Investigation of radar localization system accuracy for human gastro intestine (GI) tract

2013 7th International Symposium on Medical Information and Communication Technology (ISMICT)

Investigation of Radar Localization System Accuracy for Human Gastro Intestine (GI) Tract Perzila Arab, Michael Heimlich, Eryk Dutkiewicz Department of Electronic Engineering, Macquarie University Email: [email protected], [email protected], [email protected]

Abstract—Recent developments in capsule endoscopy have highlighted the need for accurate techniques to estimate the location of a capsule endoscope. A high accuracy localization of the order of millimeters is required for endoscopic applications. Location estimation of a capsule endoscope in the gastrointestinal (GI) tract is a challenging problem, as radio frequency signals encounter a high loss and highly dynamic channel propagation environment. In this paper, the characteristics of radio frequency signal absorption inside the human body are investigated with the aim of developing an accurate propagation model. Furthermore, the possibility of using a radar system for capsule localization is investigated and compared to its operation in the 2.4GHz ISM band. Index Terms—Capsule endoscope, Channel model, Radar

I. I NTRODUCTION Endoscopy is a medical procedure that has been used to inspect and diagnose possible abnormalities and diseases inside the human body. This technology has been developed for examination of various parts of the body. Recently, gastroscopy and colonoscopy have revolutionized the diagnosis and treatment of diseases of the upper gastrointestinal (GI) tract (esophagus, stomach, and duodenum) and the colon, respectively. The last remaining part in the GI tract is the small intestine. Monitoring the small intestine is not possible by the aforementioned endoscopy techniques due to its long length and complicated shape. Increasing the demand to improve diagnostic capabilities in the small intestine has resulted the development of a new type of endoscopy, known as capsule endoscopy. Improvements in this technology can be achieved by addressing several issues such as active locomotion, high power efficiency, high data rate, high quality of images, remote powering of the capsule and accurate location of the capsule. Among these, capsule localization is a crucial parameter that needs to be precisely estimated. The exact location information of the capsule gives the doctor a clear idea about the location of any abnormalities detected in the received video images. So far two main methods have been proposed to determine the capsule location in the digestive tract. These can be categorized as magnetic field strength and electromagnetic wave based methods [1]. Recently, a magnetic localization technique for tracking the capsule in the GI-tract has been proposed in [2],[3],[4]. Authors in [2] state that the human body has little influence on low-frequency magnetic signals, so high accuracy is achievable for the localization of the target. Moreover, this technique is faster compared to other existing methods such as computed tomography (CT), ultrasound and

978-1-4673-5772-2/13/$31.00 ©2013 IEEE

magnetic resonance imaging (MRI). Despite its high accuracy, this method is not very practical. The main reason is that the required hardware may significantly increase the capsule size. It also requires special infrastructure, which limits its application to medical settings. Electromagnetic wave based localization is the other proposed method for capsule localization. Various technologies such as Radio Frequency (RF), Radio Frequency Identification (RFID), X-ray and Gamma ray have been explored in several studies. None of these approaches can, however, produce the required accuracy of the order of millimeters. The aim of this paper is to explore the suitability of radar technology for endoscopic capsule localization. We consider three frequencies in the range of GHz for the radar operation: 3.4GHz, 4.8GHz and 5.8GHz. We also compared the results of the radar performance against a narrowband radar operating in the 2.4GHz ISM band. The remainder of the paper is organized as follows. A discussion about most common ranging techniques used for localizationis presented in Section II. In section III, we discussed radar technology and its suitability for endoscopic capsule localization. The human abdomen composition and the corresponding dielectric properties are presented in Section IV. A channel model for the human abdomen is presented in Section V and preliminary performance evaluation of the radar system is discussed in Section VI. Finally, Section VII concludes the paper. II. OVERVIEW O F RF L OCALIZATION M ETHODS In this section three traditional RF ranging techniques are introduced followed by a discussion of their pros and cons. 1) Received Signal Strength (RSS): Signal strength-based ranging is a simple technique to estimate the distance from a target node. Distance information can be extracted from energy or power of a signal sent between the two nodes. For this purpose, the relation between distance and signal energy should be considered. In this method, the strength of the received signal highly depends on the propagation channel characteristics and is influences by factors such as pass loss, shadowing, fading, reflection, scattering which should be carefully considered in the distance estimation. 2) Time of Arrival (TOA): Similar to the RSS technique, Time of Arrival is another signal parameter that can be extracted by measuring the one-way signal propagation time between a transmitter and a receiver. The distance between

144

the two nodes can be calculated based on the measured time. To achieve an accurate result, this technique needs precise synchronization between the two nodes [5]. When synchronization between the two nodes is not available, a twoway ranging technique can be used to calculate the distance between the two nodes. When multiple reference nodes with known locations are used, the above ranging techniques can be used to determine the target node’s location. Another related technique is Time Difference of Arrival (TDOA) which is used for target node localization. It uses the differences between the times of arrival of a signal from the target node to multiple synchronized reference nodes at known locations [6]. 3) Angle of Arrival (AOA): Angle of Arrival is another position related technique that determines an angle between two nodes by estimating the AOA parameter of a traveled signal between the nodes. Generally, an array of antennas are employed for estimating the AOA of the received signal [6]. To achieve a precise result in position estimation, the process of measuring the angle must be accurate. Discussion: RF signal based localization methods are usually based on one of the traditional techniques using RSS, TOA or AOA measurements. The accuracy of these techniques is highly influenced by the radio propagation environment where they are deployed. The human body is a challenging medium for RF signal propagation. It has several types of tissues varying in thickness, impedance and dielectric constant. The human body causes different radiation pattern disruptions, power absorption and shift in central frequency depending on the selected frequency of operation. Power absorption can vary according to the characteristics of the tissue [7]. The human body channel also experiences fading and multipath effects that impact on the accuracy of the localization method. In the RSS method, the strength of a received RF signal can be measured during normal data transmission. RSS measurements are generally simple to implement [8]. The RSS based methods are also less affected by the bandwidth limitation but the RSS measurements variability due to the dynamic channel conditions have a significant impact on their accuracy. In the TOA technique, ranging resolution is easily degraded by limited bandwidth. Also, the accuracy of the estimated range is highly impacted by the GI filling and emptying cycle and GI movement [9]. Furthermore, the propagation velocity inside the human tissue is not constant. Therefore, when designing a TOA based localization system, the propagation velocity inside the human body must be estimated based on average permittivity of the human organs and tissues [10]. AOA is not a reliable technique for localization of an object even in indoor environments and it is not considered as an appropriate technique for localization of the capsule, since the human body has a more complex environment than any other indoor environments [11]. III. R ADAR T ECHNOLOGY Radar is an electrical system used to determine the range altitude, direction, or velocity of an object. The basic principle of radar operation is very simple. A Radio frequency (RF)

wave is transmitted by a transmitter toward a target. In the receiver side the detection of the target will be applicable when the antenna receives the reflected wave from the target. In this project the target (endoscopic capsule) is in motion, therefore the radar technology can be a suitable candidate technique for object localization. Moreover, based on radar performance, there is no need to consider any extra transmitter/receiver at the target side since the location of the target can be calculated with regard to the reflected signal from the target. Thus, no extra hardware needs to be designed for the capsule node, resulting in a low complexity system. One of the important concepts in radar systems is Radar Cross Section (RCS). RCS is a measurement that shows the detectability of an object by a radar system. In general, an object can be detected based on the amount of energy that was reflected from it, and the RCS of a target shows how much of the incident power is intercepted, reflected or directed back toward the receiver. Thus, the target with a larger RCS can be more easily detected by a radar system. There are various factors such as size and material of the object, incident angle and reflected angle that affect the amount of reflected energy to the source and RCS respectively [12],[13]. In this study we assume that the capsule endoscope has a simple cylindrical shape with size of 11mm × 26mm so the related cross section can be calculated from the following formula: 2πrh λ2

(1)

C V =√ r

(2)

σ= where

where, r and h are the radius and the height of the capsule, respectively and λ is the wavelength. Since, the propagation velocity through the human body is different from free space, we calculate it based on the average permittivity of the human tissues. So, the aforementioned λ is the wavelength in the human body rather than free space. IV. H UMAN A BDOMEN C OMPOSITION In this section, the effect of varying thickness of the body tissues on the received power will be investigated. The interior of the human abdomen is considered as the propagation channel since the signal should pass through the abdomen area to reach its desired target in the small intestine. In this paper, a layered plannar model as shown in Figure 1 will be used to demonstrate the propagation channel of the abdomen area. As can be seen in this figure, two types of fat are considered for the propagation channel. Visceral fat refers to the fat in the abdomen cavity and between the organs, whereas subcutaneous fat refers to the fat found just beneath the skin. These layers have a different thickness and it can vary by the age and sex. Table I tabulates the thickness of these different layers in a female subject. In this table the tissue is divided

145

Fig. 1.

where

Layered planner tissue model

v " # r u  σ 2 u µ t −1 α=ω 1+ 2 ω Small intestine

Visceral fat

Sub-cutaneous fat

Abdomen muscle

Skin

into low, mid and high tertile categories based on visceral fat [14],[15],[16],[17]. TABLE I A BDOMEN T ISSUE T HICKNESS

Tissue

Range of thickness (mm)

Here α is the absorption coefficient, d is the traveled distance in the tissue (i.e. thickness of the tissue), σ is conductivity, ω is the angular frequency µ is the permeability and  is the permittivity. Based on the table I, the thickness of each tissue layer has been selected for the corresponding calculation. Therefore, the thickness of the skin, subcutaneous fat, abdominal muscle and visceral fat are fixed to 1.5mm, 17mm, 8mm and 15mm, respectively. According to the selected thickness and each tissue’s parameter the absorption loss is then calculated. B. Refraction loss

Low tertile

Mid tertile

High tertile

Visceral fat

15-36

37-47

47-98

Subcutaneous fat

17-34

17-34

20-33

Abdomen muscle

8-16

8-16

8-16

Skin

1.1-1.6

1.1-1.6

1.1-1.6

In addition to the loss due to the absorption of each tissue layer, the loss due to a reflection by the layer’s boundary should be also considered. The calculation of the reflection loss is given by: √ (5) PL (n) = r .µr

To calculate the power loss due to different tissue thickness and their dielectric properties, frequencies of 2.4GHz, 3.4GHz, 4.8GHz and 5.8GHz have been selected. Table III summarizes the dielectric characteristics of the different tissue layers at these frequencies. These properties have been obtained from [18]. It can be seen that the small intestine and the abdomen muscle have the largest relative permittivity at all frequencies. V. C HANNEL M ODEL In this study we assume that the transmitter is located on the human abdomen. It transmits 1mW power, which needs to pass through the human abdomen until it hits the capsule endoscope inside the small intestine. The incident signal will then be reflected through the human body and the receiver, which is also co-located with the transmitter, will collect the reflected signal. In the next step, based on the received signal strength and an algorithm relating the signal strength to distance, the capsule distance can be estimated. The calculation of the capsule distance from the transmitter is, however, out of the scope of this paper. To develop an efficient channel model, we consider three different losses that attenuate the signal along the path. Those can be calculated as follows. A. Absorption loss The absorption loss of transmitted power in each tissue layer as related to its thickness d is given as: PL (d) = e−αd

(4)

(3)

where r is the relative permittivity of each tissue layer that can be obtained from table III and µr is the relative conductivity that can be assumed close to 1 for most materials. C. Free space loss In this study, the transmitter and receiver are assuming to be co-located on the human abdomen. Therefore, there is no need for the free space loss calculation. However, if they were located away from the human abdomen, one-way free space loss of the propagated wave between the radar system and the human body would then be obtained from the free-space path loss formula:   4πd PL (d) = 20 log (6) λ where d is distance between transmitter and receiver and λ is the wavelength. VI. P ERFORMANCE EVALUATION OF RADAR SYSTEM Once the propagation loss has been calculated, the received power by the radar antenna can be derived based on the radar equation.

Pr =

Pt .Gt .Gr .λ2 .σ (4π)3 .d4

(7)

Where Pt is the transmuted power, Gt and Gr are the gain of the transmitter and receiver respectively, σ is the radar cross section, λ is the wavelength and d is the distance of the target from the radar system. Here, we assume the gain of 22dB for our transmitter and receiver. Figure 2 shows the theoretical results of the calculated power loss at the receiver versus distance at the four selected frequencies.

146

−100 2.4GHz 3.4GHz 4.8GHz 5.8GHz

−0.8

−1 −200

p2−p1 (dB)

Power Loss at Receiver (dBm)

−150

−250

−1.2

−1.4

−300 −1.6

−350

−400

2.4GHz 3.4GHz 4.8GHz 5.8GHz

−1.8

4

Fig. 2.

5

6

7

8

9 10 Distance (cm)

11

12

13

−2

14

Calculated power loss at Receiver versus distance

Fig. 3.

As can be seen, the system at 2.4GHz experiences lower power loss compared to the system operating at higher frequencies of 3.4GHz, 4.8GHz and 5.8GHz. This result is as expected, since the human body tissues have higher dielectric values at higher frequencies. On the other hand, in this analysis, we just considered 10 cm movement of the capsule in the small intestine. The received power loss falls below -120dBm after 1cm movement of the capsule at 2.4GHz frequency. The situation at the three other remaining frequencies becomes even worse, since their power loss is below than -140dBm from the beginning. If we assume that our receiver has a noise floor of -120dBm, the received power, in all the four aforementioned frequencies needs to be amplified due to the large amount of attenuation at the beginning of the capsule movement in the small intestine, otherwise the received signal cannot efficiently be detected. To compare the desirability of the selected frequencies in the GI localization system and evaluate the system performance, the Figure of Merits (FOM) was calculated for all four frequencies. For this purpose, the received power loss variation with respect to 1mm movement of the capsule was plotted and the FOM for all frequencies were calculated based on the plotted graph. As can be seen in Figure 3, for the small changes of about 0.2dB in the received power, the capsule should move 8.9cm, 10.3cm, 6.6cm and 27cm at each of the considered frequency 2.4GHZ, 3.4GHz, 4.8GHz and 5.8GHz, respectively. The calculated FOMs for these frequencies are listed in table II. Based on calculated FOM results, the system at 4.8GHz has the highest FOM followed by 2.4GHz, 3.4GHz and 5.8GHz. This, however, reveals that the system at 4.8GHz faces a higher path loss compared to 2.4GHz and 3.4GHz but its received power has also higher fluctuations in the small amount of capsule movement. The received power variation due to the small movement of the object is an important factor for localization, since it provides a high resolution system.

10

20

30

40 dl(cm)

50

60

70

80

Power variation due to 1mm movement of capsule endoscope TABLE II F IGURE OF M ERIT Frequency

Figure of Merit

2.4GHz

0.022

3.4GHz

0.019

4.8GHz

0.03

5.8GHz

0.0074

VII. C ONCLUSION This paper has investigated the possibility of using a radar system for GI tract localization in the GHz range frequency. In this investigation, the aim was to study the effect of the propagation loss due to different human tissue on the incident and reflected signals to and from the endoscopic capsule, in order to evaluate the suitability of the radar system for in-body localization. The study has shown that the system at higher frequencies faces higher loss, since the absorption loss in different tissues is becomes significantly high. Also, the small intestine and muscle have the highest amount of absorption loss among the other tissue layers. On the other hand, our study shows that the 4.8GHz frequency can provide better resolution for localization purposes since in the small movement of the capsule, the variation in the received power is much more significant compared to the other frequencies. However, the system at 4.8GHz faces higher propagation loss than at 2.4Hz and 3.4GHz, thus the received power may need to be captured by applying a highly sensitive receiver. Among the studied frequencies, 2.4GHz tolerates lower loss. But the important point is that the capsule should move about 6m in the high loss environment (small intestine) and the received power loss even at 2.4GHz will fall below the expected noise floor after about a 1 cm movement. Hence deploying a

147

TABLE III P ROPERTIES OF HUMAN

ABDOMEN TISSUE IN DIFFERENT SELECTED FREQUENCY

Frequency

Tissue name

Conductivity [S/m]

Relative permittivity

Loss tangent

Wavelength [m]

Penetration depth [m]

2.4GHz

Fat Muscle Skin Dry Small Intestine

0.10235 1.705 1.4407 3.1335

5.2853 52.791 38.063 54.527

0.14503 0.24191 0.2835 0.43042

0.054193 0.017069 0.02005 0.016553

0.11956 0.022785 0.022956 0.012785

3.4GHz

Fat Muscle Skin Dry Small Intestine

0.15028 2.4709 1.9655 4.0128

5.1839 51.568 37.092 52.66

0.15326 0.25333 0.28016 0.40287

0.038615 0.012183 0.01434 0.01192

0.080667 0.01555 0.016607 0.009786

4.8GHz

Fat Muscle Skin Dry Small Intestine

0.22993 3.8279 2.9076 5.5166

5.048 49.8 35.936 50.306

0.17058 0.28785 0.30299 0.41067

0.027698 0.008762 0.010304 0.0086327

0.052061 0.0098858 0.011067 0.0069623

5.8GHz

Fat Muscle Skin Dry Small Intestine

0.29313 4.9615 3.717 6.7459

4.9549 48.485 35.114 48.672

0.18335 0.31715 0.32807 0.42955

0.023125 0.0073337 0.0086106 0.0072505

0.040481 0.0075413 0.0085735 0.0056102

highly sensitive receiver is also a crucial requirement at this frequency. In GI localization using a radar system, there is no need to consider any extra hardware in the capsule. However, the down side is the necessity of considering two-way propagation loss, which requires a highly sensitive receiver. Considerably more work will be needed to determine the efficiency of the radar system in GI tract localization. Our future research will concentrate on the application of a highly sensitive receiver and also deploying a MIMO system to improve the received signal as to achieve the millimeter range accuracy required for capsule endoscopy.

[11]

[12] [13] [14]

[15]

R EFERENCES [1] T.D.Than, G.Alici, H.Zhou, and W.Li, “A review of localization systems for robotic endoscopic capsules,” Biomedical Engineering, IEEE Transactions on, vol. 59, no. 9, pp. 2387 –2399, Sept. 2012. [2] C. Hu, M. Li, S. Song, W. Yang, R. Zhang, and M.-H. Meng, “A cubic 3-axis magnetic sensor array for wirelessly tracking magnet position and orientation,” Sensors Journal, IEEE, vol. 10, no. 5, pp. 903 –913, May 2010. [3] X.Wu, W.Hou, C.Peng, X.Zheng, X.Fang, and J.He, “Wearable magnetic locating and tracking system for mems medical capsule,” Sensors and Actuators A: Physical, vol. 141, no. 2, pp. 432 – 439, 2008. [4] S.Hashi, S.Yabukami, H.Kanetaka, K.Ishiyama, and K.I.Arai, “Numerical study on the improvement of detection accuracy for a wireless motion capture system,” Magnetics, IEEE Transactions on, vol. 45, no. 6, pp. 2736 –2739, june 2009. [5] H.Liu, H.Darabi, P.Banerjee, and J.Liu, “Survey of wireless indoor positioning techniques and systems,” Systems, Man, and Cybernetics, Part C: Applications and Reviews, IEEE Transactions on, vol. 37, no. 6, pp. 1067 –1080, Nov. 2007. [6] S.Gezici, “A survey on wireless position estimation,” Wirel. Pers. Commun., vol. 44, no. 3, pp. 263–282, Feb. 2008. [7] U.I.Khan, K.Pahlavan, and S.Makarov, “Comparison of toa and rss based techniques for rf localization inside human tissue,” in Engineering in Medicine and Biology Society,EMBC, 2011 Annual International Conference of the IEEE, 30 2011-Sept. 3 2011, pp. 5602 –5607. [8] P.Swar, Y.Ye, K.Ghaboosi, and K.Pahlavan, “On effect of transmit power variance on localization accuracy in wireless capsule endoscopy,” in Wireless Communications and Networking Conference (WCNC), 2012 IEEE, Apr. 2012, pp. 2119 –2123. [9] Y.Wang, R.Fu, Y.Ye, U.Khan, and K.Pahlavan, “Performance bounds for rf positioning of endoscopy camera capsules,” in Biomedical Wireless Technologies, Networks, and Sensing Systems (BioWireleSS), 2011 IEEE Topical Conference on, Jan. 2011, pp. 71 –74. [10] Y.Ye, U.Kahn, and K.Pahlavan, “Accuracy bounds for rss and toa based rf localization in capsule endoscopy,” in Engineering in Medicine and

[16] [17]

[18]

148

Biology Society,EMBC, 33rd Annual International Conference of IEEE, 2011. T.Shah, S. Aziz, and T.Vaithianathan, “Development of a tracking algorithm for an in-vivo rf capsule prototype,” in Electrical and Computer Engineering, 2006. ICECE ’06. International Conference on, Dec. 2006, pp. 173 –176. G. R. Curry, Radar System Performance Modeling, Second Edition, 2nd ed. Artech House Publishers, 2004. Radar cross-section. [Online]. Available: http://en.wikipedia.org/wiki/ Radar cross-section S.K.Kim, H.J.Kim, K.Y.Hur, S.H.Choi, C.W.Ahn, S.K.Lim, K.R.Kim, H.C.Lee, K.B.Huh, and B.S.Cha, “Visceral fat thickness measured by ultrasonography can estimate not only visceral obesity but also risks of cardiovascular and metabolic diseases,” The American Journal of Clinical Nutrition, vol. 79, no. 4, pp. 593–599, 2004. C.W.Kim and T.S.P.See, “Rf transmission power loss variation with abdominal tissues thicknesses for ingestible source,” in e-Health Networking Applications and Services (Healthcom), 2011 13th IEEE International Conference on, Jun 2011, pp. 282 –287. Y.Lee and K.Hwang, “Skin thickness of korean adults,” Surgical and Radiologic Anatomy, vol. 24, pp. 183–189, 2002. H.Kanehisa, M.Miyatani, K.Azuma, S.Kuno, and T.Fukunaga, “Influences of age and sex on abdominal muscle and subcutaneous fat thickness,” European Journal of Applied Physiology, vol. 91, pp. 534– 537, 2004. (2012, Aug.) Dielectric poperties of body tissue. [Online]. Available: http://niremf.ifac.cnr.it/tissprop/