantoms have shown that dielectric inserts (12mm diameter) that mimic malignant and benign lesions can be successfully detected from both high and low dielectric contrast scenarios. Tumor mimicking lossy dielectric inserts with both irregular and smooth patterns have also been fabricated using chemicals to represent malignant and benign tumors respectively.
Finally, measured data from experimental prototype have demonstrated that tissue shape can be discriminated via CNRs. The experimental results confirmed that the proposed UWB antenna array is capable of picking up undistorted late-time signals from embedded tumor-mimicking dielectric inserts with different morphological profiles to offer reliable CNR extraction. Matrix Pencil Method is employed to extract CNRs from late time responses. Our investigations have confirmed that damping factors of the extracted CNRs from both spiculated and smooth inserts can be used to differentiate their shapes which are quite promising for early stage breast cancer detection.
Please use this identifier to cite or link to this item: