On the Experimental Investigation of Bone Fracture Recovery Process Using an Ultra-Wideband Planar Monopole Antenna

In this paper, a noninvasive bone fracture monitoring system is developed using a planar monopole antenna. Te proposed antenna provides an ultra-wideband response, and its overall size is 18 × 19 × 0.8 mm 3 . Te proposed ultra-wideband monopole (UWM) antenna has a maximum measured gain of 3.77dBi, and the maximum SAR value for an input power of 18 dBm is less than 1.6 (W/kg) (1g). A bovine tibia is experimentally tested using a proposed UWM antenna to monitor the fracture recovery process and then further analysed using principle component and linear regression analysis. In addition, a microcontroller with a wireless communication module is developed to monitor the data in an Android application. Te proposed system could be a promising approach for developing a portable, noninvasive monitoring device.


Introduction
Human bones are divided into fve categories based on their basic morphologies and endochondral or membranous characteristics. Tese categories are as follows: long, short, fat, sesamoid, and irregular [1]. Te tibia is a shaft connected to two ends of the epiphysis. It is the longest bone and is situated at the bottom end of the leg. Most of the time, children, athletes, and older people who do not get enough vitamins, have weak bones, or get into accidents can break their tibia [2]. Hence, the breaks in the bones are classifed into diferent categories. If the bone ends pierce the skin, the fracture is classifed as a laceration. If it is a complete break, the break is classifed as a sprain and an orientation break [3]. Normal healing of a bone fracture involves the formation of a blood clot and callus to shield the injured area. New strands of bone cells sprout on both sides of the fracture. Te connections between these strands are strengthening. When the bone has fully healed, the callus will fall of. Figure 1 depicts the stages of fracture healing. In general, it takes 6-8 weeks for a fracture to heal. However, this time frame is highly variable, both between bones and between individuals. In contrast to a tibia fracture, which can take up to 20 weeks to heal, hand and wrist fractures often mend in only 4-6 weeks [4]. Monitoring the blood clots in a fracture area can assist us in estimating the rate of recovery.
Currently, there are no standardized procedures for monitoring fracture healing; physicians rely on X-rays, which are only useful in the later phases of fracture recovery. X-rays and other conventional diagnosis techniques such as computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET) scans are used to monitor the area around soft tissue to fnd diferent types of bone fractures [5]. Tese techniques use high levels of ionised radiation to monitor the bone fractures in the human body [6]. According to the World Health Organization (WHO), the continuous radiation exposure to the human body may cause cancer [7].
Several researchers have proposed various methods for monitoring at lower and higher frequencies, such as a compact microwave device developed for fracture diagnosis [8]. Tis method uses the Lanczos imaging technique to visualize the fracture by using the transmission characteristics of a microwave ring resonator. Te reported sensor operated at 2.45 GHz and losing the high-frequency contents during image restoration results in blurring in the image. A UWB textile antenna operating in the range of 1.198-4.055 GHz is reported in [9] for various biomedical applications. Te proposed antenna is not low-profle, and there is no adequate detection system to reconstruct the image and monitor the body's subcutaneous interior damage.
In [10], a planar monopole antenna was developed to detect bone cracks or voids. Te developed system is used to detect bone cracks or voids based on a shift in the resonance frequency of the reported antenna. It would be difcult to incorporate this antenna into a portable microwave system because of its large size. A smart bone plate for monitoring fracture healing was developed [11], and electrical impedance spectroscopy was employed to monitor the healing tissue. Te impedance of the bone plates can be examined by sending current into them and measuring the change in impedance to monitor the stage of fracture healing. However, the reported system depends on the impedance variation. If the impedance is minimal, then diferentiation between them is a difcult task [12][13][14][15][16].
In [17], a meta-material-based antenna array is designed for the identifcation of cancerous cells in biological tissue. Te modelled antenna array operated over a 2-12 GHz range. In this method, a realistic phantom breast model is developed, and the model is surrounded by an antenna array, which sends the signal using a single antenna and receives it using the remaining antennas. Te magnitude and phase of received signals are recorded to diferentiate healthy and unhealthy tissue by dielectric characterization. A UWB antenna operating in a 3 to 15 GHz band is used to investigate the specifc absorption rate (SAR) for breast cancer detection, as reported in [18]. Here, the detection of abnormalities was carried out based on simulated SAR values and was not validated experimentally.
In microwave sensing [19], an antenna is used to transmit a radio-frequency signal to tissues. As a result of the higher dielectric constant of unhealthy tissue, the received signals are dispersed diferently. Te tumor is also identifed by dielectric characterization as both healthy and malignant tissue. Terefore, there is space to develop efective antennas with a robust detecting system for a wide range of biomedical applications such as for the wireless body area networks (WBANs) and wearable devices: Te planar monopole antennas used in wearable devices or WBANs typically require moderate gain [20]. Te planar monopole antenna could be used to detect abnormalities in the human body because it is easy to develop and fabricate on a dielectric substrate and easy to integrate with other devices [20][21][22][23][24].
In this work, a new approach for monitoring the recovery process of bone fractures is proposed using an ultrawideband monopole (UWM) antenna and data processing techniques. Here, machine learning preprocessing techniques are used for monitoring the recovery process and detecting the length of the blood clots to estimate the recovery rate. One is principal component analysis (PCA), and the other is linear regression analysis (LRA). PCA is used to determine whether the bone is fractured or not, and LRA is used to detect the length of the blood clot. Using PCA, the data can be explored easily to determine the most critical variables and where potential outliers could be hidden [25]. Te statistical technique known as regression analysis is used  to investigate the nature of the connection that exists between two diferent variables. Tese two methods are used to analyse the data which are received from the antenna.

The UWM Antenna Geometry
Te evolution of the UWM antenna is shown in Figure 2. Te structure of the proposed UWM antenna is composed of three layers. Te front and back layers are made of copper material, and the middle layer is comprised of fame retardant (FR)-4 material, whose permittivity is 4.3 and height is 0.8 mm. Te basic circular monopole antenna dimensions are approximated using the formulas given in [26] and [27].
Te lower frequency (f L ) of the basic monopole antenna, relating to VSWR ≤2 is where r � 1.06 mm is the efective radius of the circularshaped patch, L � 4.24 mm is the diameter of the cylindricalshaped monopole, w � 0.52 mm is the distance between the feed line with respect to the ground, and k is the empirical value of the FR-4 dielectric layer. Te basic monopole antenna is comprised of a planar microstrip circular patch coupled to a microstrip feed line, as shown in Figure 2(a). To achieve the desired frequency band, the pattern of the radiating element is changed from the initial step of evolution to the fnal step. Te basic circular monopole antenna generates a narrow operating band from 6.84 GHz to 7.37 GHz. Adjusting the distance between the patch and ground plane (w), modifying the electrical size of the patch, and changing the ground as structure help improve impedance matching, a wider bandwidth, and other performance metrics [28][29][30].To achieve the wider bandwidth with proper impedance matching, two elliptical-shaped radiators are integrated with the circular-shaped radiator, as shown in Figure 2(b). As a result, dual bands, from 3.7 GHz to 4.47 GHz and 5.98 GHz to 7.58 GHz, are obtained. In the third step of evolution, the circular radiator is further modifed by adding two more elliptical-shaped radiators at 45°and 65°, respectively, on the left and right sides, as shown in Figure 2(c). As a result, the existing bands are shifted towards the lower frequency of the UWB. In Figure 2(d), semihexagonal-shaped slots and semicircular slots are introduced in the ground plane and radiator, respectively, to achieve a wider bandwidth from 3.14 GHz to 7.16 GHz. In Step 5, the ground plane is further modifed by introducing rectangular and triangular-shaped slots, as illustrated in Figure 2(e). As a result, the desired ultra-wide band of 3.14 GHz to 11.17 GHz is achieved. Hence, the proposed UWM antenna resonates in the ultra-wide band with three resonating frequencies, 3.76 GHz, 6.7 GHz, and 9.8 GHz, with a refection coefcient better than −20 dB. Te overall size of the fnal UWM antenna is 18 × 19 × 0.8 mm 3 . Te designed parameters of the proposed UWM antenna are L su � 19, W su � 18, R 0 � 1.48, R 1 � 6.89,  Figure 4. Te fabricated antenna is measured using a Keysight Field Fox Microwave VNA (N9926A-14 GHz). Te frequency response of the fabricated prototype ranges from 3.23 to 10.83 GHz. Figure 5 illustrates that across the entire ultrawideband, the proposed UWM antenna has a realised gain of 2.1 to 3.86 dBi and a radiation efciency ranging from 87.5 to 98.3%. And it is observed that the average radiation efciency for the entire operating band is 92.3%, and the maximum gain is 3.91 dBi at 6.8 GHz.
In an anechoic chamber, the radiation performance of the proposed UWM antenna is measured. Te normalized measured and simulated radiation patterns of the proposed UWM antenna for the φ � 90°and φ � 0°at the frequencies of 3.76, 6.7, and 9.8 GHz are plotted in Figure 6. Te proposed UWM antenna has a stable omni-directional pattern for φ � 0°and a bidirectional pattern for φ � 90°. Te maximum measured gain of the proposed UWM antenna is 3.77 dBi at 5.97 GHz.
Te frequency responses of the full-wave simulated and measured are presented in Figure 7 and appear to be quite similar. Te comparative study between the proposed UWM antenna and other literature is depicted in Table 1. Te proposed UWM antenna is more compact and has good isolation than the other reported works. It has a higher gain compared to the antennas in [31,32]. It has a good average radiation efciency when compared to the antennas in [33].

Investigation of SAR.
In a full-wave simulator, a simple model of a tibia is designed using skin, fat, muscle, and bone in the shape of a cylinder, as shown in Figure 8. Te tissue properties are assigned to a four-layer model according to the values given in [34,35]. Te radii of the layers are b 1 � 15 mm, b 2 � 22.5 mm, b 3 � 30 mm, and b 4 � 30.4 mm. Te antenna is placed at a far-feld distance of F � 18 mm from the surface of the skin. Also, to make it look like a broken bone, which is usually full of blood, a blood strip of length P mm (35, 25, etc.), width 10 mm, and height 5 mm is made between the bone and muscle, and at the bottom of the strip, a 6 mm-wide crack is created. First, we check the essential parameter, SAR, to make sure that the proposed UWM antenna works safely when it is close to a human body.
Te input power (P i ) is set to 0 dBm, and the antenna is kept parallel to the surface of the tibia model without a crack at a far-feld distance from the skin surface. As a result, the simulated SAR values are 0.015 (W/kg) (1 g) at 3.76 GHz, 0.025 (W/kg) (1 g) at 6.7 GHz, and 0.028 (W/kg) (1 g) at 9.8 GHz. To validate the IEEE standard public radiation exposure limit of 1.6 (W/kg) [36,37], the input power is gradually increased from 0 to 18 dBm in various orders and plotted in Figure 9.
Terefore, the proposed UWM antenna with an input power of 18 dBm generates a SAR value of 1.239 (W/kg) (1 g) at 3.76 GHz, 1.541 (W/kg) (1 g) at 6.7 GHz, and 1.583 (W/kg) International Journal of Antennas and Propagation 3 (1 g) at 9.8 GHz. Hence, the proposed UWM antenna achieves safe radiation exposure to the human body with P i � 18 dBm and it will help to allow for deeper tissue penetration for better clinical responses [9].   Step-1

Experimental Validation with Bovine
Step-2 Step-3 Step-4 Proposed  International Journal of Antennas and Propagation [38,39]. Te schematic of the experimental setup is depicted in Figure 10. Te bovine limb is positioned in front of the proposed antenna at a far-feld distance. Te system includes a vector network analyzer (VNA), a microcontroller with a stepper-motor-based sliding plate, a personal computer with MATLAB software, and an Android application. On     International Journal of Antennas and Propagation the sliding platform, a pole with an antenna is mounted. Te mechanically moving platform is capable of forward and reverse motion. Te keysight feld fox microwave VNA functions as a transceiver that transmits electromagnetic signals via the transmitting antenna [40]. Tis VNA is also connected to the computer through a general purpose interface bus (GPIB) port, which analyses the received data using MATLAB. And fnally, the microcontroller with a wireless communication module is used to monitor the data in an Android application.

Fracture Detection.
Te experimental investigation is carried out on 6 mm-cracked and normal areas of the bovine tibia. Te |S 11 |-parameters for fractured and normal regions are recorded with 1001 frequency points and are plotted in Figure 11(a). A machine learning method, PCA, is applied to the recorded S-parameters. Te PCA is a statistical unsupervised machine learning technique that translates multidimensional data into two principal components (PCs), which lie on the X and Y axis and are represented by PC-1 and PC-2. PCA transforms data with m columns (features) into a subspace with m or less columns while preserving the original data's essence. In our application, the data set was interpreted as the refection coefcients for 1001 frequency points in a normal area (N A ) and a fracture area (F A ). Terefore, the N A and F A datasets are represented as the two observations and detecting a fracture is accomplished by tracing the two coordinates representing the two spots in the 1001-dimensional data. If there is any correlation between the observations (N A and F A data sets), the associated points are observed as PC1 and PC2 in two-dimensional space. Te PCA technique code is developed in MATLAB. Te PCA result of the refected signals from fractured and nonfractured bovine tibia regions is plotted in Figure 11(b). By observing the PCA plot in Figure 11(b), it is simple to distinguish between fractured and normal areas with the principal components, A F and A N .

Fracture Healing Monitoring.
Te experimental investigation of the fracture recovery process of the bovine tibia is shown in Figure 12. Te fractured portion of the bovine tibia is covered with blood clot strips measuring 35 × 10 × 5 mm 3 , 25 × 10 × 5 mm 3 , and 15 × 10 × 5 mm 3 (one after the other), and the corresponding refection parameters and lower cut-of frequency shifts are recorded and compared in Figure 13 to determine the relationship between blood-clot length and frequency shift, as noted in Table 2.
Te LRA is one of the most prominent modelling approaches because it can estimate the value of a response variable based on the predictor variable [41]. Using LRA, the relationship between the length of the blood clot and the frequency shift is formulated, and it is given as whereas the length of the blood clot (L b ) is the response variable, the frequency shift (F s ) is the predictor variable, and the error to best ft the response value, L b , is ∆E � 0.517. Te average percentage of error calculated for the three predictor variables is less than 1%, and the goodness of ft (R square) is 0.935. Hence, the percentage of error is very low for the diferent lengths of blood clots. Terefore, equation (2) is International Journal of Antennas and Propagation more accurate to measure the L b for monitoring the fracture healing process.
An Android application for blood clot monitoring has been developed to measure the length of blood clots on mobile devices, and it is named blood clot monitor (BCM). Te BCM application will connect wirelessly with a nearby microcontroller connected to the HC-05 Bluetooth module. Te steps to create a wireless connection in the developed BCM application are as follows: (1) Initializing the default Bluetooth device on an Android phone. (2) Checking whether the HC-05 Bluetooth module's MAC address is connected to the microcontroller.
(3) A separate thread is created in the code to initiate a connection using the MAC address, and this thread will manage what happens if a link is successfully established or fails. (4) Once a connection is established, the thread will call back for the codes that manage data exchange between two devices.
Tis thread will read the incoming data transmission from the microcontroller which is connected to the VNA through a graphical user interface, and the data are sent to the BCM application. Te BCM application enables the user or physician to determine the status of fracture healing. Equation (2) is used in microcontroller programming to    Figure 14: Te blood clots monitor application (a) before and (b) after scanning. predict the length of a blood clot based on the frequency variation obtained from the proposed UWM antenna using MATLAB. Trough the process of comparing the data points in MATLAB, the shifted frequency is transmitted to the serial port of the microcontroller. Te mobile app is developed using an open-source web app builder. Figure 14(a) depicts the user interface of the developed application. Te user interface consists of two layouts: the upper layout is for wireless connection and status, and the lower layout is for reading data from the HC-05 Bluetooth module.

International Journal of Antennas and Propagation
When the user clicks on the "click to connect" button, the application shows the available Bluetooth networks in the surrounding area. Te corresponding network should be selected, and the status of the connection is displayed on the below widget. After successful connection, user has to click on the "SCAN ON" button, and the data (blood clot length) is read from the HC-05 module via Bluetooth Technology and displayed on the respective widget, as shown in Figure 14(b). Te comparison of this work on bone fracture healing with other studies is tabulated in Table 3.
Te highlights of the proposed work are as follows: (i) A compact planar monopole antenna is used in a noninvasive, nonionized method to identify bone fractures and monitor their recovery. (ii) Te proposed UWM antenna prototype is tested on bovine bone, and the efect of various lengths of blood clots along with fracture is analysed. (iii) Te approach based on machine learning is used on refection coefcient parameters to detect fractures and their healing. (iv) Te proposed UWM antenna is integrated into the embedded hardware system. (v) A new Android application is developed to access the data from the embedded hardware system. (vi) Te length of blood clots measured by the Android application and the recovery rate can be estimated by the user. (vii) In comparison to previously reported works, the proposed diagnostic system has several advantages, including a noninvasive, precision, portability, the absence of a skilled technician, and a planar antenna (easy to integrate with RF/IoT devices) for crack detection and healing monitoring, which is relatively new.

Conclusion
A planar, 0.8-mm-thick ultra-wideband monopole antenna is designed and fabricated to monitor the fracture recovery process. Te proposed UWM antenna has a maximum measured gain of 3.77 dBi and a simulated average radiation efciency of 92.3% across the entire operating band. Te tibia model is created in the full-wave simulator to analyse the SAR for various input power levels, and the resultant SAR values are below 1.6 (W/kg) (1 g) at input power of 18 dBm. Te experimental study is also conducted on the bovine tibia and PCA and LRA is employed to detect and monitor the fracture recovery process, respectively. An embedded system-based setup with an Android application has been developed to monitor the healing status further.
Te experimental study has confrmed that it is possible to use a single antenna to monitor the stages of the recovery of a bone fracture for clinical diagnosis.

Data Availability
Te data used to support the fndings are available from the corresponding author upon request.

Conflicts of Interest
Te authors declare that they have no conficts of interest.