Enhancing Tumor Detection in IR-UWB Breast Cancer System

An ultra-wideband (UWB) microwave system for breast cancer detection is presented. The proposed system includes monocycle pulse generator, antipodal Vivaldi antenna, breast model, and calibration algorithm for tumor detection. Firstly, our pulse generator employs transmission gate in glitch generator to achieve several advantages such as low power consumption and low ringing level. Secondly, the antipodal Vivaldi antenna is designed assuming FR4 dielectric substrate material, and developed antenna element (80 × 80 mm2) features a −10 dB return loss and bandwidth ranges from 2.3 GHz to more than 11 GHz. Thirdly, the phantom breast can be modeled as a layer of skin, fat, and then tumor is inserted in this layer. Finally, subtract and add algorithm (SAD) is used as a calibration algorithm in tumor detection system. The proposed system suggested that horizontal antenna position with 90° between transmitting and receiving antennas is localized as a suitable antenna position with different rotating location and a 0.5 cm near to phantom. The mean advantages of this localization and tracking position around breast is a high received power signal approximately around mv as a higher recognized signal in tumor detection. Using our proposed system we can detect tumor in 5 mm diameter.


Introduction
Recently, microwave imaging in task of detection and location of malignant tissue in the woman breast has received a considerable amount of interest [1][2][3][4][5][6]. A microwave imaging system is considered as a viable alternative to X-ray mammography due to its several advantages such as cost and insignificant side effects. Microwave imaging includes the propagation of very low levels of microwave energy through the breast tissue. Normal and malignant breast tissue have a difference in the electrical properties as the basis for tumor detection and location.
Normal breast tissue is relatively translucent to microwave radiation, while the malignant lump contains more water and blood, leading to microwave signal backscattering. This scattered signal can be picked up by a microwave antenna and can be analyzed using an image processing to handle on computer [5]. The basic idea of breast cancer detection is to send ultranarrow low power pulses to the patient's body and make a decision depending on the received signal by comparing it with an average of backscattered signals, or a signal backscattered from normal tissues.
The pulse generator is a main block in an imaging system. It has been designed using ultra-wideband impulse radio (IR-UWB) technology [7]. IR-UWB is a very short energy pulse, typically on the order of a hundred picoseconds transmitted with the use of an antipodal Vivaldi designed wideband antenna.
Digital IR-UWB pulse generation [8,9] has been widely studied as a new approach to achieve low power characteristic. IR-UWB deployment in the medical applications is highly desirable, as there are different advantages, such as higher data rates, lower power dissipation, and enhanced medical security to the patient.
The second element on radar techniques is based on the use of ultra-wideband (UWB) antennas, in which a short pulse is transmitted into the body; then the reflected signals are detected by one or more antennas at a receiver placed in different locations [10]. The antipodal Vivaldi antennas (APVA) presented in the literatures [10,11] satisfies the 2 International Scholarly Research Notices requirements for imaging systems such as bandwidth, gain, and impulse response albeit at the expense of the size of a large volumetric. Designing UWB transmitter including pulse generator and antenna is proposed in this paper and then used in tumor detection. First the pulse generator (PG), with the stringent constraints of small area, low ringing, and low power consumption, is designed and implemented. Second the APVA which meets the above-mentioned requirement is developed. Finally, transmitting pulses generated by a proposed pulse generator are made to illuminate a cancerous breast model, with backscatter signals that is received by the receiver antenna placed around the cancerous breast. The cancerous tissue significantly affects the received signal due to a high dielectric difference between malignant and healthy tissue.
The paper is organized as follows; the new design of IR-UWB pulse generator is presented with an output monocycle pulse at 4.2 GHz center frequency compatible with breast cancer imaging system. The design, implementation, and results, using cadence tool, are also presented in Section 2. In Section 3, an Antipodal Vivaldi antenna (APVA) is designed and fabricated on a FR4 with = 4.5, and ℎ = 1.5 mm. It covers the required UWB band from 2.3 to more than 11 GHz and features directive properties. The system is completed with spherical breast phantom and a receiving APVA. The impact of tumor presence on the received signal and operational conditions are optimized in Section 4. In Section 5, an artifact removal algorithm is described to eliminate the large reflections from the skin and the incident signal. Finally Section 6 concludes the paper.

Monocycle Pulse Generator
Monocycle pulses [12] are preferred to simple pulses because they have no dc components, which could represent a limit for the spectral mask compliance and radiation antenna efficiency.
Analog filtering is the simplest one of monocycle UWB pulse generator methods. A common way of directly generating an UWB pulse without using a carrier is to first form a baseband impulse with very short time duration glitches and a high frequency bandwidth and then filter the glitches using a band-pass or a pulse shaping filter. The conventional glitch generator is based on pulse delay element such as inverters to delay pulses by specific delay value and then combining the delayed and the original pulses by an AND gate. In our design the transmission gate is used instead of NAND gate to produce a monocycle pulse of UWB with comparable low power, good balance between the positive and negative parts, and relatively small ringing level. This pulse generator is named transmission gate pulse generator (TGPG).

Pulse Generator Design.
A transmission gate pulse generator (TGPG) based on glitch generation [13] is proposed in this work to produce a monocycle pulse. Two identical glitch generators named GGA and GGB are the back bone of pulse generator circuit as shown in Figure 1. The proposed TGPG consists of transmission gate controlled by two delayed clock pulses De clk, De clk inv, and its positive glitch output connected to drain of 1 and 2 .
There are two cases observed in our design; the first case is a rising clock edge in which De clk equal "0" and De clk inv equal "1." The corresponding data is transferred to drain of 1 and 2 (node pulse o/p) as the transmission gate is ON. After 6 inv delay, delay between main clock and (De Clk), the NMOS 1 and 2 turn on and discharge the node Pulse O/P to ground so the Gaussian pulse is designed. The second case is falling clock edge (De clk equal "1" and De clk inv equal "0") in which node Pulse O/P is connected to ground via NMOS 1 and 2 at the primary time 6 inv . Afterwards, the transmission gate turns on and keeps the node Pulse O/P connected to "0." Then the output of the glitch generator is used to derive a pulse shaper consisting of two CMOS transistors and a shaping filter in form of a resonance circuit. It is an off chip filter where the inductor is a bond wire while the capacitor is added to complete the resonance circuit and the resistor is that of the load. As clear from the schematic in Figure 1, the GGA output is inverted and applied to the pull up PMOSFET 3 of the output pulse shaper; as a result, the pulse generator output at the drain of the output transistors becomes DD . The output glitches of GGB are delayed using eight-stage inverter chain and applied to the pull down transistor NMOSFET 4 that drives the output voltage to 0 V and generates the negative pulse.
The sizes of output pulse shaper stage transistor 3 and 4 have to be selected carefully for correct pulse shape. During the design procedure, the supply voltage is kept constant at its nominal value of 1.2 V. The gate sizes for the building blocks are determined to have a UWB pulse 235 ps. The transmitter output after pulse shaper transistors is applied to the package and off chip printed circuit board (PCB) through a wire bond that can be considered as an equivalent to inductance wb 2.5-nH as shown in Figure 1 [12]. A 4.2 GHz-band BPF is realized using a 0.42 pF series capacitance ps with wb .
The proposed monocycle pulses with 235 ps pulse duration and approximately operating at 4.2 GHz as center frequency are shown in Figure 2. A Gaussian monocycle of 235-ps width is obtained after PEX practices extraction test. The peak-to-peak voltage is 600 mV, and it has good balance between the positive and negative parts, and it has relatively small ringing level.

Implementation Results.
The proposed transmitter is designed and simulated using cadence virtuoso and UMC MMRF 130 nm CMOS process. The layout is shown in  The variation of the amplitude and the duration of generated pulses after Practices Extraction PEX test are shown in Figure 4 for the different operating temperatures, supply voltage, and different fabrication tolerances. The peakto-peak voltage and pulse duration of each case are recorded at different temperatures in Table 1.
Under T-T test with = 27 ∘ , and dc = 1.2 V, the duration and amplitude follow the characteristic of monocycle pulse. Slow-Slow test expanded pulse duration to 250 ps meaning a center frequency of 4 GHz at = 70 ∘ , and dc = 1.1 V, but for Fast-Fast test the duration changed to 200 ps with a corresponding center frequency of 5 GHz at = −20 ∘ , and dc = 1.3 V. These measured values are given in Table 1. From the previous results we conclude that this design has a low ringing level and compatible to breast cancer imaging system.
Comparison between the performance parameters of our PG and the previous UWB pulse generators [12,[14][15][16] presented in Table 2. Our static power consumption is calculated by cadence tools. It is seen that our UWB pulse generator has lower power consumption than all previous   works. Note also that this design exhibits one of the highest pulse peak-to-peak amplitude on standard load impedance, low static power consumption, and low level in ringing effect.

Antipodal Vivaldi Antenna
Antipodal Vivaldi antenna is a type of tapered slot antenna which is an end fire antenna. The antenna consists of the feeding line and the transition and the radiating structures. In general, the radiated structure is exponentially or elliptically tapered, which means that they are composed of two layers structure [10,17].
Antipodal Vivaldi antennas presented in the literature [18,19] satisfies the requirements for breast cancer imaging systems in terms of bandwidth, gain, and impulse response albeit at the expense of significant volumetric size.
The proposed antipodal Vivaldi antenna for inclusions in an UWB microwave imaging system [10] is shown in Figure 5. The design objective is to obtain its bandwidth requirement of 2.3 up to 11 GHz. The following design procedure is utilized in designing the proposed antipodal Vivaldi antenna: The width and length of the antenna structure, excluding the feeder, could be calculated from the following equation with the lowest frequency of operation , thickness of the substrate ℎ, and dielectric constant [10].
where is the speed of light in free space. Then, the radiating structure of the antenna is designed from the intersection of quarters of two ellipses as depicted in Figure 5. The major radii 1 and 2 and the secondary radii 1 and 2 of the two ellipses are chosen according to the following equations: International Scholarly Research Notices   where cut is used to control the lowest frequency of operation.
The microstrip transmission line feeder has a width of and dielectric thickness ℎ to give the characteristic impedance, 0 = 50 ohm that can be calculated using the equation The proposed design is verified using the commercial software package, CST microwave studio, and experimental tests. Figure 6 is a photograph showing the antenna and the test setup. Figure 7 shows the simulated and measured return loss of the antipodal Vivaldi antenna developed in FR4 ( = 4.5, ℎ = 1.5 mm) material.

Measurement Results.
One can see from Figure 7 that the −10 dB return loss bandwidth extends from 2.3 to more than 11 GHz covering the required UWB band of 3.1-10.6 GHz. The measured and simulated results are similar validating the antenna specifications. Measured and simulated far-field radiation patterns are shown in Figure 8 in E-plane and H-plane at 4.2 GHz. The measurement is made in an anechoic chamber.
Acceptable agreement is found between measured and simulated radiation pattern.

UWB Microwave Imaging for Breast Cancer Detection
The breast is largely transparent to microwave radiation for imaging. Significant, electromagnetic property contrast may appear between healthy and tumors tissues [20]. Electrical properties of tissues at microwave frequencies have been extensively studied for dosimeter, therapy, and diagnostic applications. The electrical or dielectric properties include relative permittivity ( ) and conductivity ( ). Water is a key factor in determining tissue permittivity [21]. Low water content tissues, such as fat, have lower permittivity values than high water content tissues such as muscle and skin [21]. As described in electrical properties of tissues, less attenuation and reflection are expected from normal tissues than tumor tissues. Radar-based approaches to breast tumor detection indicate the location of strongly scattering objects, rather than creating maps of the distributions of electrical properties.
There are several different approaches to UWB radar imaging. These approaches can be divided into three categories: monostatic, bistatic, and multistatic. In the monostatic case, the transmitting antenna itself acquires the backscattered signal. Often the transmitting antenna is shifted across the breast to make a synthetic aperture. In the bistatic configuration, two antennas are used, a single transmitting antenna and a single receiving antenna. At last, in the multistatic approach, the tissue is illuminated by one transmitting antenna while the backscattered signals are recorded at several antennas placed at different positions around the breast [22].

Breast Model and Investigation of Antennas Angular Direc-
tion. For the hemispherical system, the breast is modeled as a half space of sphere. The breast model is a 10 cm diameter sphere with height 6 cm surrounded by a 2 mm thick layer of skin ( Figure 9). The materials in the simple breast models (indicated in Figure 9) are assigned appropriate electrical properties. The dielectric properties of normal breast tissue are assigned random variations of up to 10% around nominal values of = 9 and = 0.4 S/m, distributed over full sphere. The assumed contrast between malignant and normal breast tissue is approximately 5 : 1 in relative permittivity and 10 : 1 in conductivity [23]. The skin is assigned the following values: = 36 and = 4 S/m. The tumor is modeled as a sphere inside the fat with 2 cm diameter. The dielectric constant and conductivity for tumor tissue are assumed to be = 50 and = 4 S/m, respectively, over the microwave frequency band. In the proposed system the Gaussian monocycle pulse (GMP) is sent to transmitter antenna and then illuminates the breast phantom model, and the receiver antenna collects the backscattered waves from breast phantom model. In order to achieve a higher level amplitude of signal for enhancement the tumor signal, the end fire direction for antenna position is introduced.  The calculated 21 curves of vertical end fire antenna are shown in Figure 11 for separation angles of 90 ∘ , 180 ∘ , and 270 ∘ between transmitter and receiver antennas, respectively. As observed that the higher power (e.g., −25 dB) at 180 ∘ is achieved from 2.3 to 6 GHz but out of this range of frequency the power is decreased to, for example, −45 dB.
The simulated 21 curves of horizontal end fire antenna are shown in Figure 13 at the angles 90 ∘ , 180 ∘ , and 270 ∘ between transmitter and receiver antennas. As observed that the higher power (e.g., −20 dB) at 90 ∘ and 270 ∘ is achieved International Scholarly Research Notices 7 Theta y z from 2.3 to 6 GHz but out of this range of frequency the power is decreased to, for example, −30 dB, which still acceptable. From previous result we conclude that horizontal end fire antenna position with 90 ∘ and 270 ∘ between transmitting and receiving antennas is an optimum configuration for enhancing a process of tumor detection. In this work we chose horizontal end fire antenna position with 90 ∘ .

Spacing between Antennas and Breast
Surface. Now the effect of horizontal antenna separation away from the breast in direction is studied.
The received signals and 21 parameters are shown for two separation processes of 5 and 15 mm in Figures 14 and  15, respectively. It is clear that the received signal at 15 mm is weaker and delayed with respect to the received signal at 5 mm which is an expected result. Also 21 is greater at the lower frequencies at the 5 mm distance than that at the 15 mm.

Detection of Tumor at Breast Model and at Different Radii.
A breast phantom model is located between transmitter and receiver antennas and illuminated with the proposed monocycle pulse. Output signal has high amplitude for the suitable horizontal end fire antenna with 90 ∘ between the two antennas as shown in Figure 16. So, detection of the tumor can be recognized (e.g., from 1.9 to 2.6 ns) at the received signal   with signals in the level of mv which is three decades higher than that observed in [24]. 21 curves of both cases: presence of tumor and free of tumor breast are shown in Figure 17. We observed that tumor effect could be recognized.
When changing tumor radius to be 5 and 10 mm in the breast model, the received signal changes as shown in Figure 18.

Tumor Detection and Calibration Approach
It is seen from the previous tumor simulation experiments that the received signal with tumor is basically different from the signal without tumor. However the difference is not appreciable such that one cannot easily judge for the presence of tumor. Therefore a special scan scenario and signal processing are needed to clearly identify the tumor.

Preprocessing for Artifact
Removal. It is found that the recorded backscattered signals consist of two parts: the earlystage and the late-stage. The majority of early-stage parts consist of incident signals and strong reflections from skin fat interface. Meanwhile the late-stage parts include tumor response, fatty tissue response, and the multireflections   between these tissues. The skin, fatty, and tumor responses indicate that the signals directly reflected from these tissues. For identification, only tumor response is needed, and all other signals are viewed as superposition early-stage artifact and the late-stage clutter [25]. These artifacts can be several orders of magnitude greater than the desired tumor response; thus they must be removed before applying any image reconstruction algorithm. In our breast cancer imaging system horizontal end fire antennas are placed with 90 ∘ between transmitted and   Figure 19 illustrates the shape of received signals at different antenna position. Calibration steps are depending on the assumption that the signals recorded at different antenna locations have similar incident pulse and skin backscatter content [25]. The tumor signal is different in amplitude and time as a result of multipath propogation and different tumor antenna distance. This variation is observed (e.g., from 1.8 to 2.6 ns) as the tumor response is observed.

Analysis of Artifact Removal and Tumor Detection.
In order to remove the artifacts and extract the tumor signal one used a calibration algorithm called subtract and add, SAD algorithm. In the first step of the algorithm one subtracts the signals between two adjacent locations and second step is summed subtract signal from adjacent location at the same time instant.
Assuming that the received signal from the breast at position is ( ) and at the position + Δ is +1 ( ), then the tumor signal ( ) can be expressed by difference [26].
By repeating the measurements at × -locations and building the difference between two neighboring locations, one can fortify the tumor identification signal by summing    Figure 20 shows the tumor signals ( ) at different locations from 2 cm to 3.5 cm in steps of 0.5 cm. It is clear that the different subtracted signals are appreciable and in the order of mV. Figure 21 depicts the overall accumulated tumor signal resulting from the all scanned positions of the receiving antenna. It is clear from the figure that the tumor signal becomes amplified and so very identifiable.

Conclusions
A system of UWB breast tumor detection is designed. It consists of ultrashort monocycle pulse generator, transmitter and receiver antennas, and breast with and without tumor model. The transmitter employs an impulse-generator network to achieve less distortion, low ringing, and low power consumption. It produces monocycle pulses with 235 ps pulse duration and about 600 mV peak-to-peak voltage. The waveform produced by this transmitter is characterized by low ringing level and good balance between the positive and negative parts. Antipodal Vivaldi antenna with an extremely wide −10 dB return loss bandwidth from 2.3 GHz to more than 11 GHz has been developed. This antenna element features directive over the required ultra-wideband.
Many simulation experiments have been carried out to identify and maximize the tumor signal. It is found that one can achieve a tumor signal in the order of mill volts. One can multiply this signal by accumulating the tumor signal resulting from different scanning positions of the receiver antenna.
This work proves that one can build an effective breast tumor detector using the UWB pulse technique together with some signal processing.
It is intended to complete building the system in the future.

Conflicts of Interest
The authors declare that they have no conflicts of interest.