Developing a wireless sensor network based on a proposed algorithm for healthcare purposes

Abstract

This letter describes a developed wireless sensor network based on a proposed algorithm for monitoring the environmental parameters in healthcare intentions. This proposed algorithm contains a frame with different packets that are implemented on the developed wireless sensor network. The developed wireless sensor network consists of one central node as well as four sensor node that has been equipped with various sensors such as temperature, humidity, CO, CO₂, and passive infrared sensor. In order to test the presented algorithm and the developed wireless sensor network, the sensor nodes are situated in four different rooms in a hospital for recording essential parameters of the environment while the central node is put in the nurse station for warning to nurses. The obtained result of the proposed sensor nodes in comparison to gold standards shows root mean square error 1.1%, $0.35{}^{\circ }\text{C}$, 0.98% for humidity, temperature and gas, respectively. Also, the obtained results illustrate that the system gives accurate feedback from environmental temperature, humidity, and CO, and CO₂ to the nurse station in order to increases the possibility of a healthy environment condition for patients.

Introduction

Wireless sensor networks (WSNs) employ different sensors to monitor physical or environmental conditions such as temperature, sound, pressure, etc [1]. The WSNs pass their data through the networks to other locations. Although the WSNs had first proposed for specific applications, today, such networks are used in many consumer applications, such as industrial process control as well as health monitoring [2].

The expanding use of wireless networks besides the miniaturization of electrical devices have improved the development of wireless body area networks (WBANs) [3]. A WBAN makes a continuous health monitoring of a patient without any constraint on daily life movements. WBANs considerate a number of innovative applications such as smart health care, assisted the elderly living, emergency response and interactive gaming [4]. The first WBANs had many drawbacks such as high power, low bit rate, low reliable in transmitting data, high-cost and high error. While today WBANs have largely resolved the problems, and they can be appropriate for different applications especially in healthcare. Many types of research classified WBAN in medical and non-medical applications [5]. The WBSNs for medical applications can be divided into several types: implanted, wearable, and environment-embedded [6]. In other words, the purpose of WBSNs for medical use is a distinction between in-body and on-body applications [7]. The implantable medical devices are inserted inside the human body [8]. While wearable devices are used on the body surface of a patient. Wearable WSBNs can be divided into a variety application comprise of telemedicine and remote patient monitoring, rehabilitation [9] and therapy [10] as well as biofeedback [11]. Telemedicine allows medical professionals around the world that benefit integrated health systems and telecommunication technologies to the delivery of patient care [12–14]. According to this, telemedicine networks for the delivery of several healthcare services have essential effects on declining healthcare costs especially among the aging population [15]. In order to hold a correct motion pattern in a rehabilitation course, patients’ movement needs to be continuously monitored and also rectified utilizing the WBNS [9]. Biofeedback systems benefit WBSA to control and modify the physiological activity with the aim of improving health and performance employing the measurement of physiological activity. Another medical WBSN application is environment-embedded systems which employ sensors contained in an environment to monitor its essential parameters [16].

In this letter, a simple algorithm of wireless sensor networks for developing a wireless sensor network based on low-cost commercial wireless modules in an environment-embedded application is proposed. To this end, a wireless sensor network is developed which consists of four sensor nodes and one central node. Then, the proposed algorithm, which was written with Bascom-AVR software, is mounted on the processors of the sensor nodes and central nodes. The obtained results show that the system has more reliability than [11, 17, 18] and was established with minimum cost than [11, 17, 18]. Also, the cost, as well as the power of the proposed system, is lower than in comparison [19–22]. Furthermore, the proposed system records movements of patients in the room that knowing the condition accounts for the nurses an important parameter. Furthermore, the proposed algorithm can easily be mounted on a common commercial module such as HM-TR or RF-12.

The proposed algorithm

In order to implement the proposed algorithm, it is mandatory familiarizing with the employed modules. The modules (HM-TR) use the FSK modulation. The communication in the modules is half-duplex which can perform two works sending and receiving. The module operates at industrial, science, medical (ISM) band to send and receive the information and it does not need to permit the radio and frequency management for having the communication. A proposed frame is defined as Fig. 1 for the communication of the central node and sensor nodes.

Fig. 1.

The proposed frame for communication between the central node and the sensor nodes

The central node sent the frame for all of the sensors nodes. As it is shown in Fig. 1, the value of 255 is the start of the frame. After the start packet, an address packet of 32 bites is defined which has been changed by the central node to connect with every special sensor node. Every sensor node has a special address that its value is saved on sensor nodes’ flash memory. Indeed, the central node works based on time division multiplexer and it allocates every time a specific address and then sends it to all of the sensor nodes and wait to respond from the sensor nodes. If the address value that is made by the central node equals the address of the sensor nodes, which is saved on the sensor nodes’ flash memory, the sensor node put data on the data packet considering the function packet. Otherwise, the sensor nodes do not carry out any communication with the central node. The function packet is an ability of the central node to access every type of sensor data in the sensor nodes such as temperature, humidity and so on. For example, if the central node only seeks temperature data of the sensor node, it gives 1 to the function. While for other sensor data, the central node gives other numbers to the function package. The content of the function is put on the data packet in the frame.

If the data packet contains two values comprise of 255 and 254, it causes to have some problems because the start and end of the frame are 255 and 254, respectively (Fig. 1) in the proposed algorithm. In order to separate the data from the start and end of the frame, a protocol packet is defined, as shown in Fig. 1. The protocol packet helps to diagnose the data from the start and the end of the frame. If a 255 or 254 emerge in a data packet, the value of the protocol packet changes from 1 to 0 in every sensor node. It should be noted that the protocol packet is 1 byte. Accordingly, the data packet is 8 byte. After identifying the packet of the address, protocol, function as well as the data, the CRC (cyclic redundancy check ) is gotten from the packets and its value is saved in the CRC packet. Finally, the end of the frame is closed with 254 value in the sensor nodes and the sensor node sends the frame to the central node. The frame is received by the central node and first, its CRC is gotten and then is saved in the central node. Then, the CRC in the central node is compared to the CRC made the sensor node that is transmitted wirelessly. If the central node CRC is equal to the sensor node CRC, the central node saves data and is prepared for communicating with the further sensor nodes. If the central node CRC is not equal to the sensor node CRC, the central node transmits a command to the sensor node for resubmitting the frame. All explanations of the proposed algorithm have been shown in Fig. 2.

Fig. 2.

The proposed algorithm for the central node and sensor nodes

The component of developed system

The developed system for implementing the proposed algorithm comprises a central node, a relay node as well as four wireless sensor nodes as shown in Fig. 3a. As it is shown in Fig. 4b, the sensor node consists of a microcontroller, a transceiver as well as temperature, humidity, light, passive infrared (PIR), and gas sensors. The microcontroller is an Atmega 32 AVR series that carry out tasks such as converting the data from analog to digital, calibration and then sends the data to a transceiver. The transceiver, which operates in the ISM band at frequency 433 MHz, is connected to the microcontroller through the serial protocol in the sensor nodes as well as central and the relay node. One of the features in the transceiver is in sleep mode. Indeed, when the transceiver does not need to work, the microcontroller can change its mode to sleep for saving energy.

Fig. 3.

The prototype of the developed system; a wireless sensor network, b the sensor node, c the relay board, d and the central node

Fig. 4.

Schematic of measurement setup; a sensor node, b central node, c and relay node

In order to determine an environmental temperature, an LM75 sensor is used. The sensor output is digital. The change of temperature in the environment causes a set of alterations in the sensor output. By using the $I_{2}C$ protocol of the microcontroller, the changes can be read through two pins SDA and SCL of the microcontroller (Fig. 4a). Another sensor in the sensor node is the humidity sensor. The sensor is a capacitive sensor that its value changes with the moisture of the weather. In order to bias the sensor, an NE 555 IC and 4 resistances are used [23]. Indeed, the IC is employed to make a typical astable design with variable frequency when the capacitive capacity of the humidity sensor alters (Fig. 4a). In order to read out the variable frequency, T1 pin of the microcontroller (Atmega32) is used. For this end, Timer#1 is defined as a counter and its value increases when a rising pulse occurs. Then, Timer#2 is employed in order to set an accurate time for countering the rising pulse. With the number of the rising pulse and its time (for 1 s), the frequency of the sensor is obtained. According to datasheet [23], every frequency is related to the amount of humidity and finally, by having the frequencies, the humidity is estimated. Gas sensors are a resistive sensor. The gas sensors have 4 pins in which two pins are related to warm its core with a pulse voltage. While two other pins are the voltage output that change encountering with gases consists of CO and CO₂. Finally, the analog to digital converter (ADC) of microcontroller, which is 10 bits, record the changes (Fig. 4a). Considering to utilizing ADC of Atmega32 for two sensors of gas and light, it is mandatory to enable an anti-noise circuit for the decrease of the noise because their ADC values vary when users surround their hands to the circuit. Moreover, a PIR sensor is used for identifying the motion of humans and other living species surrounding the sensors. The sensor output is given to the ADC of the microcontroller after amplifying it (Fig. 4a).

The central node consists of a transceiver, a microcontroller as well as a TTL to USB converter (Figs. 3d, 4b). Indeed, the transceiver receives the data and sends them to the microcontroller. Then, the microcontroller transmits the information to a personal computer (PC) to visualize and further process. Also, the microcontroller shows the information on an LCD in the central node if a PC is not available. Furthermore, the central node can connect to a phone socket to call with users. Indeed, the users can save several numbers in the central node besides other setting such as maximum and minimum light, temperature, humidity, CO₂ and CO and also activate or deactivate PIR. Ultimately, The relay board is connected to the central node wirelessly to receive commands for ON or OFF the cooler or heater. The Board consists of a transceiver and several relays as well as an atmega8 and also an LCD (Figs. 3c, 4c).

Results and discussions

Accuracy investigation in the developed sensors

Figure 5a illustrates a Bland–Altman plot for temperature measurements which were obtained 60 pairs of the temperature sensor in [24] as a reference and the developed sensor. This plot is procured by calculating the difference and average of the temperature measurements as the vertical and horizontal axis, respectively. As the results reveal, the mean error is $0.21{}^{\circ }\text{C}$ with 95% limits of agreement between $-0.34$ to $0.76{}^{\circ }\text{C}$. The Bland–Altman plot is based on the quantification of the agreement between two quantitative measurements by studying the mean difference and constructing limits of agreement. The Bland–Altman plot analysis is a simple way to evaluate a bias between the mean differences and to estimate an agreement interval, which is named 95% limits [25]. Data can be analyzed both as unit differences plot and as percentage differences plot. In another figure (Fig. 5b), the Bland–Altman plot of the humidity sensor for 60 pairs of humidity measurements consist of the developed sensor and a reference in [26] is shown. The measurements illustrate a mean error 0.02% while 95% limits of the agreement are restricted to $-2.16\%$ and 2.21%. As Fig. 5c shows, the 120 pairs of gas measurements from the developed sensors and reference in [27] for the Bland–Altman plot are used that mean error is $-0.05\%$ whereas 95% limits are $-1.97$ and 1.87. Also, the Bland–Altman plot gives more information such as standard deviation and root mean squire error. According to this, the standard deviation (SD) and root mean squire error (RMSE) are 1.12% and 1.1% for light, respectively. The SD and RMSE of the temperature measurement are $0.28{}^{\circ }\text{C}$ and $0.35{}^{\circ }\text{C}$ while for the gas measuremen obtained 0.98% and 0.98%, separately. As these values indicate, a very good agreement was obtained between the proposed system and the references. It should be noted that some results have deviations of more than 5%. The deviations are from the sensor because if some problems occur in transmission such as the lack of data in receiving in the central node, the sensor nodes send new data again to the central node. Therefore, the data is extracted from the sensor and the occurrence consist of accuracy and deviations also are related to sensors.

Fig. 5.

The Bland–Altman plot of the measurements for the developed sensor and references ; I temperature, II humidity, and III gas (CO₂, CO)

Environmental parameters monitoring

In order to provide a healthy environment for a patient, the sensor nodes were situated in a sick room while the central node is put in nurse’ stations. The nurses can set the maximum and minimum values for all the sensors with the keys on the central node (Fig. 3c).

To investigate the performance of the developed system, every sensor is separately tested. First, the temperature sensor is employed for setting a suitable temperature for patients. To this purpose, the maximum and minimum temperature is set in a room’s hospital in the central node by the nurses. The values were set on 37 and 24 Celsius, respectively. After warming the room by using a heater, the environmental temperature increases from the setpoint ($37{}^{\circ }\text{C}$) and the temperature sensor followed the variations. When the temperature sensor reaches values more than the maximum setpoint, the nurse’ stations inform from the variations (Fig. 6I). On the other side, the cooler is used to decline the temperature. When the temperature decreases from the minimum setpoint ($24{}^{\circ }\text{C}$), the central node warns the nurses. The experiments simultaneously are performed in two rooms with separated sensor nodes and the obtained results show high accuracy in awareness from environmental temperature (Fig. 6I). Also, any variation is not observed in another room because of not being active in the heater or cooler in the place (Fig. 6I). In another test, humidity test, warm air of the mouth on the humidity sensor is blown. By performing that, the amount of capacitor capacity of the sensor changed and result in increasing humidity. So, the sensor output exceeded the defined setpoint (60%) and the variations were seen in the central node in the nurses’ station (Fig. 6IV). Indeed, the humidity sensor is for setting moist in the room which is an essential parameter for patients with lung problems. In order to test the gas sensor of CO₂ and CO, little smoke was used. When the smoke hits the sensor, the output of the sensor altered, and because their values expanded more than setpoint for CO₂ and CO (10, 15%, separately), the central node makes a beep in the nurse’ station (Fig. 6II, III). The main reason for using the gas sensor is that some people or patients smoke in hospitals when any nurses are not in their rooms. However, with the proposed system they cannot carry out this work and nurses inform from the situation. It should be noted, the nurses can also activate the PIR sensor for informing from the movement of patients, especially for specific patients. Thus, they can know the useful information of the movement of patients.

Fig. 6.

The snapshot of the obtained results of the developed sensor for two a room including I temperature, II CO₂, III CO and IV humidity

By having the proposed system in hospitals, the nurses obtain more information about the environmental condition of the patient and can respond to the variation with high speed. Also, it should be noted that the developed system not only can be used in the hospital but also can be utilized in the home for determining whether the place is a healthy environment for a patient or not.

As Table 1 shows, the proposed system owns a lower bit rate in comparison to Zigbee and Bluetooth in [21, 22]. However, the proposed system benefit from two main features consists of low-power and low-price than Zigbee and Bluetooth in [21, 22]. The low-power help to the presented system for running with a battery for a long time and low-price provides more users for it. Other methods such as Lora and SIGFOX in [19, 20], need a special substrate for this purpose in which results in high-cost for implementation. Additionally, the Lora and SIGFOX can update their data in limited times (124 times) during a day and it is not very applicable for sensitive use. One advantage of them is lower-power than the presented system for sending the data and accordingly, they work with low energy. Nonetheless, they are not recommended for sensitive applications that need to send data with more than 124 times in a day.

Table 1.

Comparison of the presented system with other works

References	Wireless tech.	Frequency of operation (MHz)	Data rate (kbps)	Power mW	Price $
[26]	LoRa	868,915 MHz	50	100	30
[27]	SIGFOX	868,915 MHz	0.3	25–160	25
[19]	ZigBee	2.4 GHz	250	40	40
[20]	Blutooth	2.4 GHz	114	30	10
The proposed system	HM-TR	433 MHz	19.2	20	9

Recommendations and limitations

Two main factors are considered in the proposed system. First, the placement of the system is very important for maximum coverage of every room. If the sensor nodes are not situated well in a place, they have low accuracy. Second, In a power outage, the hospital needs to emergency electricity for supplying power. If hospitals are not associated with the option, occurring an abnormal condition would be plausible in the hospitals. In the presented system, the variables of the main menu are saved in EEPROM memory and setpoints do not change with the power outage.

One of the limitations of the system is the size. However, it is possible that we minimize the size. But the cost increases very much. Thus, we carried out a trade-off between size and cost. For example, there are many temperature sensors in the market with different size. If the size is miniature, the price also would increase. We selected a temperature sensor with a medium price and a moderate size. The temperature sensor gave good accuracy and we reached an acceptable result. Nonetheless, it is possible that we change the temperature sensor with a smaller size and more accurate. If we selected a high accuracy sensor for sensing, the cost would be high. Thus, users determine what type of sensor with which size can be used.

In the future, we will implement the algorithm on new wireless technology for reaching lower power as well as will develop the system with a smaller size.

Conclusions

In this letter, a comprehensive description of developing a wireless sensor network based on a proposed algorithm in healthcare applications was presented. The presented algorithm contains a frame with several packets that were considered for transmitting and receiving data in the nodes. After developing the wireless sensor network which comprises four sensor nodes and one central node, the proposed algorithm was implemented on them. Then, the system was mounted in hospital rooms and a nurse station for monitoring environmental parameters. The outcome results reveal the system is suitable feedback for monitoring humidity, temperature and gas (CO, CO₂) with RMSE of 1.1%, $0.35{}^{\circ }\text{C}$, 0.98%, respectively. So when some variations of temperature, humidity, and gas occur in the sensor nodes and that moves beyond the setpoints which are defined by the nurses, the central node acts a beep. Consequently, the proposed algorithm and the developed system can make a healthy environment for the patient.

Compliance with ethical standards

Conflict of interest

Reza Abbasi-Kesbi, Zahra Asadi and Alireza Nikfarjma declare that they have no conflict of interest.

Ethical approval

This article does not contain any studies with human participants or animals performed by any of the authors.

Informed consent

Not applicable.