In this work, we design a multisensory IoT-based online vitals monitor (hereinafter referred to as the VITALS) to sense four bedside physiological parameters including pulse (heart) rate, body temperature, blood pressure, and peripheral oxygen saturation. Then, the proposed system constantly transfers these signals to the analytics system which aids in enhancing diagnostics at an earlier stage as well as monitoring after recovery. The core hardware of the VITALS includes commercial off-the-shelf sensing devices/medical equipment, a powerful microcontroller, a reliable wireless communication module, and a big data analytics system. It extracts human vital signs in a pre-programmed interval of 30 min and sends them to big data analytics system through the WiFi module for further analysis. We use Apache Kafka (to gather live data streams from connected sensors), Apache Spark (to categorize the patient vitals and notify the medical professionals while identifying abnormalities in physiological parameters), Hadoop Distributed File System (HDFS) (to archive data streams for further analysis and long-term storage), Spark SQL, Hive and Matplotlib (to support caregivers to access/visualize appropriate information from collected data streams and to explore/understand the health status of the individuals). In addition, we develop a mobile application to send statistical graphs to doctors and patients to enable them to monitor health conditions remotely. Our proposed system is implemented on three patients for 7 days to check the effectiveness of sensing, data processing, and data transmission mechanisms. To validate the system accuracy, we compare the data values collected from established sensors with the measured readouts using a commercial healthcare monitor, the Welch Allyn® Spot Check. Our proposed system provides improved care solutions, especially for those whose access to care services is limited.
By digitizing the physical realm, the Internet of Things (IoT) technology enables substantial societal benefits and magnifies the scope of real-time monitoring systems. Recently, the wide adoption of IoT-enabled applications has amplified the explosion of real-time monitoring systems for different domains including home automation [
To leverage the benefits of IoT technology, healthcare organizations must develop a system where they can measure, process, and assess big clinical data in a profitable and scalable way. Big data analytics enable the medical sector to reshape its domain efficiently. On the top-end, these organizations can employ data analytic systems to make actionable verdicts from enormous data and support medical staff to save lives by making early clinical interventions [
Systematic vital signals monitoring is an integral part of patient care, which targets to expedite the early detection of abnormal parameters of worsening patients. Instability in vitals is an extremely sensitive predictor of clinical deterioration and imminent adversarial happenings including heart attack or even death [
The major objective of this work is to develop a multisensory online vitals monitor for the remote patient monitoring system. The key contributions of the manuscript are four-fold. We identify the drawbacks of traditional vital sign monitoring systems and their inappropriateness in remote health services. We explore the necessity for developing real-time physiological signs monitoring system and decision making by integrating IoT and big data analytics technologies to provide a reliable, real-time, and transparent interconnection among stakeholder components of the proposed system. We implement our system on three patients for 7 days to check the effectiveness of sensing, data processing, and data transmission mechanisms. To validate the accuracy of the proposed system, the outputs of the system are related to the sensed data values through a commercial healthcare monitor, the Welch Allyn® Spot Check. We develop a mobile application to send the measure parameters and statistical graphs to doctors and patients in order to enable them to track health conditions remotely.
The article is framed as follows. The following Section provides existing healthcare systems aiming to provide continuous remote healthcare monitoring and classify patient data. In Section 3, we present the architecture of the proposed system VITALS in detail. Section 4 describes the implementation details of our model and evaluates the results obtained from a real-time scenario. In Section 5, we conclude this paper.
Of late, adopting IoT-sensor technologies and big data analytics to monitor patient health has increased exponentially. Several IoT-based health monitoring applications have already been commercialized and existing in the market owing to the aptitude of these systems to deliver fast, secure, and lucrative solutions [
Mohammed et al. developed an IoT-enabled healthcare monitoring system using suitable sensors (for measuring ECG, BT, PR, and SPo2), the MySignals development shield, and a low-power long-range (LoRa) communication technology [
Cloud computing is employed as an enabling technology that allows IoT systems to deliver reliable and easier communication between multiple sensors/devices and stakeholders of the healthcare system. Al-Kababji developed an IoT-enabled fall detection and ECG tracking mechanism using cloud computing and a mobile application [
The abovementioned works reveal that recently, numerous architectures, models, and approaches have been emerged and implemented to deliver better-quality care for individuals. Conversely, very few dependable models have developed that can successfully implement a holistically real-time, lucrative, and intelligent method in the medical industry. Some extant healthcare systems focused on how diverse smart things efficiently assimilating, while others focused on the security of patient data. Real-time data processing and prediction are the most important endeavors in healthcare ecosystems, particularly in emergency care units [
The proposed monitoring system is intended to measure four bedside vital signals using different biosensors. For this purpose, we develop a working prototype model with appropriate sensors to measure PR, BT, BP, and SPo2. Once the vital parameters are measured, they are transmitted to data analytics tools using a WiFi module for further analysis or long-term storage. The overall architecture of VITALS is shown in
Our proposed system uses an AVR-IoT WG development board (AC164160) as shown in
In this work, the AVR-IoT WG is employed to collect vitals from sensors including MCP9808 (to measure body temperature), MAX30100 (to measure SPo2), and medical equipment called Healthgenie BPM01W (to measure blood pressure and pulse rate). Our proposed system captures patient physiological signals as “live data streams” using multisensory maneuvers and other medical apparatus.
The AC164160 employs the divide and conquer technique through smart elements such as a powerful microcontroller, a secure cryptographic coprocessor chip, and a WiFi communication module to decrease the complexity of the algorithm implementation. The ATECC608A chip in this board is used to store private keys, authenticate the firmware, and provide a secure boot process for the connected maneuvers. This coprocessor chip engenders both the public and private keys by a random number generator and enables devices to create secure communication. The ATWINC1510 WiFi communication module is particularly designed for low-power sensing applications. It has an option of an embedded antenna or a micro coaxial connector for an external antenna. This WiFi unit also assimilates a power amplifier, low-noise amplifier, switch, and power controlling module, which leads to a compact structural design.
The sensing elements and other maneuvers are implemented and controlled through embedded C programming codes. The sensors and Healthgenie BPM01W equipment capture different physiological parameters and transmit them to the analog to digital converter to obtain digital signals. Nonetheless, data collected from the related biosensors/equipment are vulnerable to loss before ingesting by Kafka due to (i) obstacles and distance between connected sensors and the master node; (ii) congestion in the overwhelmed network setting; and (iii) failure in the sensors itself. In this situation, the caregivers cannot make the correct decision about the patient health status or store data for further analysis. In order to prevent this missing data problem, data preprocessing methods must be employed before making any decision or storing data. Hence, these signals are preprocessed by the ATmega4808 microcontroller and sent to the big data analytics system for analysis and visualization through a selected WiFi module as shown in
The big data analytics system used in VITALS contains Apache Kafka to gather live data streams from biosensors every 30 min, Apache Spark to categorize the vital signs and notify the medical professionals while detecting abnormalities in patient vitals, HDFS to store data streams for future analysis. For information retrieval and visualization, VITALS employs Spark SQL and Hive to explore and understand the health status of the individuals and Matplotlib to visualize the results. In addition, we develop a mobile application called VITALS to receive and display patients’ vital signals on a smartphone platform in a text and graphical form. In this work, we develop a working prototype model with appropriate sensors to measure the vitals of the patient remotely as given in
Vital signals reflect the operation of the body’s homeostatic mechanisms. Monitoring and inferring the physiological signals are significant tasks of a healthcare system that can provide knowledge about the basic health condition of the individuals. Also, they are of paramount significance in defining treatment and triage. Indeed, vitals act an important role in calculating medical deterioration in critical care. The rate of vitals anomalies reflect the persistent patient condition, frequency of readmission to clinics, return emergency room visits, and exploitation of healthcare assets in the hospitals. Body temperature, pulse rate, blood pressure, and oxygen saturation are standard vital parameters to reflect the status of the life-sustaining functions and severity of the disease. Mostly, vitals vary with age, gender, body mass index, fitness, and overall health.
Life span | Body temperature (°C) | Oxygen saturation |
Pulse rate (bpm) | Blood pressure | |
---|---|---|---|---|---|
Diastolic (mmHg) | Systolic (mmHg) | ||||
Older adult (>70 years) | 35–37.2 | 95–100 | 60–100 | 60–80 | 90–120 |
Adult (> 19 years) | 36.5–37.2 | 96–100 | 60–100 | 60–80 | 90–120 |
Adolescent (≤ 19 years) | 36.5–37.2 | 96–100 | 60– 90 | 62–80 | 94–120 |
School-age (6 – 12 years) | 36.6–37 | 97–100 | 75–110 | 54–80 | 84–120 |
Preschooler (3 – 5 years) | 37–37.2 | 98–100 | 80–120 | 50–78 | 82–110 |
Toddler (1– 2 years) | 37.2–37.6 | 98–100 | 80–130 | 50–80 | 80–112 |
Infant (2 months – 1 year) | 37.4–37.6 | 98–100 | 80–160 | 50–70 | 74–100 |
Neonate (0–2 months) | 35.3–37.5 | 98–100 | 70–190 | 20–60 | 60–90 |
The selection of biosensors and medical equipment for capturing patient vitals hinges on the reliability, availability, affordability, and compatibility of the sensors/devices with AVR-IoT WG and ATmega4808 microcontroller. The designated devices are connected to the development board using their appropriate interface units. Each sensing node collects and pre-processes the data and transfers it to the big data analytics system. The sensitivity and ruggedness of the sensors also act an important role in developing VITALS. A small variation in the readout of the sensor will modify the implication of the vital signals. For a reliable analysis, high-quality off-the-shelf sensors are designated.
In this framework, two sensors, MCP9808, MAX30100, and non-smart medical equipment Healthgenie BPM01W are selected for measuring BT, SPo2, BP, and PR to provide a common interpretation of patient health condition but can be scaled up in the context of the number of sensors if needed. The ATWINC1510 WiFi unit transmits the measured parameters, date, and time to a master node directly. ATmega4808 is used to collect and pre-process the sensed signals. A compact ATECC608A secure element is used as standby storage for captured parameters. It uses a battery to energize the sensors and other devices in VITALS. The specification of these devices is given in
Sensor/medical equipment | Measuring parameters | Range | Accuracy | Response time |
---|---|---|---|---|
MCP9808 | Body temperature | –40 to 125°C | ± 0.5°C (25°C) | ≤ 1 min |
MAX30100 | Oxygen saturation | 60 to 100% | ± 0.1% (25°C) | < 5 s |
Healthgenie BPM01W | Pulse rate | 30–180 bpm | ± 5% (25°C) | < 30 s |
Blood pressure | 0–299 mmHg | ± 3 mmHg (25°C) | <1 min |
The core temperature of the body remains constant except the user develops a febrile illness. The standard BT of humans relies on various factors including age, gender, ambient temperature, time of day, exercise, hot or cold drinks consumption, eating habits, etc. For example, the normal BT can range from 97.8°F (36.5°C) to 99°F (37.2°C) for a healthy adult. BT may be anomalous owing to hypothermia (<95°F) or fever (>98.6°F). In order to measure patient BT, we use an integral MCP9808 digital temperature sensor in the development board which is given in
The blood oxygen saturation level of individuals is a critical parameter for predicting the improvement and severity of illness. SPo2 indicates the peripheral saturation of hemoglobin by oxygen. It reflects the general health status of the individuals. SPo2 of a normal healthy adult is 96%–100%. It drops if somebody has a respiratory disease or any other sickness [
Blood pressure (BP) values are the predominant determining factor of therapeutic decisions as it specifies blood flow when the heart is relaxing (diastole) and contracting (systole). It is impacted by cardiac output, the volume of blood, peripheral vascular resistance, and thickness and elasticity of the vessel wall. Trends or variations in BP values reveal primary pathophysiology or the body’s efforts to sustain homeostasis. For example, a reduction in BP is a common indication in patients prior to heart attacks [
Pulse rate is the count of heart beats per minute (bpm). In order to collect the value PR of an individual, most medical devices use the volume of blood flow. Typical PR ranges from 60 to 100 bpm for a healthy adult. The normal relaxing PR for adult females 75 bpm and males is approximately 70 bpm [
The big data analytics used in VITALS is implemented on top of the Apache Spark streaming platform. VITALS consists of the following data analytic tools: (i) Apache Kafka to gather live streams from connected biosensors; (ii) Apache spark to classify the patient data and send an alert to the healthcare professionals while detecting abnormalities in patient physiological parameters; (iii) HDFS to store data streams for future analysis, and (iv) two information retrieval tools (Spark SQL and Hive) and one graphing tool (Matplotlib) to enable medical staff to access/visualize medical records of patients and to analyze/understand the health status of the patient.
Our system employs Apache Kafka to collect data streams from patient vitals such as BT, PR, BP, and SPo2. Generally, Kafka is working on the idea of “topics” and input streams are stored as keys. We install Kafka on the master node and create a topic, called “Patient_Vitals”, to gather parameters from biosensors then transmit them to a Spark streaming for data processing. In this work, after receiving data streams from Kafka, Spark streaming explore them in real-time and send warning messages to the medical professionals during an emergency case detection. For this purpose, we develop a risk prediction and recommendation module that engenders an alert when an abnormal vital signal is detected. Also, it endorses an apt action that should be taken by the medical staff whenever the patient vitals deviate from the threshold values.
The real-time prediction process in VITALS includes the following modules: the learning module and the deployed model. The learning module accepts the input batches to train the model and directs the training sequences to the deployed model to learn and generate the result. The learning module receives the batch result and then prefers one-to-one analysis from the training sequences and calculates the score (weight) for new data. This model implements the learning process constantly and it updates parameters for each result, which is almost “learning-on-the-fly”. It supports envisaging differences in distribution rapidly and increases the accuracy in many cases. For each vital sign data collected from the connected sensors is related to early warning score between 0 and 3 where 0 denotes the score for normal health status where other values signify the anomaly. Therefore, the severity of the disease is increasing with the score as given in
Vitals | 3 | 2 | 1 | 0 | 1 | 2 | 3 |
---|---|---|---|---|---|---|---|
BT (°C) | <35 | 35.1–36 | 36.1–37.9 | 38–39 | ≥39.1 | ||
SPo2 (%) | <85 | 85–92 | >92 | ||||
PR (Bpm) | <40 | 41–50 | 51–100 | 101–110 | 111–130 | >130 | |
BP (mmHg) | <70 | 71–90 | 91–170 | 171–200 | >200 |
For example, the data gathered from the connected sensors are 38°C (BT), 90% (SPo2), 48 (PR), and 120 (Systolic BP), then the calculated score set is {1, 2, 1, 0}. After computing the score set for every time instance, Apache Stream evaluates the score and directs an alert to the medical professionals with proper recommendations in emergency cases. In our work, we employ the clinical responses guide, National Early Warning Score (NEWS) developed in [
Cumulative score | Criticality level |
---|---|
≥7 | High |
≥5 or 3 in one parameter | Medium |
1–4 | Low |
0 | None |
For the aforesaid example, the cumulative score is 4. Now the master node creates a notification to medical staff alerting that the frequency of surveillance of the vital signs should be once in 4 to 6 h. Hence, it is mainly imperative to consider NEWS for giving appropriate recommendations and enable healthcare workers to monitor changes in bedside signals and consequently the early predictions of patient deterioration.
The VITALS uses two information retrieval tools (Spark SQL and Hive) and one graphing tool (Matplotlib) to enable the medical staff to obtain the patient vitals from HDFS. Hive is used for data warehousing to process queries and analyze big datasets archived in HDFS. In addition, it allows customers to generate metadata storage with tabular forms or views in a relational database. This makes our system more efficient by reducing the access time. Moreover, it supports the practitioners to compute the criticality level of diseases. The VITALS installs Hive on the driver and uses a table in the main directory of HDFS. Then, the doctors can search the patient clinical data through the HiveQL console. SparkSQL is used to realize a data abstraction called DataFrames for structured data processing. It enables SparkSQL to use schema and it is realized by domain-specific language. Matplotlib is a comprehensive library in Python and uses its mathematical extension NumPy to create cooperative, active, and static visualizations. Our VITALS enable us to write a Python script to access data stored in HDFS periodically and to visualize them using Matplotlib. We develop a mobile application to send statistical graphs to doctors and patients to enable them to track the health conditions remotely.
The proposed healthcare monitoring framework is built over the Apache Spark version 2.3.1 which involves one driver node and three slaves. Our system employed Ubuntu 14.04 virtual machines to create the clusters. The master, as well as executor nodes, contains quad cores, 16 GB of RAM, and 100 GB disk storage. We developed a working prototype model with appropriate sensors. The proposed risk prediction and recommendation module is implemented to identify the criticality level of the health condition of the patient and send an alert to the medical professionals with suitable recommendations.
The accuracy of the sensors for BT, SPo2, BP, and PR are also assessed by relating their observed data values to that of the Welch Allyn® Spot Check which is commercial multi-parameter vital signs monitor given in
P.ID | Body temperature | Oxygen saturation | ||||
---|---|---|---|---|---|---|
VITALS | Welch Allyn® Spot Check | Diff | VITALS | Welch Allyn® Spot Check | Diff | |
VITALS 001 | 36.4 | 36.1 | 0.30 | 99.6 | 99 | 0.6 |
VITALS 002 | 37.5 | 36.8 | 0.70 | 97.8 | 98 | −0.2 |
VITALS 003 | 36.7 | 37.2 | −0.50 | 98.3 | 98 | 0.3 |
Average Difference | 0.17 | Average Difference | 0.23 |
It can also be observed in
P.ID | Blood pressure | Pulse rate | ||||
---|---|---|---|---|---|---|
VITALS | Welch Allyn® Spot Check | Diff | VITALS | Welch Allyn® Spot Check | Diff | |
VITALS 001 | 180/110 | 178/107 | 2/3 | 116 | 114 | 2.00 |
VITALS 002 | 185/93 | 184/91 | 1/2 | 119 | 120 | –1.00 |
VITALS 003 | 127/84 | 128/88 | –1/–4 | 108 | 105 | 3.00 |
Average Difference | 0.67/0.33 | Average Difference | 0.33 |
We developed a mobile application for patients and doctors (who use smartphones) named VITALS. This app is created on Android OS by means of the Java language. The app enables patients to visualize their measured physiological parameters with time stamps. It contains a patient module and a physician module. The patient’s account is created by requesting the patients to enter their details such as name, email-id, gender, age, address, and contact details. Each new user registering on our app will be provided with a user id and password. The complete system workflow is that patients register themselves on the system and enter their details such as their user id and password to access their vital signals. Doctors register themselves on the portal and access patient vitals.
The authorized medical professionals can access patient’s data through this application. They can select a particular patient (refer to
The integration of IoT, big data analytics system, and mobile applications is a predominant approach in the real-time healthcare monitoring system. This convergence is intended to decrease the total medical cost and increase the quality of care delivery to individuals, especially patients. In this work, we design a multisensory IoT-based real-time vitals monitor to sense BT, SPo2, BP, and PR and constantly transfer these signals to the big data analytics system which aids in enhancing diagnostics at an earlier stage. For this purpose, we use the AVR-IoT WG development board to collect vitals from sensors including MCP9808 (to measure BT), MAX30100 (to measure SPo2), and medical equipment called Healthgenie BPM01W (to measure BP and PR). The developed kit extracts vital signs in a 30 min interval and sends them to the big data analytics system through the WiFi module for further analysis. We use big data analytic tools including Apache Kafka, Apache Spark HDFS, Spark SQL, Hive, and Matplotlib. In addition, we develop a mobile application to send measured data with an overall health condition to the patients and doctors. To validate the accuracy of the system, we implement our system on three patients for 7 days. We compare the data values collected from established sensors with the measured parameters using the Welch Allyn® Spot Check. The VITALS provides improved healthcare facilities to patients, especially for those whose access to care services remotely.
We plan to extend the scope of the application of VITALS to (i) generate an automatic notification to ambulance, family, or friends. The alert will specify the criticality level and the GPS position of the patient to rush an ambulance from a nearby hospital to the patient location. The ambulance will exploit the GPS coordinates to get to the particular location hastily and concurrently tracks the vitals and conveys them to the concerned hospital; (ii) add a module for medicine dispensing system to send alarms to the patient to remind him/her of the scheduled timely medication/injections and out of schedule medicine dosages; (iii) design wearable system (e.g., accelerometer sensor) to monitor potentially infected Covid-19 patient and send an alert to the concerned people in case of emergency and based on violation of self-quarantine regulations; and (iv) send doctor prescriptions to selected pharmacies so that the patients can get the medicine delivered to their doorstep.