Internet of Things-Based Electromagnetic Compatibility Monitoring (IEMCM) Architecture for Biomedical Devices

Abstract

Electromagnetic compatibility is the capability of electrical and electronic equipment to function properly around devices radiating electromagnetic energy, without mutual disturbance. Hospital environments contain numerous devices operating simultaneously and sharing resources. Undetected electromagnetic interference can cause medical devices’ malfunctions, exposing patients and staff. Traditional monitoring is time-consuming and relies on expert interpretation. An Internet of Things-enabled embedded system architecture for remote and real-time monitoring of electromagnetic fields from medical devices is proposed. It integrates frequency probes, a Raspberry Pi 4, and a communication module. A three-month study conducted at Muhima District Hospital, Kigali, Rwanda, demonstrated the system’s effectiveness in monitoring electromagnetic field levels and cloud transmission. The signals were benchmarked against International Electrotechnical Commission and Rwanda Standards Board standards. Alerts are triggered when thresholds are exceeded, with results plotted on website and mobile interfaces. Emissions were highest at noon when the equipment was most active and lower after 1:30 PM, indicating reduced activity. The sample recorded statistics of electric fields include mean (1.0028), minimum (0.7228), and maximum (1.3515). Among the five filters evaluated, the Savitzky–Golay performed better, with MSE (0.235) and SNR (9.308). A 412 ms average latency and 24 h operation was achieved, offering a portable solution for hospital safety and equipment optimization.

1. Introduction

The healthcare industry increasingly uses advanced electronic and wireless-enabled medical devices to improve patient care. These devices emit electromagnetic radiation (EMR) during their operation, which could otherwise cause interference with other devices in their vicinity, requiring vigilance regarding their compatibility to prevent risks of interference [1]. Therefore, ensuring the electromagnetic compatibility (EMC) of medical devices is crucial because interference-related failures can have adverse consequences ranging from mild to fatal [2,3]. Electromagnetic interference (EMI) is a phenomenon that may occur when an electric device is exposed to electromagnetic waves [4]. The International Electrotechnical Commission (IEC) regulates EMC in medical devices. National standards, such as those of the Rwanda Standards Board (RSB), are also applied as needed for regulation. The RSB standards are adopted by the African Electrotechnical Standardization Commission (AFSEC), which is a governing body that promotes harmonization and adoption of IEC standards in African countries [5]. Compliance is evaluated before devices are released on the market and installed in medical facilities. Existing guidelines and regulations may not fully address the EMI in these devices [3]. The electromagnetic environment of the hospital is increasingly complex, influenced by factors such as the use of mobile phones and the presence of various electronic medical and non-medical devices, which are absent in EMC testing laboratories [5]. Medical devices lose electromagnetic resilience or immunity by undergoing repairs and aging [6]. Previous research works indicate that many such devices face undetected EMI-related problems, leading to unclassified failures due to limited monitoring systems [7,8]. Monitoring the environments of these medical devices for EMC ensures the reliability of the equipment as well as the overall safety of patients and medical personnel [9].

Due to their role in human treatment, biomedical engineers must manage the EMC of medical devices [10]. The issue of EMC in healthcare institutions has been the subject of research efforts. Kurta et al. performed measurements before and after the hospital became operational to assess the electromagnetic environment and determine safety levels. They found EMI with one defibrillator and two electrocardiograms [11]. Other studies measured and characterized electromagnetic fields in hospitals to investigate their effects on medical equipment. Although no EMI was recorded, they suggested that continual EMC assessments be performed [12,13]. The influence of mobile phones on medical devices EMC was assessed. Although the measured electromagnetic field strengths were not very high, it was concluded that if interfering devices are near medical equipment, EMI could occur [14,15]. Most of these studies were conducted manually by experts, as they were the ones who could interpret the results. However, manual measurements are time-consuming and impractical for continuous and large-scale monitoring due to the need for expert fault-labeling. Automated monitoring offers an appealing solution, especially amid safety concerns and a shortage of experts [16].

To address these challenges, researchers have explored the use of the Internet of Things (IoT) for automated radiation monitoring. The IoT technology is crucial for remote monitoring because it facilitates the connectivity of sensors and devices over the Internet, providing convenience and real-time data on critical parameters [17]. Tocchi et al. implemented an IoT-enabled wireless sensor network for environmental radiation monitoring in decommissioned nuclear plants. The system stored data on a central server, and access to monitoring data was not in real time [18]. Saleem et al. developed an IoT-based remote real-time radiation measurement system that targets ionizing radiation. They used a Waspmote, a radiation sensor board, a Geiger tube, and a 3G module. The system used the Advanced Encryption Standard (AES) to encrypt the data. Their work did not include remote notification to the caretaker, which is crucial for remote monitoring [19]. Nyakuri et al. implemented a WSN capable of capturing real-time radiation data to evaluate medical device EMC. However, they focused on prediction and barely mentioned the monitoring aspect [20].

An IoT-based WSN was developed for radiation monitoring; however, the system relied on a personal computer (PC) for data processing, limiting mobility [21]. An advanced embedded EMR monitoring system was developed and implemented near cellular tower base stations to monitor radiation emissions. It focused on cellular tower stations and relied on PCs to access data [22]. Radio monitoring systems were developed using calibrated equipment, a structured algorithm, an omnidirectional wideband antenna, and a spectrum analyzer. The implementation required a connection to a computer with dedicated software and an administrator to manage the measurements, and access to the data was not real time [23]. Vega et al. designed an IoT-based non-ionizing radiation monitoring prototype for outdoor environmental radiation monitoring. The system was not specifically developed for medical devices and required dedicated software for data interpretation [24]. Rana developed an IoT-based electromagnetic radiation detector for emissions generated by computers and other electronic devices. However, it did not offer remote notification when radiation levels were too high and was also not specific to medical devices [25]. Raman et al. applied the IoT technology to develop a monitoring and control system for industrial applications, integrating real-time alerts when radiation fields are high, to ensure optimal field conditions and improve maintenance efficiency. Similarly, the system can be applied to detect faults caused by EMI in medical devices [26].

The existing automated systems do not support either real-time or remote monitoring with alerting. There is a need for a trade-off between a low-cost EMC monitoring system and a well-calibrated one [27]. A pressing challenge in EMC monitoring is in balancing the cost of installing and maintaining accurate and well-calibrated sensing infrastructures against affordability and the reduced accuracy of low-cost sensors. Moreover, most calibrated sensors rely on dedicated software, limiting customization and adaptability. There is a need for monitoring systems that enable integration and interoperability of the equipment with other devices and technologies. Most existing solutions either lack real-time visualization, secure and mobile data access, dedication to medical devices, or remote notifications. We thus propose integrating a well-calibrated EMC sensor with a Raspberry Pi 4 microcontroller, equipped with a communication module for sending data to the cloud and alerting personnel. This study explores the possibility of applying an IoT-based architecture for continuous monitoring in long-term equipment EMC evaluation. We monitor EM signals to verify the compatibility of medical devices and assess EMI risks to inform equipment management decision-making. EMI monitoring enables the timely detection of radiation spikes, helping engineers/technicians prevent equipment failures and minimize radiation exposure to staff and patients [28].

The major contribution of this work is the design of the dedicated IoT-enabled architecture that detects and monitors EMC in medical devices. The pioneering implementation of an EMC monitoring architecture for medical devices in Rwanda. The design eliminates the cost and skills gap in hiring an EMC expert to assess EMC in hospitals. It utilizes well-calibrated sensors for EMC detection and enables the possibility of remote and real-time monitoring. The rest of the paper comprises Section 2, Section 3, Section 4 and Section 5: materials and methods, results, discussion, and conclusions.

2. Materials and Methods

Ensuring that medical devices meet EMC criteria for the duration of their useful lives is crucial for hospital safety. Clinical/biomedical engineers should perform medical device electromagnetic foot printing and fingerprinting and continuously monitor their ambient environment for spectral peaks and transients. Measurements can be selective or wideband. Selective frequency measurements allow targeted analysis, compliance testing, and equipment diagnostic testing. Measurements can also be performed near- or far-field. We conducted near-field measurements using an electromagnetic RF meter to measure emissions near medical devices. The EMC measurements were performed at 0.3 m from the medical device, according to the proximity assessment guidelines presented in the IEC 60601-1-2014 [29]. It permits shorter measurement distances for in situ evaluations to account for near-field effects in clinical settings. The ECG or ultrasound machines under surveillance were the only pieces of equipment present in the hospital room. Measurements were conducted in the hospital rooms while the medical devices were in operation. The EMF drops off at the inverse square of the distance, meaning that the field strength rapidly decreases when the distance from the source increases. Therefore, we can assume that the electromagnetic fields detected are specifically from the medical device under surveillance [30].

The IEMCM system architecture depicted in Figure 1 comprises (a) Medical devices being monitored, (b) EM Signal detection system, is the sensing unit that includes the sensors connected to a microprocessor equipped with a communication module, (c) Monitoring and Control center is a secure IoT platform enabling transmission, storage and access control to the stored data, and (d) Visualization component includes availing data on the IoT platform, website and mobile interfaces. The categories in Figure 1 are explained in detail in the following sections: Section 2.1 for Medical Devices, Section 2.2 for EM Signal Detection System, Section 2.4 for Monitoring and Control Center, and Section 2.5 for Visualization. The IEMCM architecture employs two sensors to detect electromagnetic signals within the Industrial, Scientific, and Medical (ISM) band, spanning from 100 kHz to 3 GHz. This frequency range was chosen because it encompasses the typical operating frequencies of most medical devices and potential sources of electromagnetic interference in hospital environments. The broad range ensures the system can detect low- and high-frequency emissions. The frequency range poses significant challenges to both immunity and emissions. Lower frequencies entail reduced energy transfer to susceptible devices, whereas extremely high frequencies can be effectively shielded. Frequencies ranging from 10 kHz to 1 GHz are particularly problematic because they combine high energy potential with greater penetration, making them more difficult to manage in terms of EMI and shielding [31]. The data is transmitted to the Raspberry Pi microcontroller for processing, and the determination of the peak signal is based on the strict safety values of medical devices according to regulations. It was chosen for its compact and lightweight form factor while possessing enough computational capability to process real-time data, making it suitable for deployment in diverse hospital settings. According to the IEC 60601-1-2 standard, all critical equipment must support a 10 V/m electric field, and all non-critical equipment must support 3 V/m at a frequency within the 80 MHz to 2.5 GHz range [32]. The peak values trigger a short message service to inform the biomedical engineer of the environmental status, prompting him to take the necessary measures. The Global System for Mobile Communication (GSM)/General Packet Radio Service (GPRS)/Global Navigation Satellite System (GNSS) HAT communication module was selected because it enables wireless data transmission to the cloud and the sending of Short Message Service (SMS) alerts to engineers when thresholds are exceeded. SMS notifications provide instant alerts, ensuring immediate corrective action. The measured data is stored on the cloud, chosen because it is a lightweight and scalable platform for IoT applications, offering free storage, analysis tools, and real-time visualization capabilities. The parameters were selected considering cost, performance, and adaptability.

2.1. Medical Devices

A maternal and childcare hospital in Muhima district in Kigali, the capital city of Rwanda, was selected for experimentation. Selection was based on its location, the served population, and the availability of medical equipment. It has an approximate total of 560 pieces of medical equipment, and a population of 326,478 people [33]. The selection of appropriate medical devices for testing an EMC monitoring system is a crucial factor that directly influences the reliability and efficacy of healthcare technologies. The selection of medical devices to be monitored was based on their specifications, usage, and compliance with regulatory standards. Medical devices can be categorized as critical or noncritical based on their operation. Due to their crucial roles in patient diagnosis and condition monitoring, the monitored medical devices are electrocardiographs and ultrasound machines. ECG machines detect the heart’s electrical activity to assess cardiac health and evaluate the functioning of intracardiac conducting tissue [34,35]. ECG machines must accurately transmit data for the prompt detection of cardiac abnormalities. ECG waveforms can be distorted even by small EMI, which could jeopardize patient care by causing misinterpretation of results [36]. Similarly, ultrasound machines use sound waves to create finely detailed images of inside organs and tissues, making them essential medical imaging instruments [37,38]. These devices are susceptible to EMI, even with shielding, which can cause artifacts in ultrasound images that obscure diagnostic data and interfere with clinical judgment [39]. ECG and ultrasound diagnostics are essential for patient care; therefore, they require strict EMC monitoring to minimize interference risks and ensure optimal device performance.

2.2. EM Signal Detection System Design

The electromagnetic signal measurement and detection system comprises an EMF-839 RF meter, a GSM/GPRS/GNSS HAT, and a Raspberry Pi 4 Model B for data processing. Both battery and AC power sources were utilized, as the measurements were conducted indoors. The proposed IEMCM system is operated on the power bank’s rechargeable battery that is charging on the rectified AC power supply. The principal sensor for detecting electromagnetic signals is the EMF-839 RF field meter (Manufacturer: Lutron Electronic Enterprise Co., Ltd., Taipei, Taiwan). The meter employs a custom-built large-scale integration (LSI) circuit on a single microchip to measure frequencies ranging from 100 kHz to 3 GHz. It was powered by a 9 V battery. Two probes are used with the RF meter: the EP-04L (a low-frequency probe that covers 100 kHz to 100 MHz) and the EP-03H (a high-frequency probe that covers 100 MHz to 3 GHz). The probes exhibit exceptional sensitivity, enabling the detection of minute variations in the electric field. With an accuracy of less than 2 dB, they capture even subtle fluctuations in signal strength, ensuring high fidelity in measurements. The probes have an input impedance of 50 ohms and are completely passive, making them ideal for monitoring medical devices. Their triaxial configuration simplifies the measurement process by determining the total field strength using three orthogonal axes, x, y, and z, without requiring a change in sensor position. The device computes the total electromagnetic field value “E” for each reading using (1).

$$\mathrm{E}=\sqrt{x^2+y^2+z^2}$$

(1)

where x, y, and z are the EMF values sensed from the x, y, and z directions, respectively.

The value of E is measured in V/m. The EMF-839 RF meter was calibrated before deployment to ensure accurate measurements. Calibration was performed according to the manufacturer’s specifications and ISO/IEC 17025:2017 [40] standards for testing and calibration laboratories. The process involved exposing the device to a reference electromagnetic field source with known strengths across its operational frequency range (50 MHz to 3.5 GHz) under controlled laboratory conditions. The readings from the EMF-839 were compared to the reference values, and any deviations were corrected using the meter’s internal calibration settings. The RF meter only outputs data using the Recommended Standard (RS)-232 Personal Computer serial interface, and all previous studies that have used such devices for data collection have required a direct connection to the computer. Connecting to a Raspberry Pi enables mobility and real-time processing from anywhere, thus maximizing the benefits of ubiquitous computing and IoT technology. A Raspberry Pi is a low-priced minicomputer with low power consumption and the computing power of a desktop [3,41]. Raspberry Pi 4 Model B was selected for processing because it allows universal serial bus connections. It can balance between processing power, affordability, and energy efficiency. It is equipped with a quad-core ARM Cortex-A72 processor and ample memory, which is suitable for processing the EMC data. Additionally, it has versatile connectivity options; Wi-Fi and General-Purpose Input/Output interfaces, making it suitable for deploying a low-cost, scalable EMC solution. Figure 2 shows a block diagram of the proposed IEMCM system. We configured the system to use dual communication channels, which utilize a Waveshare GSM/GPRS/GNSS Raspberry Pi HAT and the hospital Wi-Fi for communication. It is a low-power HAT featuring multiple communication functionalities: GSM, GPRS, GNSS, and Bluetooth.

2.3. Embedded System Dataflow

Minicom 2.10 facilitated communication between the Raspberry Pi and the meter. Minicom is a text-based serial port communications application that enables bidirectional communication with devices connected to the serial port. It enables users to submit commands to the connected device and receive results.

Minicom helps in the configuration, testing, and troubleshooting of various devices. The sensor reading is displayed as a 16-digit data stream on Minicom, labeled D_15 to D_0.

D_0 is the end of the word, and the actual reading/measurement is displayed from D_1 to D_8. D_9 denotes the decimal point (DP) in the radiation reading, where 0 means no DP, 1 means 1 DP, 2 means 2 DP, and 3 means 3 DP. D_10 is a binary representation of the reading polarity, where 0 denotes positive and 1 denotes negative. D_11 and D_12 describe units of measurement, which can be A9 = [W/m]², A8 = [mW/cm]², or A7 = V/m. The readings were set to V/m for easy comparison with the thresholds set in the standards. D13 indicates whether the reading’s upper display data is =1 or the lower display data is =2. D_14 is denoted by 4, and D_15 is the beginning of the word = 02. A pseudocode given as Algorithm 1 was used to convert the sensor data readings into an understandable format.

Algorithm 1: Convert Sensor Data

1. Procedure CONVERTSENSORDATA(data_stream)

2.   data_stream = D15 D14 … D1 D0

3.   Read data_stream

4.   decimal_point = D9

5.   polarity = D10

6.   unit = ‘V/m’ when (D11, D12) == (0, 1)

7.   display_type = ‘upper’ if D13 == 1 else ‘lower’

8.   reading = D1 : D8

9.   If decimal_point == 1 then

10.     reading = reading[:4] + ‘.’ + reading [4:]

11.    Else if decimal_point == 2 then

12.     reading = reading[:3] + ‘.’ + reading[3:]

13.    Else if decimal_point == 3 then

14.     reading = reading[:2] + ‘.’ + reading[2:]

15.    End if

16.    If polarity == 1 then

17.     reading = ‘-’ + reading

18.    End if

19.    Return reading, unit, display_type

20. End procedure

Algorithm 2 is a pseudocode used to verify the radiation value against the thresholds in the standards and trigger a notification to the biomedical engineer/technician while sending all sensor data to the cloud. Medical devices are classified into two categories: critical and non-critical. The EMC threshold for critical devices is 10 V/m and 3 V/m for noncritical devices based on the IEC 60601-1-2 standard [30,39]. When the data is within acceptable ranges, it is sent to the cloud for storage without notifying the engineer. All sensor readings are continuously transmitted to the cloud for real-time visualization and analysis, to ensure comprehensive data coverage and minimize the risk of delayed or missed events. We implemented role-based access control on both the website and the mobile application to ensure that general users, hospital staff, and health authorities have appropriate levels of access to data trends based on their roles. We used 3G cellular communication for data transmission, which offers enhanced security compared to Wi-Fi.

Algorithm 2: EM Radiation Reading Process

1. For each EM radiation reading, do:

2.   Check the device type:

3.   If device == critical then

4.     Set threshold = 10 V/m

5.     If EM radiation > 10 V/m, then

6.       Send SMS Notification

7.     Else

8.       Send data to the cloud

9.     End if

10.  Else if device == noncritical then

11.      Set threshold = 3 V/m

12.      If EM radiation > 3 V/m, then

13.       Send SMS Notification

14.      Else

15.       Send all readings to the cloud

16.      End if

17.  End if

18. End for

2.4. Monitoring and Control Center

The monitoring and control center comprises an IoT platform that is used to receive, store, and disseminate sensor data. We used a privately managed cloud to store the data to ensure full control over data security, accessibility, and scalability. The Message Queuing Telemetry Transport (MQTT) communication protocol was utilized to transmit data from the Raspberry Pi to the cloud via the GSM module. We selected the MQTT protocol due to its lightweight design and efficiency in real-time and low-latency communication, which is essential in IoT applications. The MQTT publish-subscribe messaging model enables seamless transmission and reliable delivery of sensor data to the cloud for storage and analysis. To ensure data integrity and confidentiality, the MQTT protocol supports Transport Layer Security (TLS) over port 8883, with the microcontroller establishing a secure connection to the cloud broker using digital certificates for authentication. A database server and an application server were also implemented. We used My Structured Query Language (MySQL) for the general database and an application server to service the website and the mobile application. To ensure secure access to radiation data through the website and mobile applications, strict access control measures have been implemented. The controls are based on user roles and authentication credentials to determine visibility and functionality.

2.5. Data Visualization

Measurements were taken over three months in the hospital rooms used for bedside ultrasound imaging and ECG testing. The collected data was analyzed to visualize the emission trends in the electromagnetic environment of the medical device. The data recorded in the system included sensor data from the two probes, timestamp data, radiation values, and frequency of the instances. The 4-layered IoT architecture depicted in Figure 3 was adopted in the design of the prototype development. It includes a device, communications, data processing, and an application layer.

The device layer comprises devices, sensors, and actuators that gather radiation data from the medical device environment. The data is transmitted from the device layer to the cloud through the communication layer. The system allows data to be available on the cloud for easy access. The collected data was then retrieved using the API and processed on the application server, at the application layer. The data can be used for decision-making in the monitoring of medical device radiation and is accessible to end-users through a web interface and mobile application.

Sensor measurements were subjected to filtering, and extraneous disturbances and noise inherent to the true values were removed. Five simple filtering techniques were compared to determine the best filtering technique. The simple moving average (SMA) [42], the exponential moving average (EMA) [43], the median filter [44], the low-pass filter [45], and the Savitzky–Golay filter [46] were explored. To demonstrate the operation of filtering techniques, a window size of 3 was used for SMA, a 3-period EMA, a kernel size of 3 for the median filter, a third-order low-pass filter, a window length of 5, and an order of 2 for the Savitzky–Golay filter were utilized. A dataset of 5000 occurrences was used to evaluate the filtering techniques. The outcome of the filtering techniques and the metrics used to evaluate their performance include the Mean Squared Error (MSE) and the Signal-to-Noise Ratio (SNR). The MSE measures the average of the squares of errors between the desired and the actual signal that is input to the filter. SNR is a comparison and quantification of the level of the desired signal against the background noise. When the SNR is high, it indicates that the signal is strong and stands out clearly from the noise.

3. Results

The experimental findings on the architecture of the IEMCM system are organized into four levels: the selection of medical devices, the design of the EM radiation Detection System, the Monitoring and Control Center, and Data Visualization.

3.1. Results for the Selection of Medical Devices

For the hospital considered in this study, we selected ECG and ultrasound machines for monitoring due to their availability and known susceptibility to EMI. Another factor considered was their crucial role in patient diagnosis and monitoring of the condition.

3.2. Results for the EM Radiation Detection System

The IEMCM architecture prototype, shown in Figure 4, was tested by monitoring and collecting data over three months at the hospital. The prototype includes the sensors, the microcontroller, the communication module, and a power bank as an alternative power source. A sampling frequency of 5 s was employed, and a total of 16,8000 readings were recorded. The system demonstrated the ability to detect electromagnetic field intensities reliably and notify the biomedical engineer when the received data exceeds the recommended emission values established by IEC standards and the RSB. The system lasted 24 h without recharging, meaning that it can continuously monitor EM signals throughout the day without human intervention. The system development and user training have a one-off investment cost of $2000. The cost of hiring a biomedical engineer to measure and interpret results is $7000 per session. Medical equipment in hospitals can cost up to $8000 per minute [47].

3.3. Results for the Monitoring and Control Center

It incorporated a cloud dashboard for data storage and real-time visualization of radiation data. As shown in Figure 5, a representation of the collected radiation data can be monitored online in real-time when the data is sent to the cloud. MySQL was used in the design and development of a database to effectively store and manage data, guaranteeing scalability and safe access to large datasets. To access data through the website and mobile application, users must log in, and based on their allocated privileges, different levels of access and functionality are granted.

3.4. Results for Data Visualization

A website and a mobile application were available for hourly, daily, and monthly representations of the EMC of medical devices. Access to a more detailed view of the monitoring data requires authentication. However, a general view of daily recordings can be accessed through the website. This enables remote monitoring of the radiation levels. Data can also be analyzed to assess radiation levels in the vicinity of medical devices, providing insight to monitor EMC. Figure 6 provides a graphical representation of the sample daily average radiation recorded from 15 January 2024 to 8 March 2024.

The mean recorded radiation value for the period was 1.0028 V/m, with a minimum of 0.7228 V/m, and a maximum of 1.3515 V/m. The average radiation recorded on 19 January 2024 was 0.97 V/m, and the standard deviation was 0.76 V/m, as shown in Figure 7. The data represented was collected from 10:00 a.m. to 3:00 p.m. The recorded values were within the standard compliance range for EMC in medical devices.

The output of the five filtering techniques is shown in Figure 8. The raw data displays significant variability with sharp spikes, making it noisy and hard to analyze. The SMA smooths out fluctuations and reveals larger patterns by reducing some of the noise, though it misses a few distinct peaks. The EMA plot shows a quicker response to changes because it emphasizes recent observations. The median filter removed sharp noise spikes and outliers while maintaining the main data structure. The low-pass filter effectively removed most high-frequency noise and significantly smoothed the data compared to the SMA or EMA. The Savitzky–Golay filter smooths the data while keeping its shape.

Table 1. Evaluation of the filtering techniques.

Filtering Technique	Mean Squared Error (V 2)	Signal-to-Noise Ratio (dB)
Simple Moving Average (SMA)	0.385122	6.940508
Exponential Moving Average (EMA)	0.189109	9.995846
Median Filter	0.349088	7.684709
Low-Pass Filter	0.320519	7.731447
Savitzky–Golay Filter	0.235142	9.308036

represents the results of the evaluation of filtering techniques. Two metrics were used to evaluate the filtering techniques, and the Savitzky–Golay filter was chosen based on a small MSE, a high SNR, and the ability to smooth the signal while maintaining its shape.

3.5. Overall Performance and Findings

A comparison between our work and previous work is summarized in

Table 2. Comparison between previous works and the proposed architecture. The symbol ✔ indicates availability of a functionality and ⮿ represents the absence of the functionality.

Monitoring System Ref	Application	Operating System	Remote Monitoring	Cloud Storage	Wireless Transfer	Remote Notification	Portability	Real-Time
[2]	Environmental Radiation monitoring	Linux	✔	✔	✔	⮿	✔	⮿
[3]	High voltage Transformers	Dedicated software	⮿	⮿	⮿	⮿	⮿	⮿
[21]	Cellular Base Stations	Dedicated software	⮿	⮿	⮿	✔	⮿	✔
[4]	Environmental Radiation monitoring	Dedicated software	✔	✔	✔	⮿	✔	✔
[5]	Environmental Radiation monitoring	Any	✔	✔	✔	⮿	✔	✔
[6]	Computer Generated EM radiation detector	Any	✔	✔	✔	⮿	✔	✔
[7]	Environmental Radiation detector	Any	✔	✔	✔	⮿	✔	✔
This work (IEMCM)	Medical equipment	Linux	✔	✔	✔	✔	✔	✔

. Compared to the old frameworks where a computer with dedicated software was required to interpret and visualize the recorded data [47], the proposed approach introduces greater flexibility. The wireless transmission of data from IoT devices enables real-time radiation detection and brings adaptability and portability [48]. Traditional methods require an expert to manually measure electromagnetic fields, which is infeasible for continuous monitoring [49]. When thresholds are exceeded, biomedical personnel will be notified via a short message service; other previously published works used a buzzer to alert personnel. Although the potential for IoT utilization for radiation monitoring is evident, practical implementations and EMC handling in medical equipment are understudied in Rwanda.

Implementing the system using durable components like Raspberry Pi, well-calibrated sensors, and a reliable Internet connection ensures the system’s long-term reliability. Regular system inspections are conducted to verify sensor calibration, assess battery health, ensure reliable Internet connectivity for data transmission, and confirm the accuracy and completeness of the transmitted data. Potential enhancements to the IEMCM system architecture include its integration with hospital management systems and extension of sensor arrays, such as temperature and humidity sensing. Monitoring electromagnetic field levels and alerting engineers in real-time when thresholds are exceeded ensures medical devices are operating within safe thresholds. This assures accurate diagnoses and has a direct impact on patient outcomes. It also ensures that radiation is within the limits, implying patient medical staff protection from excessive exposure. Continuous monitoring of electromagnetic fields enables early detection of abnormal operating conditions that could degrade or shorten the useful life of the device. It ensures the ongoing compliance of medical equipment with regulatory requirements.

4. Discussion

At the prototype level, the IEMCM system architecture demonstrated its capacity to provide accurate real-time monitoring of electromagnetic fields in the vicinity of medical devices. It allows users to select measurements depending on the specific device under surveillance. Compared to the usual handheld use of RF meters, the system enables remote monitoring of EM signals. Although we considered a few medical devices, the recorded results align with the strict values of the electric field established in the standards [5]. We conducted a three-month study in Muhima District Hospital, located in Kigali, the capital city of Rwanda, which was selected after conducting a nationwide survey in referral hospitals. For the data and results in this manuscript, Muhima Hospital is the representative of other district hospitals at the same level. For device validation, more referral hospitals will be considered for the data collection, storage, and analysis based on big data. The proposed architecture of the IEMCM system is an advancement of an IoT-based system developed by Tocchi et al.: they used a BeagleBone Black microcontroller and stored data on a central server [24]. Our design stores data in the Internet cloud, making it accessible from anywhere and at any time. We used an SMS to notify the biomedical engineer, unlike prior research that used a buzzer when the threshold was surpassed [9]. Utilizing a single-board computer minimizes the system’s size, enabling remote continuous monitoring of electromagnetic signals in sensitive environments like hospitals. In most previous works, data processing was performed using a personal computer, which limits mobility and remote monitoring [50]. Data processing was also performed on the PC using specific software with the measuring device. In this study, the Raspberry Pi was interfaced with the RF meter via Minicom, allowing easy access and data communication. The proposed architecture sends all the collected radiation data to the cloud.

The system operates on the power bank’s rechargeable battery, charging on the rectified AC power supply. It can operate for 24 h without requiring recharging. This implies that it can function for an entire day without the need for human intervention, even when electrical disturbances are encountered. It also highlights the potential for remote monitoring. To minimize Internet connectivity disruptions and ensure reliable data transmission, the system was configured to support dual communication modes, employing both the GSM HAT and the Wi-Fi network. This enables the system to seamlessly switch between the two communication channels whenever reduced bandwidth is detected on either network.

To minimize Internet connectivity disruptions and ensure reliable data transmission, the system was configured to support dual communication modes, employing both the GSM HAT and the Wi-Fi network. This enables the system to seamlessly switch between the two communication channels whenever reduced bandwidth is detected on either network. The proposed system offers a financially sustainable alternative to traditional assessment methods. With a one-time investment of $2000, covering sensing, processing, and communication components, the system substantially reduces the recurring costs of hiring experts for professional EMC assessments by 70% of the estimated $7000 per session. The system cost is only 25% of the medical equipment downtime, estimated at $8000 per minute.

The exploration of filtering techniques was evaluated using the MSE and SNR. The SMA, with an MSE of 0.385 and SNR of 6.94, reduces noise but introduces lag, blurring sharp transitions in the data. The EMA, with an MSE of 0.189 and SNR of 9.996, offers a quicker response by emphasizing recent data. However, it still smooths over finer details. The median filter, with an MSE of 0.349 and SNR of 7.684, removes short-duration noise while preserving edges, making it ideal for data with sudden spikes. The low-pass filter, with an MSE of 0.321 and SNR of 7.731, cleans the signal by removing high-frequency noise, but may also oversmooth important details. The Savitzky–Golay Filter, with an MSE of 0.235 and SNR of 9.308, is preferred in this case study. Although it may not be suitable for other types of EM signals, it was selected because it smoothens the data to reduce noise while preserving the trend. Despite its limitations in low-frequency noise suppression, window size sensitivity, and computational demands, the Savitzky–Golay filter effectively preserves signal characteristics. This makes it well-suited for processing electromagnetic signals from medical devices with minimal distortion compared to other filtering techniques [46].

Cloud-based storage facilitates rapid data visualization, complemented by a web and mobile application tailored for biomedical engineers and technicians. Access control measures were implemented based on user roles and authentication credentials. Biomedical engineers and technicians were granted full access to all the radiation data, along with the capability to analyze it. Hospital management would be granted access solely for visualizing the radiation distribution [51]. This will enable them to make timely decisions on medical device management. The data collected includes GPS coordinates for identifying the location of the devices under surveillance. This is quite useful, especially since the objective of the work is to monitor several hospitals concurrently in the long run. The EMC of medical devices can be influenced by nearby electronic devices and hospital communication systems, which emit electromagnetic waves. In this operational hospital, safety inspections were conducted, and the medical device under surveillance was isolated in the room. The EMC can also be influenced by temperature and humidity; however, we did not include all these factors since we focused on developing an embedded system architecture. The proposed IEMCM system is currently at the Technology Readiness Level 8. The system has been fully developed and tested and is ready for implementation in various hospitals. The next steps needed to advance it toward real hospital deployment are seeking funding and marketing the system. In future work, we will consider other factors that influence EMC. The work enables remote electromagnetic radiation monitoring by integrating IoT and already calibrated radiation sensors.

The proposed architecture reduced costs by minimizing reliance on expert EMC assessments, while enabling real-time radiation monitoring and diagnostics, thereby lowering the likelihood of EMI occurrences. An average of 1.0028 V/m indicates that the radiation values are within the acceptable ranges. The ability to track the levels of emissions around medical devices helps ensure the safety of patients and healthcare workers. The maximum recorded value was 1.3515 V/m, which is below the thresholds set for the medical devices. This ensures optimal device performance, making it practical for remote monitoring and reducing reliance on on-site experts. The IEMCM architecture also enables long-term monitoring, allowing big-data collection that can be used for the prediction of the future EMC of medical devices. The study’s scope was limited to a single hospital and two categories of medical devices. While the choice of the hospital was influenced by its location and similarity to other district hospitals, this may constrain the generalizability of the findings. Hospital infrastructures may vary in terms of equipment density and operational protocols; therefore, results from one facility may not fully represent the settings of other hospitals. Similarly, monitoring two device categories restricts the diversity of electromagnetic emission and susceptibility characteristics considered and may not be representative of all the devices in the hospital.

5. Conclusions

This study explored the design of an IoT-based embedded system architecture to detect and monitor electromagnetic field intensities around medical devices to verify EMC. A Lutron EMF-839 RF meter equipped with two probes for low and high frequencies was connected to the Raspberry Pi 4 microcontroller as a sensor. A GSM module equipped with GPS to determine location coordinates and GPRS to send data to the cloud was used for communication. The data was collected continuously over three months to evaluate the IEMCM system architecture’s performance in a hospital in Rwanda. The proposed architecture detects and monitors the EMC in real-time and transmits data to the cloud for storage and further analysis. The SMS notification mode is triggered when radiation levels exceed the IEC thresholds of 3 V/m for general medical devices and 10 V/m for critical care equipment, thereby indicating instances of electromagnetic incompatibility. The system operates autonomously for 24 h without recharging, demonstrating its suitability for remote monitoring. Of the five filters explored, the Savitzky–Golay was selected to smooth the collected data to enhance the visualization of the sensor data. The proposed IEMCM architecture will improve hospital safety by verifying the efficacy of the medical devices. In the next publication, a machine learning approach will analyze the radiation data collected to predict future EMC levels for medical devices and automated compliance reporting. Continuous monitoring will also be expanded to include more medical devices in multiple hospitals.