Proposal of a Microstrip Patch Antenna Array for the Osseus: An Equipment for the Aid of Diagnosis of Osteometabolic Diseases


 BackgroundApplications on electromagnetic waves in the field of biotelemetry have increased in the latest years, being used to prevent, diagnose, and treatment of several diseases. In this context, biotelemetry allows minimally invasive monitoring of the physiologic, improving the comfort and patient care and significantly reducing hospital costs. Aiming to assist the mineral bone density classification, through a radio frequency signal (RF), for a later diagnosis of osteoporosis, Osseus was proposed in 2018. This equipment is a combination of the application of techniques and concepts of several areas such as software, electrical, electronic, computational, and biomedical engineering, developed at a low cost, with easy access to the population, and non-invasive. However, when placed on evaluation, potential improvements were identified to increase the stability of Osseus operation.ResultsIt is proposed the implementation of improvements in the antennas used by Osseus, aiming its miniaturization, improvement in the reception of the RF signal, and better stability of the equipment's operation. ConclusionTwo antennas were then built, one of which was used as a project for the second, which is an array. The array showed significant improvements in the parameters relevant to the application, being a candidate to replace the antennas currently in use at Osseus.


Background
Applications of electromagnetic waves in the field of biotelemetry have increased in the latest years. Currently, these waves are used in the prevention, diagnosis, and treatment of a variety of diseases and play an important role in analysis and imaging, thermal therapies, administration of drugs and sensors 1,2 . In this context, biotelemetry allows minimally invasive monitoring of physiological parameters, improving patient comfort and care and significantly reducing hospital costs 3,4 .
The monitoring of physiological parameters can lead to the diagnosis and, consequently, to the early treatment of diseases, being essential for the implementation of preventive measures. Thus, in the case of osteoporosis, the risk of bone fractures can be reduced 5 .
Annually, millions of bone fractures resulting from osteoporosis happen worldwide. The most common bones with fracture incidence include those of the spine, forearm, proximal humerus, and hip6. Hip fractures are those with the highest morbidity and mortality and cause the highest direct costs to the health services, increasing exponentially when they affect the elderly population. It is estimated that in 2050 the number of hip fractures in men and women, within the range of 50 to 64 years in Latin America, will increase by 400%. In Brazil, ten million people, about one in seventeen, suffer from osteoporosis 6 .
To reduce costs with procedures for the treatment of bone fractures caused by osteoporosis, it is essential to implement preventive measures in view of the growing number of elderly people in Brazil and the worsening of the calamity that this disease has caused, hindering early detection and subsequent appropriate treatment 7 . In Brazil, the population has not been given adequate access to early diagnosis of the disease. One of the factors that lead to this is the complexity and high cost of the distribution of equipment throughout the Brazilian territory.
Based on what was exposed, in 2018, Osseus was proposed. An equipment capable of assisting in the classification of bone mineral density, by means of a radio frequency (RF) signal that would lead to a later diagnosis of osteoporosis. This instrument is a combination of the application of techniques and concepts from different areas such as software, electrical, electronic, computational, and biomedical engineering, developed at a low cost, with easy access to the population, and non-invasive 6 .
This procedure is done through the reading of a 2.45 GHz RF signal, which passes through the medial phalanx of a patient's middle finger, and, after the signal is received, it is processed by a system that contains an embedded computational intelligence technique. Analyzing the intensity of the received RF signal, the system can provide a prognosis informing the necessity to refer or not the patient to perform an exam of bone densitometry.
Upon being put under evaluation, potential improvements were identified to increase the stability of the Osseus functioning. The intensity of the RF signal varied in the same patient for several measures, whether due to the positioning of the patient's finger between the antennas, or due to random factors. Another improvement of interest was the manufacturing of equipment with smaller dimensions, in order to increase its portability.
Thus, the physical configuration of the equipment was analyzed, as well as the antennas in use. Throughout the research, it was concluded that it was necessary for the pair of antennas used in the transmission/reception to be well-calibrated and to have very similar operating characteristics in terms of signal radiation.
In this scenario, microstrip antennas were a good alternative, as they are known for their miniaturization capacity and simplicity, and low manufacturing cost, despite their narrow bandwidth. However, bandwidth is not a relevant parameter for this application, as no information is being transmitted, but only an RF signal. Other parameters, such as impedance matching and directivity, are essential for the proper functioning of the device.
This article presents state-of-the-art microstrip antennas used in biotelemetry applications. Then we present the main improvements implemented in Osseus, the design of the proposed antennas to be used in the equipment, as well as the main results achieved.

Latest researches
The use of antennas in biomedical applications is vast, ranging from sensoring to disease diagnosis. In the literature, there are countless researches with proposals for microstrip antennas, in biomedical applications. In this section, we try to identify the main publications in the last 4 years. Experimental results of the reflection coefficient and the axial ratio were presented. The fractional bandwidth and axial ratio were 5.45% and 5.69%, respectively.
Another circularly polarized antenna was proposed by Z-J. Yang 13  All of these researches have antennas with characteristics that do not match the requirements of Osseus, mainly because we need a directive antenna, with a Half Power Beam Width (HPBW) lower than 70º and with a gain greater than 2.5 dBi. In addition, the antenna has dimension restrictions due to the size of the equipment. Thus, in this research we propose an array of two-element microstrip antennas, to reduce the HPBW and provide a satisfactory gain, as well as occupying a reduced space.

Results
After the designing, the microstrip antenna and the microstrip antenna array, simulations of the main parameters of these antennas were performed in the commercial software ANSYS HFSS, as well as simulations for the Yagi antenna too. The purpose of these simulations is to compare the performance of the array with the other antennas.

Simulated results
The first parameter analyzed is the reflection coefficient. Figure   Another important parameter that must be analyzed is the radiation diagram. This parameter was simulated for co-polarization and cross-polarization, which are illustrated in Figure 2. Figure 2(a) illustrates the simulated radiation diagrams for co-polarization.
The Yagi antenna has the highest directivity, around 5.7 dBi, while the patch antenna has the lowest directivity, reaching around 2.5 dBi while the antenna array has 4.7 dBi.
However, we can see that the antenna array has a better HPBW of 70º, while the patch antenna has 120º and the Yagi has 95º. Regarding the cross-polarization diagram, illustrated in Figure 2(b), we can notice that the Yagi antenna does not present polarization purity, with a difference of only 16 dB, comparing its co-polarization with the cross-polarization. As for the antenna array, the cross-polarization level reached more than 40 dB below the co-polarization level, which shows a high polarization purity. The patch antenna also features high polarization purity, with a cross-polarization level of more than 35 dB below the co-polarization level. The next parameter to be analyzed is the Smith chart, which shows how the antenna impedance behaves as a function of the frequency. Figure 3 illustrates the Smith chart of the three antennas. For the Yagi antenna, the Smith chart is illustrated in Figure   3(a). It can be seen that the resonance occurs at 2.29 GHz and the normalized impedance was 1.065 + j 0.025 Ω. For the patch antenna, the resonance frequency occurs at approximately 2.45 GHz, but it can be seen in Figure 3  Based on the data listed in Table 1, we can see that the antenna array has a lower performance than the antenna currently in use in Osseus, as the patch antenna too.
However, the proposed array has smaller dimensions than the Yagi antenna used. While the array has dimensions of 3.5 cm × 7.0 cm, the Yagi antenna has dimensions of 6.6 cm × 7.0 cm. Both have equivalent performance in terms of directivity and front-to-back ratio. The antenna array has a lower bandwidth, but that is not relevant for the application.
In addition, the array has a lower HPBW, which can promote better results and an improvement in equipment performance, since stability is an issue. Finally, the way in which the coaxial cable is connected to the antenna array eliminates the instabilities found in the Yagi, which can prove to be a significant improvement. Another fact to be noted is that the Yagi antenna is positioned along the longest axis of the Osseus equipment, while the antenna array will be placed perpendicular to that axis. In this way, the antennas will operate in the distant field region. This will prevent variations in the reading of the signal intensity from occurring.
From the decision of the antenna that would be used in the equipment, some prototypes were manufactured for experimental validation and to analyze the improvements in the equipment. Figure 4 illustrates the Yagi antenna currently in use compared in scale to the built antenna array.

Measured results
The measurement setup was set up with various equipment. The Agilent two-port network vector analyzer, model E5071C, with a measurement range from 300 KHz to 14 GHz was used. Cables appropriate for this frequency range were used. In addition, for gain measurement, we used standard horn antennas model SAS-571 (double ridge guide horn antennas). We tried to measure all possible parameters, with the available setup. The parameters measured were: the modulus of the reflection coefficient (|S11|), gain of the array, and the Smith chart.
The first parameter measured was S11, which allows us to analyze the resonance frequency of the antenna and its bandwidth. We measured this parameter for four manufactured arrays. Figure 5 illustrates this parameter. We can see that the antennas responded in a similar way. The antenna arrays that presented the best impedance matching were installed in Osseus. Figure 6 illustrates the frequency response of S11 only for these two arrays. We can see that the results are very similar.  In Figure 7, we compared the simulated S11 with the measured response for antenna array 2. We can see that there was a small difference between the results. This is perhaps due to the manufacturing process, which was not done with a milling machine. This difference is not a problem, as Osseus has a voltagecontrolled oscillator, whose oscillation frequency can be regulated between 2.25 and 2.65 GHz. Figure 7 -Comparison between the simulated and measured S11 for antenna array 2. Figure 8 illustrates the measured Smith chart. We can see a reactance greater than that obtained in the simulation, which was 7 Ω, while the measured value was 12 Ω. This difference represented in the reactance caused a reduction in the resonance frequency.

Discussion
One of the main improvements that needed to be made in Osseus was to reduce possible variations in the reading of the signal intensity. It is important that the signal-tonoise ratio (SNR) received remains constant, or that it undergoes minimal changes. This will allow computational intelligence tools to achieve a more accurate classification when doing the screening. It is observed that the previous version of the Osseus, with the Yagi-Uda antennas, presented considerable SNR variations, depending on the position of the finger inserted in the equipment. To perform the measurements, two positions of the finger were considered, as shown in Figure 10. The first measurement is made with the finger inserted in the position considered correct (Figure 10(a)) and the second measurement is made with the finger inserted in the position considered rotated ( Figure   10(b)). This will cause the RF signal to reach a greater or lesser finger surface. The desired result is that the SNR does not vary, even with the finger inserted in a rotated manner, allowing a more stable reading. The received signal is obtained in a form of a level of direct current (DC) with a certain level of background noise coupled to it. The DC signal has a maximum level when there is no obstacle between the antennas and this value gets lower when the finger is inserted into the equipment. Also, a shield made with absorbers was inserted inside the box, so that the noise level remains constant, regardless of the reading, with or without the finger. Figure 11(a) illustrates the DC level received for a direct reading (without the finger insertion), for the case of the first version of Osseus, with the Yagi-Uda antennas.
In Figures 11(b) and (c) we can see the readings with the finger insertion in the correct and rotated position, respectively. It can be seen that there is a considerable level of noise, in addition to noticing a variation in the level of the received signal, for the different positions of the finger. Without the finger, the received signal is about 800 mV. For the reading with the finger in the correct position, the received signal is 620 mV, while for the reading with the finger in the rotated position, the received signal is 520 mV. In the frequency domain, we can verify whether the noise varies or not. Thus, the same measurements were made and the results can be seen in Figure 12. The noise level for the three situations considered above was -61.4 dB, with no variation for the situations with or without the finger. Based on the measurements made, it was possible to obtain the signal-to-noise ratio for all situations considered, for both versions of the equipment. Table 2 illustrates the SNR values obtained for each situation considered. We can observe an increase of more than 8 dB in the SNR, in all situations, which represents a very significant improvement.

Conclusions
The antennas currently in use in Osseus were characterized, to find out if they were the most appropriate for the equipment. The main parameters of the antenna were simulated in the HFSS software. It was observed that there is no uniformity of these parameters within the entire operating range. The antenna showed a satisfactory gain in the resonance frequency, but this gain suffered a great variation throughout the operating range. In addition, the HPBW of the antenna proved to be high. The measurements showed that the prototypes of the antenna arrays constructed showed similar performances, regarding the measured parameters. We selected the two best arrays to be installed in Osseus With the insertion of new antennas and a reduction in the size of the Osseus box, significant improvements in the intensity of the received signal, in the noise level, and in the signal-to-noise ratio were observed. SNR increased by more than 8dB. These improvements will allow for more accurate classification of the equipment.

Methods
The first part of the research aimed to identify possible improvements in Osseus.
Thus, characterizing the antennas used previously was an important action to start up this task. The antenna used a Yagi type, which is a directive antenna, with a gain of approximately 4 dBi. The input impedance of the antenna is 50 Ω, which allows easy matching for most of the feed networks.
The Yagi antenna is illustrated in Figure 15. It has a 50 Ω single-ended feed input and this transmission line is converted into a two-wire feed through the conductive path. One of the main problems of this antenna is the way that the coaxial cable is connected to it. Figure 16 illustrates this form of connection. It can be seen that a minimum error in welding can lead to an impedance mismatch changing the way it operates. The ground mesh must be welded on the reflector part and the central conductor of the coaxial cable must be welded on the dipole. Another problem refers to the region in which the electromagnetic field of this application is located. This type of antenna must operate in the distant field region, that is, the two antennas must be separated by a distance of 2R, where R is calculated as: In which D is the largest linear dimension of the antenna and λ is the wavelength of the antenna's resonance frequency.
This field regions are illustrated in Figure 17. to a volume restriction of the equipment, the antennas were separated by less than 30 mm, which would make them operate in the near field region, which could lead to erroneous readings of the signal intensity.
Finally, we must have an antenna whose HPBW is the smallest possible, for this application, aiming to concentrate the RF signal as much as possible. It is worth mentioning that the total volume of the equipment limits this factor, as it does not allow us to have highly directive arrays, with many elements. The analyzed Yagi antenna has a 95º HPBW.
Thus, we conclude that the Yagi antenna was not suitable for use in Osseus, and another antenna, or an array of antennas, had to be proposed to meet all the criteria listed in this section.

Project of the antenna and antenna array
From what was exposed in the previous section, we listed some parameters of the antenna in use to be improved. The antenna needs to be as small as possible to meet the portable device's final volume requirements. The antenna feed must be done in a more consistent way, via SMA connector soldered to it. The antenna needs an improvement in the gain, directivity, and, consequently, in the reduction of the HPBW. To guarantee certain radiation parameters, the antenna needs good impedance matching, along with a consistent way to input the signal of 2.45 GHz.
Based on the issues that need to be improved, a microstrip patch antenna and a patch antenna array have been proposed. As an initial proposal, we will start with a simple concept of a planar microstrip antenna with rectangular patch geometry. Unlike the initial antenna, this antenna has a ground plane completely filled over the entire length of the dielectric. The chosen feed was via a microstrip line and the impedance matching with the input is made by inserting a line gap in the rectangular patch (Inset Fed). An illustration of this model can be seen in Figure 18. For the design of this type of antenna, the classic approach presented by S. A.
Kumar and T. Shanmuganantham 15 was used, which systematically describes how to design this kind of geometry. It is important to choose the appropriate dielectric material.
In this approach, it is evident that the physical dimensions of the antenna can be reduced if we use a dielectric material with high relative electrical permittivity. Thus, the ROGERS RO3010 was chosen, with ɛr = 10.2 and thickness h = 1.27 mm. The increase in ɛr tends to reduce the bandwidth, so we choose the largest thickness available from this manufacturer, compensating for this reduction 15 .
Initially, the dimensions of the rectangular patch 15 are determined as: In which, Besides that, c is the speed of light in the vacuum, fr is the desired resonance frequency, εeff is the effective permittivity of the dielectric, and ∆L is the extended electrical length of the rectangular patch, due to the fringe effect 16 .
The next step is to design the gap in the patch, which will be responsible for the impedance matching for the antenna with its microstrip feed line. To do this, we need to calculate the patch impedance: where G1 is the approximate conductance of one of the patch slots, Rin is the approximate resistance of the patch, Rin (y = y0) is the resistance of the patch after the insertion of the gaps and y0 is the length of the gap.
Thus, using the expressions above, the physical dimensions for the antenna are listed in Table 3. The proposed array is illustrated in Figure 19. In this case, the impedance matching will not be done by an inset fed, but by a network feed, whose physical dimensions are determined from the desired impedance values. The elements of the array are identical to the rectangular patch antenna. Figure 19 -Antenna array and its proposed network feed.
The literature shows us that an array has as its basic characteristic an improvement in directivity and, consequently, the maximum gain along with the reduction of the HPBW, which are the parameters that we want to improve. Therefore, the design of the parallel array comes down to create a network feed that injects the signal in an equivalent way for two identical patches. This network feed must have a good impedance matching between each of its steps, to ensure that the injected signal reaches the radiator elements with the least possible loss 17,18 .
Initially, it is necessary to design a power divider. There are many different models of power dividers, but the T-type junction was chosen for ease of design and manufacture. This junction can be seen in Figure 20 and will serve to equally divide the power of the injected signal. In the T-junction, the impedance of each line determines its width, and the "V" shape with dimension "a" is necessary to minimize the parasitic reactances that arise due to the discontinuity of this junction. Figure 20 -T-type junction.
In Visser 17 , it is recommended for "a" to be 1.8W1. Whenever there is a discontinuity the same rule will be applied. It is important to note that if we want to divide the signal intensity equally for both antennas, we need the junction lines to have twice the impedance of the line that receives the power. We can verify this through the following expressions: In which: P1 is the power delivered to the array; P2 is the power delivered to one of the elements of the array; P3 is the power delivered to the other element of the array; K is a factor that determines the proportion how the power will be divided (K=1 for equal division of power); Z0 is the characteristic impedance of the feed line; Z2 is the impedance of one of the sides of the T-junction; Z3 is the impedance of the other side of the T-junction.
As the lines that follow from the T-junction have an impedance different than 50 Ω, it is necessary to use an impedance transformer so that the network feed is matched with the antennas that we initially designed. The transformer used will be the one-quarter wavelength transformer. Its width is determined from the desired impedance, that is 50 Ω. Figure 21 illustrates the insertion of the λ/4 transformer. To calculate the dimensions of the transformer, we consider the length λ/4 and the width obtained from (9), knowing that the impedance of the transformer must be: = √ 01 02 (17) In which Z01 is the impedance of the line of the T-junction and Z02 is the impedance of the antenna.
In this way, we completed the design of the array's network feed by obtaining all of its physical dimensions. Lesser adjustments will need to be made, due to the fact that some of the expressions are approximate and, mainly, due to the fact that the high permissiveness we use to reduce the dimensions of the antenna causes a greater fringe effect. Table 4 lists the dimensions of the network feed for this antenna array.