Wideband Characterization of RF Propagation for Time-of Arrival Localization of Wireless Video Capsule Endoscope Inside Small Intestine

Transcription

1 Wideband Characterization of RF Propagation for Time-of Arrival Localization of Wireless Video Capsule Endoscope Inside Small Intestine Zhuoran Liu, Jin Chen, Umair Khan, Bader Alkandari and Kaveh Pahlavan Abstract Localization of Wireless Capsule Endoscope (WCE) inside the small intestine is a very challenging problem, which is very important for the diagnosis of certain diseases. Previous studies have shown that Time-of-Arrival (TOA) localization can be more effective than Received Signal Strength (RSS) based localization. Performance evaluation of TOA-based localization techniques for WCE requires understanding of wideband RF propagation characteristics from the internals of small intestines. Since these measurements are difficult, we need to resort to simulation techniques and validating the results of simulations with empirical measurements. In this paper, we first validate the measurements of wideband characteristics of RF signals pertinent to TOA in homogeneous tissues using a SEMCAD X simulation software platform. Then we use the software for simulation of TOA from inside the human body and the small intestine. Finally, we analyze and assess TOA techniques regarding the nonhomogeneous environment inside the human body. Keywords SEMCAD X, simulation, nonhomogeneous tissue, Wideband measurement, BAN application, small intestine, Wirelss Capulse Endoscopy I. INTRODUCTION Localization of WCE inside the human body is of great importance in many applications of Body Area Networks (BAN). WCE provides images to gain information in order for the doctors to locate internal abnormalities such as bleedings, infections and tumors [1]. Common techniques that provide an estimate of the location of these internal abnormalities use the TOA and the RSS of the transmitted signal from the WCE to locate the capsule [2]. Research in localization inside of the human body has reached a bottleneck due to the difficulty of conducting measurements inside the human body. Two major limitations causing difficulty are the existence of a nonhomogeneous environment and difficulties in antenna implantation inside the human body for experimental purposes. Previously, the efficiency of different simulations around a human body was assessed and theoretically analyzed [3]. Phantoms with emulated tissues were used to validate surface measurements of a human body [4]. However, the simulation analysis of the small intestine remains to be done. Center for Wireless and Information Networks Studies (CWINS) Department of ECE, Worcester Polytechnic Institute (WPI) {zliu, jchen3, uikhan, baalkandari, kaveh}@wpi.edu The simulation was carried out using SEMCAD X. SEMCAD X is a full-wave electromagnetic simulation platform based on the Finite Difference Time Domain (FDTD) method. This software provides an abundant library of anatomical nonhomogeneous human body models for waveform transmission problems. The models can be used to simulate wave propagation in and around the human body. Additionally, this software runs faster than other electromagnetic simulation platforms due to its algorithm optimization [5]. The FDTD method was first introduced by Yee in It solves Maxwell s curl equations in the time domain. The FDTD method has been proven to be an effective simulation method in terms of the accuracy of obtaining electrical and magnetic field parameters [5]. It has been widely used in indoor localization and microwave simulations [11]. In our work, we simulated the waveform propagation in both homogeneous and nonhomogeneous tissues. By comparing simulation results and empirical measurements, we show that the SEMCAD X platform is a reliable tool for waveform transmission. In this paper we first describe the fundamental difference between free space RF propagations and propagation in the liquid like lossy environment inside the human body. Then, by matching the results of empirical wideband measurements in an emulated homogenous body tissue and the results of computer simulation using SEMCAD X, we demonstrate the validity of simulations to replace wideband empirical measurements. Finally, we use complex simulations inside the full body with organs to extract characteristics of wideband signals pertinent to localization inside the human body. Following this section, this paper is outlined as follows: Section II describes the scenario and setup of experimental measurements and simulations in homogeneous tissues. Section III compares the measurement findings and simulation results and accordingly evaluates the reliability of SEMCAD X. Section IV discusses simulation in nonhomogeneous tissue inside the human body with emphasis on the small intestine, and analyzes the effects of bandwidth and distance on the TOA. Finally, Section V summarizes our findings and concludes the paper.

2 II. MEASUREMENT AND SIMULATION SCENARIOS IN HOMOGENEOUS TISSUE In free space propagation, the power transfer ratio is given by the Friis equation. This equation is valid if the transmission media has negligible losses. However, homogeneous tissues are mediums with heavy losses mainly due to absorption. This requires a modification on the Friis power formula. This extended formula is given by where P r is the received power in Watts (w), P t is the transmitted power in Watts (w), G t and G r are the average transmitter and receiver antenna gains, respectively, λ is the wavelength in meters (m), r is the distance between the transmitter and receiver antennas in meters (m) and α is the absorption coefficient of the transmission medium. In path loss terms it can be written as where is the total path loss and PL 0 is the path loss in the first meter. Eq. 2 shows that the path loss is a function of the classical log-distance term and a linear term. The linear term arises from the absorption property of the lossy propagation medium. Figure 1 shows the plot of Eq. 2 versus the effect of the absorption loss term, the log-distance term and the model presented in [9]. The model in [9] uses a log-distance relationship. For small distances ranging from 2 cm to 8 cm, the model in [9] and the result in Eq. (2) are very close. A. Measurement Scenario In this paper, we conducted the measurements at multiple bandwidths from 50MHz to 2GHz. The center frequency of these measurements is 6.5GHz. All the measurements are obtained in a rectangular box using a Vector Network Analyzer (VNA) and UWB antennas. The following is a description of all the measurement equipment and setup. The length, width and height of the box are 59.69cm 21.59cm 20.32cm respectively, as shown in Fig. 2. When performing measurements, the box was filled with homogeneous tissue in order to emulate the propagation environment inside the human body. Based on results published in [6], the effect of multipath can be ignored if the transmission antenna is inside homogenous tissue. The permittivity of the homogenous tissue ranges from 79 to 81 A plastic plate was placed on top of the box. The plastic plate is used to hold the transmitter and the receiver antennas. The transmitter antenna was fixed on one end and the receiver antenna was moved in incremental step distances. The incremental step size for the (1) (2) distance is 1 cm. The maximum distance between the transmitter and receiver in this measurement scenario is 11cm. An E8363B Vector Network Analyzer (VNA) is used to measure the channel characteristics between the transmitter and receiver antennas in the box, as shown in Fig. 3. The VNA sweeps the channel in the frequency domain using single tones. The transmitted tone power is set to 0dBm and the bandwidth at the receiver to detect the tone is set at 3 khz. The spectrum is swept from 50MHz to 7GHz (center frequency is 6.5GHz), by changing the frequency of the tone with discrete values. The number of points during the measurement is 401. The measured frequency domain transfer function S21 is stored in an external computer. The Inverse Fourier Transform (IFT) of the measured frequency response is used as the channel impulse for the analysis of multipath arrivals. In practice, when we obtained the location information from the impulse response, we always used a Hanning Window on the measured frequency response to reduce the effects of side lobs and then used the Inverse Chirp-Z transform [7], rather than traditional IFT, to convert the data in a more computationally efficient manner. B. Simulation Scenario to Match Measurements In order to compare the simulation results with measurement results, we modeled the measurement flat box in SEMCAD X with the dimensions shown in Fig. 2. The plastic box is filled with homogeneous tissue. The homogeneous tissue permittivity is set to 81. We can simulate a variety of homogeneous tissues with different permittivity values is by mixing different substances with water [12]. The box is made from polyethylene, with the permittivity of In the simulation, dipole antennas are used at the transmitter and the receiver. The impedances of dipole antennas at the two ports are matched at 50Ω. The amplitude of the input voltage signal is set to 1V. Fig. 1. Comparison of the linear and logarithmic models for path-loss in lossy liquid-like homogeneous tissues.

3 Fig. 2. Flat box used for measurement in homogeneous tissues. Fig. 3. Measurement scenario for the analysis of the effects of distance received signal waveform. Fig. 4. Flat box model for simulation of radio propagation in homogenous tissues using SemCAD X To match the measurement settings, the center frequency of the system is 6.5GHz and the bandwidth ranges from 50MHz to 7GHz. We used a transmitted power of 10dBm for measurements and calibrated the results of measurements and simulation to compensate for extra path losses on the cables used in measurements. To simulate the input signal from the Vector Network Analyzer, we use the Hanning pulse as the input signal given by make full use of SEMCAD X, we must verify that the FDTD simulation system in SEMCAD X is reliable. We simulated the homogeneous tissue measurement at different bandwidths and distances. Afterwards we compared the simulation results and the measurement results to check how well they fit. A. Bandwidth Limitation Analysis in Homogeneous Tissue Figure 5 shows a comparison between the simulation results of homogenous tissue with measuement results with bandwidths of 50, 100, 200 and 400 MHz. It is easy for us to distinguish the different bandwidths througth the width of the pulse of each figure. We can see from Fig. 5 that the simulation results closely fit the measurement results when the bandwidth is smaller than or equal to 400MHz. When the bandwidth exceedes 400 MHz the deviation between the simulation results and collected measurement results become larger. We conclude that the simulation matches measurements in homogeneous tissue when the bandwidthis less than 400MHz with a grid size of 0.16mm. After we confirmed that the SEMCAD X simulation is reliable, we wanted to know the effect of distance on Time-of-Arrival (TOA) in homogeneous tissue. We have done the measurement at 50MHz of different distances from 2cm to 6cm in water. B. Effect of Distance on TOA in Homogeneous Tissue During the measurement, we collected ten sets of data at each measurement distance. Then we applied the Inverse Chirp Z transform to the collected S21 parameters to both simulation results and collected measurements to obtain the impulse response of the channel each distance. Figure 6 shows the RSS versus distance for the simulation model and the measurements. There are ten measurement result points and one simulation result point at each distance increment. We see that the difference in RSS between measurement and simulation is very small when the distance is smaller than 6cm. Also, from the left part of Fig. 7 we can see that the TOA values of the measurement and the simulation are very close when the distance is smaller than 6cm. At a distance of 6cm, the TOA of the measurement result is 36.7ns and the average of received signal amplitudes is -80dB, which is significantly inaccurate. We conclude that the distance limitation in homogeneous tissue 5cm at the bandwidth of 50MHz and propagation medium permittivity of 80. (3) where N represents the width of the pulse. In SEMCAD X, we can easily obtain the S-parameters of 2-port networks. To obtain the impulse response of the system we apply the Inverse Chirp Z transform to the S21 parameter. III. VERIFICATION OF SEMCAD X SIMULATION WITH MEASUREMENT IN HOMOGENEOUS TISSUE From the previous studies [4], we already know that the simulation system based on FDTD method is a LTI system. To

4 TOA [ns] DME [cm] amplitude [db] Fig. 5. Simulation and Measurement results: Amplitude vs. distance at 50MHz, 100MHz, 200MHz and 400MHz water 50M amplitude VS distance simulation result measurement result measurements and simulation at different distances, which are shown to fit well with each other. IV. SIMULATIONS IN NONHOMOGENEOUS TISSUES It is difficult to do the measurement inside the human body as we mentioned in the previous sections. In this section, we investigate if we can obtain useful localization information inside the human body. We conduct simulations inside the human body using the models provided in SEMCAD X. A. Simulation Scenario in Nonhomogeneous Tissue We simulated an in-body waveform transmission. We imported a model of a 34-year male and positioned two sensors internally around the human torso. In these simulations, we fix the transmitter at one location and move the receiver at increments of 1 cm in a similar fashion to the experimental setup. The distance range between the transmitter and receiver is 1cm to 10cm. The human body model is shown in Fig. 8. with all the organs in the torso area. Figure 8 also shows the layout of the sensors. The red point indicates the position of the transmitter and the blue points indicate the ten positions of the receiver TOA water 50M 5 Distance Measurement Error at 50MHz in water simulation result measurement result 4.5 simulation result measurement result logarithmic of distance (log10(d)) Fig. 6. Received Signal Strength (RSS) versus distance at 50MHz From Fig. 6, we can obtain the line with the relationship (4) Fig.7. TOA versus distance (left) and Distance Measurement Error (DME) versus distance (right) at 50MHz in water where represents the y-coordinate value and represents the x-coordinate value. The Distance Measurement Error (DME) of both measurement and simulations can also be calculated from DME versus distance shown in Fig. 7. Propagation velocity in a homogeneous tissue is given by [9]: (5) Then the DME can be obtained from: (6) Where ε is the DME, is the relative permittivity, d is the actual distance, is the speed of light and is the velocity in the homogeneous tissue. Figure 7 shows the DME of for Fig.8. the Human Body Model (Left), Locations of Transmitted and Received Sensors (Upper Right) and Path in Small and Large intestine (Lower Right)

5 amplitude [db] time [ns] To be consistent with the results of measurements, we used a Hanning window on the input signal in the SEMCAD X simulations in nonhomogeneous tissue as well as in homogeneous tissue. We conducted a set of simulations at different bandwidths at the distance of 2cm between the transmitter and receiver. The bandwidth range is from 50MHz to 400MHz. Additionally, we wanted to know the detailed waveform transmission situation in the human GI tract. We simulated the path in the small intestine using a series of receiver sensors inside it with the help of the Python scripting tool in SEMCAD X. The layout of edge sensors is shown in Fig. 8. The distance between adjacent edge sensors is 1 cm. B. Simulations Inside Human Body Around Torso Part As mentioned in Section II, we placed two edge sensors inside the human body model at the torso area. One sensor acts as a transmitter and the other sensor as a receiver. We fix the transmitter at a single location and move the receiver at increments of 1 cm. Using this procedure we obtain the values of TOA and received signal strength at different distances. The results are shown in Fig. 9. In the left part of Fig. 9, we fitted a linear line to the ten points of the amplitude at different distances. The polynomial equation of the fitted line is presented as where we use and to represent y-coordinate and x-coordinate values respectively. In a similar fashion, the relationship between TOA and distance can be described the polynomial equation of the fitted line shown in Fig. 9 by (7) On the other hand, according to Fig. 9, we find that the amplitude is lower than -140dB which is too small to be detected from the noise in real measurements when the distance is larger than 5cm. In conclusion, in order to ensure the accuracy of both simulations and measurements, the distance limitation at the human torso area inside human body is 5cm at the bandwidth of 50MHz. By averaging the permittivity inside we can solve future measurement problems much more easily. For example, if we know the permittivity values for several nonhomogeneous tissues, we can replace nonhomogeneous tissue with a homogeneous tissue that uses the average permittivity. C. Simulations of the Path in GI tract We already know the situations around the human torso part. Because of the difficulty in getting localization information in the small intestine, we decided to focus our preliminary simulations in that area. The CAD model of the small and large intestine shown in Fig. 8 was imported into the human body model already present in SEMCAD X [10]. The Chirp Z transform of the S21 parameters gave us the impulse response at each sensor point, as shown in Fig. 10. Figure 10 shows that there are some multipath effects inside the body, as well as significant path loss as the transmitter travels away from the surface of the abdomen. Figure 11 presents the RSS at different location points on the path inside the small and large intestine. We can calculate distance power gradient α inside this GI tract from the RSS readings resulting in α = 80/18 = 4.44.] We compare our results with the path loss model obtained in [9], which is given by (11) To compute the average velocity in the human body, we take the inverse of the slope in Eq. 8 and obtain (8) (9) where is the reference distance, and is the path loss gradient. S represents the deviation in db caused by different materials and antenna gain in different directions [8]. The path loss gradient under different situations is given in Table 1. We find that our calculated simulation path loss gradient is a close match to the to the deep tissue path loss gradient in Table Amplitude VS Distance fitted curve TOA VS Distance fitted curve Then we can calculate the average permittivity of the human body using Eq. 6 resulting in (10) Fig. 9. RSS versus distance (left) and Time-of-Arrival (TOA) versus distance (right) inside human body

6 permittivity inside the human body around the torso area is Therefore, we have solved the localization problem inside human body using a TOA-based technique on wideband signals. Fig. 10. Ten consequtive channel impulse response from inside the small intestine VI. ACKNOWLEDGEMENT The authors would like to thank Mr. Guanqun Bao from the Department of Electrical and Computer Engineering, Worcester Polytechnic Institute (WPI), for helping establish the simulation model. The authors would also like to thank other members in the CWINS lab of WPI for their useful suggestions and constructive criticism. REFERENCES [1] S. Li, Y. Geng and K. Pahlavan, Analysis of Three-Dimensional Maximum Likelihood Algorithm for Capsule Endoscopy Localization, 5th IEEE International Conference on Biomedical Engineering and Informatics, Chongqing, China Oct [2] J. He, Y. Geng and K. Pahlavan, Modeling indoor TOA Ranging Error for Body Mounted Sensors, 2012 IEEE 23nd International Symposium on Personal Indoor and Mobile Radio Communications (PIMRC), Sydney, Australia Sep [3] U. Khan, K. Pahlavan, Umair Khan, Computational Techniques for Comparative Performance Evaluation of RF Localization inside the Human Body, Fig. 11. RSS versus distance of the path in small intestine (upper) and Path loss versus distance of the path in small intestine (lower) Table 1. Referenced path loss model parameters for the human body V. CONCLUSIONS In this paper we have validated the simulation inside the homogeneous tissues. The limitations in bandwidth and distance are also discussed. We imported the path of movement inside the small intestine into SEMCAD X and demonstrated that the TOA characteristic of the signal transmitted from the inside of the small intestine can be measured. The bandwidth of simulation in homogeneous tissue is limited to 400MHz when the boundary is as large as the plastic box, while the distance in homogeneous tissue of measurements is limited to 5cm. We also found the distance limitation in nonhomogeneous tissue to be 5cm. The averaged [4] P. Swar, K. Pahlavan and U. Khan, Accuracy of localization system inside human body using a fast FDTD simulation technique Medical Information and Communication Technology (ISMICT), La Jolla, CA, USA, March, 2012 [5] Schmid & Partner Engineering (SPEAG), SEMCAD-X Manual [6] Jin Chen, Yunxing Ye, Kaveh Pahlavan, Comparison of UWB and NB RF Ranging Measurements in Homogenous Tissue for BAN Applications, in April 17-19, Phoenix, Arizona, IEEE WTS, 2013 [7] df [8] Makoto Kawasaki, Ryuji Kohno, A TOA based Positioning Technique of Medical Implanted Devices [9] K. Sayrafian-Pour, W.-B.Yang, J.Hagedorn, J. Terrill, and K. Yazdandoost, A statistical path loss model for medical implant communication channels, in Personal, Indoor and Mobile RadioCommunications, [10] G. Bao, Y. Ye, U. Khan, X. Zheng and K. Pahlavan, Modeling of the Movement of the Endoscopy Capsule inside G.I. Tract based on the Captured Endoscopic Images, International Conference on Modeling, Simulation and Visualization Methods, Las Vegas, [11] H. Sawada, et al., "Channel model for wireless body area network," 2nd IntI.Symp. on Med. Info. and Comm. Tech. (ISMICT), Dec [12] S. Gabriel, et al., "The dielectric properties of biological tissues: III. Parametric models for the dielectric spectrum of tissues," Phys. Med. BioI., Vol. 41, pp , 1996 [13] Jaechun Lee and Sangwook Nam, Effective Area of a Receiving Antenna in a Lossy Medium. IEEE Transactions on Antennas and Propagation,Vol. 57, No.6, June 2009.