Simulation and Design of an UWB Imaging System for Breast Cancer Detection

Breast cancer is the most frequently diagnosed cancer among women. In recent years, the mortality rate due to this disease is greatly decreased thanks to both enormous progress in cancer research, and screening campaigns which have allowed the increase in the number of early diagnoses of the disease. In fact, if the tumor is identied in its early stage, e.g. when it has a diameter of less than one centimeter, the possibility of a cure can reach 93%. However, statistics show that more young aged women are suered breast cancer. The goal of screening exams for early breast cancer detection is to nd cancers before they start to cause symptoms. Regular mass screening of all women at risk is a good option to achieve that. Instead of meeting very high diagnostic standards, it is expected to yield an early warning, not a denitive diagnosis. In the last decades, X-ray mammography is the most ecient screening technique. However, it uses ionizing radiation and, therefore, should not be used for frequent check-ups. Besides, it requires signicant breast compression, which is often painful. In this scenario many alternative technologies were developed to overcome the limitations of mammography. Among these possibilities, Magnetic Resonance Imaging (MRI) is too expensive and time-consuming, Ultrasound is considered to be too operatordependent and low specicity, which are not suitable for mass screening. Microwave imaging techniques, especially Ultra WideBand (UWB) radar imaging, is the most interesting one. The reason of this interest relies on the fact that microwaves are non-ionizing thus permitting frequent examinations. Moreover, it is potentially lowcost and more ecient for young women. Since it has been demonstrated in the literatures that the dielectric constants between cancerous and healthy tissues are quite dierent, the technique consists in illuminating these biological tissues with microwave radiations by one or more antennas and analyzing the re ected signals. An UWB imaging system consists of transmitters, receivers and antennas for the RF part, the transmission channel and of a digital backend imaging unit for processing the received signals. When an UWB pulse strikes the breast, the pulse is re ected due to the dielectric discontinuity in tissues, the bigger the dierence, the bigger the backscatter. The re ected signals are acquired and processed to create the energy maps. This thesis aims to develop an UWB system at high resolution for the detection of carcinoma breast already in its initial phase. To favor the adoption of this method in screening campaigns, it is necessary to replace the expensive and bulky RF instrumentation used so far with ad-hoc designed circuits and systems. In order to realize that, at the very beginning, the overall system environment must be built and veried, which mainly consists of the transmission channel{the breast model and the imaging unit. The used transmission channel data come from MRI of the prone patient. In order to correctly use this numerical model, a simulator was built, which was implemented in Matlab, according to the Finite-Dierence-Time- Domain (FDTD) method. FDTD algorithm solves the electric and magnetic eld both in time and in space, thus, simulates the propagation of electromagnetic waves in the breast model. To better understand the eect of the system non-idealities, two 2D breast models are investigated, one is homogeneous, the other is heterogeneous. Moreover, the modeling takes into account all critical aspects, including stability and medium dispersion. Given the types of tissues under examination, the frequency dependence of tissue dielectric properties is incorporated into wideband FDTD simulations using Debye dispersion parameters. A performed further study is in the implementation of the boundary conditions. The Convolution Perfectly Matched Layer (CPML) is used to implement the absorbing boundaries. The objective of the imaging unit is to obtain an energy map representing the amount of energy re ected from each point of the breast, by recombining the sampled backscattered signals. For this purpose, the study has been carried out on various beamforming in the literature. The basic idea is called as "delay and sum", which is to align the received signals in such a way as to focus a given point in space and then add up all the contributions, so as to obtain a constructive interference at that point if this is a diseased tissue. In this work, Microwave Imaging via Space Time (MIST) Beamforming algorithm is applied, which is based on the above principle and add more elaborations of the signals in order to make the algorithm less sensitive to propagation phenomena in the medium and to the non-idealities of the system. It is divided into two distinct steps: the rst step, called SKin Artifact Removal (SKAR), takes care of removing the contributions from the signal caused by the direct path between the transmitter and receiver, the re ection of skin, as they are orders of magnitude higher compared to the re ections caused by cancers; the second step, which is BEAmForming (BEAF), performs the algorithm of reconstruction by forming a weighted combination of time delayed version of the calibrated re ected signals. As discussed above, more attention must be paid on the implementation of the ad-hoc integration circuits. In this scenario, due to the strict requirements on the RF receiver component, two dierent approaches of the implementation of the RF front-end, Direct Conversion (DC) receiver and Coherent Equivalent Time Sampling (CETS) receiver are compared. They are modeled behaviorally and the eects of various impairments, such as thermal, jitter, and phase noise, as well as phase inaccuracies, non-linearity, ADC quantization noise and distortion, on energy maps and on quantitative metrics such as SCR and SMR are evaluated. Dierential Gaussian pulse is chosen as the exciting source. Results show that DC receiver performs higher sensitivity to phase inaccuracies, which makes it less robust than the CETS receiver. Another advantage of the CETS receiver is that it can work in time domain with UWB pulses, other than in frequency domain with stepped frequency continuous waves like the DC one, which reduces the acquisition time without impacting the performance. Based on the results of the behavioral simulations, low noise amplier (LNA) and Track and Hold Amplier (THA) can be regarded as the most critical parts for the proposed CETS receiver, as well as the UWB antenna. This work therefore focuses on their hardware implementations. The LNA, which shows critical performance limitation at bandwidth and noise gure of receiver, has been developed based on common-gate conguration. And the THA based on Switched Source Follower (SSF) scheme has been presented and improved to obtain high input bandwidth, high sampling rate, high linearity and low power consumption. LNA and THA are implemented in CMOS 130nm technology and the circuit performance evaluation has been taken place separately and together. The small size UWB wide-slot antenna is designed and simulated in HFSS. Finally, in order to evaluate the eect of the implemented transistor level components on system performance, a multi-resolution top-down system methodology is applied. Therfore, the entire ow is analyzed for dierent levels of the RF frontend. Initially the system components are described behaviorally as ideal elements. The main activity consists in the analysis and development of the entire frontend system, observing and complementing each other blocks in a single ow simulation, clear and well-dened in its various interfaces. To achieve that the receiver is modeled and analyzed using VHDL-AMS language block by block, moreover, the impact of quantization, noise, jitter, and non-linearity is also evaluated. At last, the behavioral description of antenna, LNA and THA is replaced with a circuit-level one without changing the rest of the system, which permits a system-level assessment of low-level issues.