Abstract
A recent reflector-backed antenna model is proposed in this paper for wireless capsule endoscopy localization. The antenna is designed to operate at the lowest 802.15.6 mandatory UWB (ultrawideband) channel, i.e., 4 GHz center frequency with 500 MHz bandwidth. The antenna achieves a good directivity and radiates well over the frequency band of interest. The proposed antenna was constructed within three successive steps. Initially, a planar omnidirectional antenna was designed of 3.15 dBi gain at 4 GHz. Since the antenna aims to operate as a receiving antenna, good directivity is preferred. Thus, an air-filled cavity was included backing the planar antenna to bolster the directivity toward the radiating element. The cavity-backed antenna has a measured gain of 6.4 dBi. The antenna was evaluated next to the homogenous and multilayer models. Then, the antenna design was optimized, by reducing its size, to a reflector-backed antenna structure reaching a maximum gain of 5.3 dBi, which is still promising for the regarded application. The body effect on the antenna matching was evaluated by means of multilayer and voxel models simulating the human body. This was followed by on-body measurements involving real subject. The depth of in-body propagation, from skin to small intestine, was studied using the multilayer and voxel models. Simulations were run using the CST Microwave Studio tool. While prototyping, free-space and on-body measurements took place at University of Oulu, Finland.
1. Introduction
In a world that is dominated by rapid change, mostly in novel technologies, wearable devices become important and unavoidable in life starting from home to hospitals. Nowadays, wearable antennas are regarded as priority gears for sensing, diagnosing, and treating illnesses in the medical field [1, 2]. An important medical application is wireless capsule endoscopy localization [3, 4]. This specific application aims to provide an accurate position of a swallowed capsule, which travels through the GI (gastrointestinal) tract from the esophagus to the small bowel organ [5]. In fact, this gastroendoscopic approach relies fundamentally on the disposition of a high-quality telemetry system. In other terms, a well-designed capsule [6, 7] along with an efficient directional receiving antenna is highly recommended for an efficient communication link budget. Furthermore, the setup of an accurate localization system [8] requires a careful receiving antenna design. The design of any antenna structure starts from the decision regarding the operational frequency. In this context, the UWB (ultrawideband) range has generated a large number of studies, from ISM [9] and MedRadio bands [10, 11] to low-UWB band (3.75–4.25 GHz) [12–14]. However, comprehensive design of an efficient receiving antenna for capsule localization is lagging. As the antenna is supposed to be held by the patient under control, the antenna should provide a directional radiation pattern to improve the overall link budget inside the human body. Related to the aforementioned reasons, exhaustive investigations have been conducted recently by the authors to emphasize the challenges of UWB directive antennas designed for the application under interest. As a result of these works, new antennas were designed and presented in detail in [15–18]. Moreover, the antenna behavior next to a human body was discussed in [19], and on-body and channel propagation features were studied in [20–22].
This paper revolves around the optimized new design of a recent receiving antenna operating at 4 GHz, center frequency of lowest UWB mandatory channel, and covering 3.75–4.25 GHz bandwidth defined in the IEEE 802.15.6 standard [23]. The proposed antenna is a continuity of the recently published antenna structure [23] and is characterized by good directivity and high efficiency over the frequency band of interest. The antenna design steps are described in detail over the paper. Initially, a planar monopole antenna was modelled to which a cavity was included later. Then, the resulting structure was optimized, by reducing its size, to obtain the proposed reflector-backed antenna design answering to the required needs. The antenna structures were measured in free-space and in different on-body scenarios. By considering the external antennas available in literature and serving of on-body antennas for wireless capsule endoscopy systems, the proposed reflector-backed antenna provides a small-sized structure with good directivity normal to the surface body compared to available external antennas found in [12–14].
The paper is organized as follows: the antenna structure and analysis of the planar, cavity-backed, and reflector-backed antenna structures are presented and discussed in Sections 2 and 3, respectively. Finally, conclusions and future work perspectives are presented in Section 4.
2. Planar and Cavity-Backed Antenna Structure
The planar antenna was recently designed and originally published in [24]. The planar monopole antenna has an overall size of 36 × 36 × 1.6 , as depicted in Figure 1(a), and its detailed parameters with tuning trials are described in [24]. The optimized antenna parameters are collected in Table 1. The planar monopole antenna has a reflection coefficient below −10 dB over 3.55–4.69 GHz bandwidth, in free-space, as presented in Figure 2. The input impedance plotted in Figure 3 shows that the planar antenna had originally 48.25 + j0.98 Ω at 4 GHz. It seems that the antenna prototype will not create any impedance mismatching with a 50 Ω SMA connector for practical use.
(a)
(b)
According to results presented in Figure 4, the modelled planar antenna operates well in free-space. This is revealed by the simulated total efficiency between −0.95 dB and −0.15 dB range. The planar monopole antenna has basically an omnidirectional radiation pattern with maximum directivities of 3.28 dBi, 3.15 dBi, and 3.08 dBi at 3.75 GHz, 4 GHz, and 4.25 GHz, respectively, as mentioned in [24]. Maximum gain values in free-space of the planar antenna structure are summarized in Table 2. Results reveal that the main radiation is in Theta = 90° direction, in free-space.
The designed antenna is targeted to represent a receiving antenna to enable the wireless communication link with the wireless capsule endoscope that will be situated in the small-intestine area of the patient. Hence, the receiving antenna is preferred to be directive enough providing a moderated good gain with main directivity normal to the body surface. For this aim, the planar structure was corrected to a new antenna structure. In this context, the cavity-backed approach was applied to the original planar antenna recently published in [24]. The antenna was embedded within the air-filled cavity box in simulations. It was concluded from the tuning cavity parameters, discussed in details in [24], that the ideal antenna position inside the cavity is in the center with X = 30 mm, Y1 = 35 mm, Y2 = 25 mm, and Z1 = Z2 = 25 mm as depicted in Figure 1(b). To remind the antenna features available in [24], a comparison between the planar and cavity-backed configurations is delivered and discussed based on the reflection coefficient, the input impedance, and the efficiency, as can be seen in Figures 2–4, respectively. Results show that the cavity-backed antenna covers the required bandwidth of 3.75–4.25 GHz. Furthermore, it has good radiation properties over the covered frequency band. Comparison of 3D radiation pattern, in free-space, of the planar and cavity-backed structures is illustrated in Figure 5. From the figure, the gain is clearly improved from 3 dB to 8 dB by introducing the cavity.
(a)
(b)
The manufactured cavity-backed antenna, seen from top and side views, is illustrated in Figures 6(a) and 6(b), respectively, while Figure 6(c) gives the focus on the contact feeding point of the antenna. From the figure, it is clearly seen that the antenna prototype was fed differently from the simulated model. The reason for this feeding method was the easy implementation. However, this approach seemed responsible for the resulting mismatch in the impedance of the antenna seen in Figure 7. Regarding this issue and according to the results of this first prototype trial, authors came to these conclusions: the feeding had to be done differently as in simulations. Besides, grounding the monopole to the cavity may cause a mismatch in practice appearing as “galvanic effect”. For the aforementioned reasons, it was agreed to repeat the matching result measurements with a modified and improved prototype. At this regard, the new prototype incited a small size and easy manipulation for on-body application. To this end, an improved and optimized antenna structure was built, later named “reflector-backed structure.” This will be discussed in the following section.
(a)
(b)
(c)
The measured realized gain of the cavity-backed antenna at 3.75 GHz, 4 GHz, and 4.25 GHz is depicted in Figure 8. Results show a good correlation with simulated gains, which are grouped in Table 3. Maximum measured gains are 6.61 dBi, 6.39 dBi, and 4.25 dBi at 3.75, 4, and 4.25 GHz, respectively. Therefore, the antenna is considered a good candidate for directive receiving antennas.
(a)
(b)
(c)
The measured total efficiency of the proposed cavity-backed antenna in free-space is given in Figure 9. The measured total efficiency is between −2.45 dB and −0.85 dB over the frequency band of interest. At 4 GHz, the total efficiency is −1.2 dB, corresponding to 75% in linear scale. The total efficiency is radiated power integrated over a sphere related to power delivered to the antenna port. It is evident that this cavity-backed configuration meets the requirements needed for enabling good in-body communication link. However, its large size presents a drawback for practical on-body operational use. Therefore, reducing the antenna size is highly required by maintaining the good antenna performances. This will be discussed more in detail in the following section.
3. Reflector-Backed Structure and Analysis
3.1. Antenna Structure
In this section, the cavity-backed antenna structure was revised to meet practical on-body application. Some changes applied on the cavity box were brought about, as demonstrated in Figures 10(a) and 10(b). First, the model was simulated in CST with the aim to reduce the size of the cavity. The proposed reflector-backed antenna structure was fed in simulation by coaxial cable to approach the measured prototype, as seen in Figure 10(a). The metallic reflector has a size of 56 mm length and 54 mm width. The antenna parameter values are X = 10 mm, Y = 20 mm, and Z = 20 mm. The reflector is selected as the suitable approach to minimize the backwards radiation that would create interference with other medical devices operating at the same 3.75–4.25 GHz range, such as 5G devices [25]. Later, the antenna prototype was fabricated by supporting the planar antenna with a Rohacell piece [26], as illustrated in Figure 10(c). The prototyped antenna has an input impedance of 51.22 − j15.83 Ω, which is adapted to real measurements involving 50 Ω SMA connector, as seen in Figure 10(a). The reflection coefficient of the antenna was measured as a first attempt (Measured 1), as mentioned in Figure 11. Then, a tuning was made in practice at the feeding point of the antenna resulting in plot (Measured 2). Both measured bandwidths are 3.38–6.00 GHz below −10 dB, which proves the good operation of the proposed antenna at the requested low-UWB range of 3.75–4.25 GHz. The prototype shows a total efficiency in −2.0 and −1.2 dB range, as presented in Figure 12, which demonstrates the good radiation over the considered frequency band 3.75–4.25 GHz. Figure 13 illustrates the measured radiation pattern of the reflector-backed antenna compared to simulations, and results prove a good agreement. The detailed comparison of the gain values are grouped in Table 4. It results that the prototype has a good directivity with a maximum gain in Phi = 0° direction of 4.2 dBi, 5.29 dBi, and 5.82 dBi at 3.75 GHz, 4 GHz, and 4.25 GHz, respectively. This confirms the good application of the reflector-backed antenna, in free-space.
(a)
(b)
(c)
(a)
(b)
(c)
3.2. On-Body Simulations
The free-space investigations were completed by on-body studies by using the multilayer model described in Table 5 [27–30] and Laura 2018 voxel model from CST Library, as depicted in Figures 14 and 15(a), respectively. The proposed reflector-backed antenna is aimed to receive signals transmitted by an endoscope capsule travelling through the small-intestine (SI) tract. Thus, the wireless telemetry link could be assessed by using an external receiving antenna. As the designed antenna will perform near the human body, it is greatly appreciated to predict the antenna interaction with body proximity. By assuming the fat layer (visceral and abdominal) to be the most likely tissue differentiating persons, we consider two person cases called “PA” and “PE,” with the latter having three times the fat thickness as the former, as described in Table 5. When using voxel model-based simulations, the antenna was placed at the navel level, as presented in Figure 15(b), and only the abdomen section of the voxel model was evaluated to save the computation time. Thus, the boundary settings were set to “Open” at the top and bottom of the voxel cut section in order to avoid any radiation overflow at these boundary conditions. As can be seen from Figure 15(b), unlike the multilayer model, the voxel model presents a realistic model close to the real anatomical body construction since it includes other tissues such as bone, blood, colon and etc. Besides, the voxel shape follows the real body morphology including the organ emplacement inside. The antenna is distanced from the skin layer for both multilayer and voxel models by the parameter d, as mentioned in Figures 14 and 15(b).
(a)
(b)
By referring to antennas for on-body applications, the analyses of signal absorption by lossy tissues are inevitable [31]. At this regard, SAR (specific absorption rate) estimations were calculated in this paper at first, by considering PA model case(Figure 16). For d = 4 mm, the maximum SAR value over 10 g is 0.084 W/kg at 4 GHz, which is below the SAR limit according to the IEEE C95.3 standard [26]. This value was calculated by CST, using the averaging method IEEE C95.3, with a low input power of 0.00316 W. By the distance increase up to 30 mm, the peak value decreased to reach 0.005 W/kg. Assuming this distance value, the reflection coefficient of the proposed antenna for the several tissue combinations PA and PE are tested, as presented and compared to voxel model results in Figure 17(a). At this fixed distance, a frequency bandwidth below −10 dB of 3.45–5.49 GHz is covered. The undesired resonant frequency occurs at 2.45 GHz achieving a bandwidth of 2.36–2.55 GHz, which is within 2.4 GHz ISM (Industrial, Scientific, and Medical) band; however, its presence can be neglected in this study. Compared to the voxel model result, close bandwidth is remarked about 2.36–5.41 GHz with the appearance of the unexpected 2.4 GHz resonant frequency. By decreasing the distance by 10 mm, as plotted in Figure 17(b), the bandwidth was narrowed and consequently it covered 4.00–5.32 GHz. By further decreasing the distance to 12 mm and 10 mm, the impedance matching was lagging and the undesired resonant frequency was shifted to 2.68 GHz with a maximum return loss of −40.26 dB. At a minimal separation distance of 2 mm, the mismatch was remained with the disappearance of the undesired resonant frequency by using both multilayer and voxel models.
(a)
(b)
(a)
(b)
The input impedance results of the reflector-backed antenna next to multilayer/voxel models are studied, and maximum gain values are collected in Table 6. Overall, the antenna has a capacitive behavior regardless of the distance, the human model, and the study case. Besides, according to the table, at d = 30 mm, the real part of the input impedance is in 51.76−55.31 Ω range for the multilayer model, which is close to the free-space case 51 Ω, against 41.92 − j26.41 Ω for the voxel model case. With the distance decreased down to 20 mm, the input impedance declined to 29.31 Ω using the multilayer model. This value continued to decrease to reach 12.83 Ω with a distance decrease by 2 mm. However, as the distance got smaller about 2 mm, the real part of the input impedance notified a sharp increase up to 126.59 Ω and 92.05 Ω for PA and PE person case, respectively. This input impedance jump explains the impedance mismatch remarked in Figure 17(b) [32]. On the other hand, from the same comparison table, at d = 30 mm, the realized gain is between 5 dB and 6 dB for the studied multilayer and voxel models. By decreasing the distance, the realized gain decreased to 0.34 dB at 10 mm. Further approximation to the human model about 2 mm produces high absorption of the radiations by the several/multiple layer tissues and consequently leads to a poor radiation around −8 dB gain. By considering this 2 mm distance, the realized gain does not change much by altering the tissue thicknesses like for PA and PE. Besides, 3D insight of the gain using the human models can be found in Figure 18. Consequently, a good agreement is seen between multilayer and voxel model results. From these findings, one can predict that fat amount (visceral and/or abdominal) will not affect the concluded results for real candidates, which is a very important issue to consider in real measurements.
(a)
(b)
(c)
(d)
(e)
(f)
The on-body simulations previously conducted using multilayer and voxel models remain theoretical. To approach practical application, real measurements are of vital importance. In this context, the antenna was attached to a real candidate by means of elastic bands and the measurements were realized in an office room at the University of Oulu, Finland. Initially, the antenna was directly held by the candidate at the navel in the abdomen area, in direct contact with thin clothes, as illustrated in Figure 19(a). Secondly, the antenna was placed at the navel position to evaluate the impact of the separation distance on the reflection coefficient, by filling the air space between the antenna and clothes with Rohacell pieces [33] of different thicknesses. Various distances were applied of 2 mm, 4 mm, 6 mm, 8 mm, 10 mm, 20 mm, and 30 mm by means of Rohacell pieces, as figured in Figure 19(b). Later, the antenna was moved to various on-body positions with 0 mm separation distance, as described in Figure 20. The chosen on-body positions under investigation are A (navel), B (left flank), C (left from the navel by 15 cm), D (right flank), E (right from the navel by 15 cm), and F (up from the navel by 10 cm).
(a)
(b)
For d = 0 mm, the reflection coefficient was widened from the original free-space case, as plotted in Figure 21, and the reported results are collected in Table 7. It seems that the antenna has a large bandwidth behavior in direct contact to the body through thin clothes. Interestingly, this bandwidth enhancement was not predicted by previously discussed multilayer/voxel models. It is important to remind that the voxel model does not consider the clothes' impact on the antenna performance. Besides, it is reported in the literature the significant effect of the real human body to further increase the covered bandwidth, which is mainly related to the input impedance change of the antenna close to real human body [32, 34, 35]. However, Figure 22 proves how the antenna distance d alters the impedance matching significantly. One can see that the required impedance bandwidth delimited by −10 dB of 3.75–4.25 GHz is covered for d = 6 mm and d = 8 mm (3.64–4.54 GHz) and for d = 10 mm (3.64–4.26 GHz). Beyond these three distances, the bandwidth of interest is lost. These on-body measurements are close to the discussed results in Figure 17. The reasons behind these findings will be discussed in detail, in a coming paper, using multiple voxel models and by means of other antennas working at the same frequency range.
3.3. Simulated In-Body Investigation
In this section, the antenna performance for in-body communications is analyzed. First of all, two antennas were placed at 2 mm distance from the multilayer model PF, corresponding to the placement shown in Figure 23(a). This aimed to expect the path loss of a signal travelling the skin towards the SI layer. The resulting transmission coefficient at 4 GHz is estimated averaging 62 dB, as presented in Figure 23(b). To confirm this estimation, multiple aligned E-field and H-field probes were set at various tissue interfaces with the purpose of estimating the Poynting vector at these locations to predict the power consumption by each tissue. The probes are aligned according to the middle of the antenna itself, as described in Figure 24. Table 8 explains the probes names. The power loss is estimated using Poynting vector defined aswhere is the cross product and “” is the conjugate value.
(a)
(b)
(a)
(b)
E and H are the electric and magnetic complex vectors provided by the probes, respectively.
Considering the multilayer model combination for PF case, the resulting calculated power loss by formula (1) is 65.4 dB from the skin to the SI layers, starting from 29.3 dB to end with −36.1 dB at the end surface of the SI layer. This total path loss matches well with the reported value of 62 dB in Figure 23(b). Therefore, the following investigations can rely on Poynting-based analyses to average the power consumed in the human body.
Considering the reported results in Table 9, at the probes A, B, and C, the power average is around 19.8 dB. This value is decreased to 15.3 dB at a point located just before the skin layer surface (0.1 mm). This is evident since the power is reverse proportional to the travelled distance. At the skin surface, the power has almost fallen down roughly by 4 dB. Accordingly, as the signal goes deep within the multilayer model, for instance at fat1 layer interface, the power further decreases for all the model combinations. This power value continues to decrease by passing by muscle and fat2 layers. At SI level, the wall layer is faced at first, which is responsible for around 1 dB decrease in received power. The power consumption by SI content layer can be estimated about 21 dB, regardless of the considered multilayer model. It is concluded that the average power loss from the skin to the SI layers is in the range of 40–49 dB for PA and PE model cases.
With regard to the impact of the separation distance on the path loss, deep investigation was done considering the presence of the model PA with distinct distances d of 20 mm, 12 mm, 10 mm, and 4 mm. The evaluated path loss values were grouped in Table 10. At the skin layer, by decreasing the distance from 20 mm to 4 mm, the path loss increased by 13.7 dB. Moreover, the reported path loss values are 3 dB, 6 dB, 3 dB, 4 dB, 0.7 dB, and 14 dB corresponding to the mentioned thicknesses in Table 5 for skin, fat1, muscle, fat2, SI wall, and SI content layers, respectively. It is concluded that the resulting power average increases with the distance decrease, which goes back to deep signal penetration through the tissues because of the high antenna directivity. These results were followed by in-body investigations conducted using Laura voxel model, by placing few E-field and H-field probes, as marked by the green circles mentioned in Figure 24(b) within the distinguished tissues by their names. An initial probe was placed just before skin tissue (JBS), followed by a probe placed just on the skin tissue in a line direction. Then, the start of the fat tissue was chosen for the following probe position. To reach the SI tissue, two probes were placed at the start and the end of the SI tissue to approximate the average power consumed by the lossy tissues, from skin to SI organs. The reported path values are grouped in Table 11 for d = 4 mm and d = 30 mm case. It is clearly seen that the total loss reported using voxel model is minor than the multilayer case, which is in 20–30 dB range. This proves that the EM waves travelling through the lossy human model choose an easier and rapid propagation path to reach the SI organ. In this direction, an upcoming paper will be prepared by presenting an exhaustive simulation and measurement investigation for channel model and in-body propagation using the same antenna presented in this section.
4. Conclusion and Perspectives
As the world continues changing in a variety of ways, the necessity for new devices will significantly increase in many fields, especially in medical applications. On this regard, this paper comes to propose a new external reflector-backed antenna design for wireless capsule endoscopy systems. The proposed antenna has good directivity and operates at 3.75–4.25 GHz bandwidth with 4 GHz center frequency, in accordance with the IEEE 802.15.6 standard. The proposed reflector-backed antenna was assessed in free-space and next to the human body. The antenna bandwidth behavior next to the human body was evaluated by means of multilayer and voxel models and showing a good result agreement. Moreover, on-body measurements were conducted as well and showing the enlargement of the free-space bandwidth to 2.86–9.14 GHz in direct contact to the body. Besides, the proposed antenna is designed basically to establish a good communication link with a capsule travelling through the SI organ. Therefore, in-body propagation depth is evaluated by estimating Poynting vector inside the considered human models. It is concluded that the path loss from skin to SI is approximated to be in 40–49 dB range for the planar multilayer case. However, this value is expected to be less than 30 dB by considering a realistic human model, in this study Laura voxel model. These studies will be completed by channel model and on-body propagation investigations validated by measurements as a next step.
Data Availability
The simulated results are included as whole in the paper. The experimental measurement data used to support the findings of this study have not been made available because the paper describes in a lot of detail the experimental scenario of how the data were collected. Interested parties can replicate the same scenario to get their own data.
Conflicts of Interest
The authors declare that there are no conflicts of interest regarding the publication of this paper.
Acknowledgments
The authors would like to thank all the staff at the Microelectronics Research Unit and Centre for Wireless Communications teams for the help on antenna design and measurements. The work was partly supported by the Academy of Finland project 6Genesis Flagship (grant no. 318927).