The Internet of Things (IoT) is one of the key drivers of extensive automation of the healthcare sector. IoT offers a seamless platform to enable interactions between humans and different virtual and physical things, including personalized healthcare domains. Recent advances in the IoT have further improved aspects like patient care and monitoring in the healthcare sector. This article explores how IoT transforms healthcare with machine learning enhancing its applications.
Image Credit: Sergey Nivens/Shutterstock
Introduction to IoT in Healthcare
In recent years, the IoT has played an important role in enabling automation in diverse fields like smart and remote healthcare systems. The branch of IoT dedicated to healthcare is also known as healthcare IoT (HIoT). IoT is increasingly employed as a critical technology in health management systems with the rising focus on making healthcare more cost-effective, proactive, and personalized.
Personalized healthcare allows remote monitoring of patients, which enables early detection, diagnosis, and prevention of diseases, specifically chronic diseases like heart disease, arthritis, cancer, obstructive pulmonary disease, and diabetes. Additionally, the growing elderly population with chronic ailments and their needs for remote monitoring, rising medical costs, increasing focus on telemedicine in developing countries, and lack of access to healthcare resources are also increasing the importance of IoT in healthcare systems.
IoT can reduce the strain on sanitary systems and provide tailored health services to enhance the quality of life. This technology comprehensively redefines people, applications, and devices in the healthcare domain. For instance, IoT enables health monitoring anywhere and anytime around the human body using wireless body area network technology, prevents hospital infections, and assists in post-discharge care and emergency management.
In healthcare, the implementation of IoT is categorized into automatic data collection and sensing, people identification and authentication, and tracing individuals, medical teams, and staff. Healthcare environments can be revolutionized using IoT opportunities like the Internet of Medical Things (IoMT) technology, which involves connected special medical devices or medical sensors to ensure a tailored approach to healthcare delivery.
The application of IoMT technology in healthcare/HIoT assists in creating proper therapeutic strategies for patients by connecting several medical devices to the internet and performing different telehealth services like computer-assisted rehabilitation, teleconsultations, telemonitoring, and supervision of elderlies. Moreover, the convergence of IoT and big data in sanitary systems results in the smart management of healthcare processes. Big data analytics has enabled predictive, autonomous, and prescriptive analysis of healthcare approaches.
HIoT Architecture
Three primary requirements, including interoperability, reliability and bounded latency, and security and privacy, must be fulfilled by all HIoT frameworks for their implementation in real healthcare environments. A HIoT system contains an end-to-end network consisting of three major operational layers for data collection, data storage, and a data processing.
The data collection layer collects the medical data from different sensor devices attached to the test subject/patient that needs to be examined/monitored, while the data storage layer stores the big data obtained from different sensors and transmitted through the internet. The data processing layer analyzes the data stored in servers to generate the response using computing algorithms.
This layer also performs the visualization and compilation of results. Diverse technologies like unique identifiers, radio frequency identification, Zigbee, Bluetooth, global positioning system, accelerometers, gyroscopes, and electrocardiogram sensors are employed to implement these layers.
Additionally, dimensionality reduction algorithms like linear discriminant analysis (LDA) and principal component analysis (PCA), linear regression, logistic regression, support vector machine (SVM), k-nearest neighbors (KNN), k-means, decision tree (DT), random forest (RF), naive Bayes (NB), gradient boosting and AdaBoost, convolutional neural networks (CNN), artificial neural networks (ANN), reinforcement machine learning, and natural language processing (NLP) are the supervised and unsupervised learning algorithms commonly utilized in HIoT.
HIoT Approaches
Sensor-based approaches, resource-based approaches, communication-based approaches, application-based approaches, and security-based approaches are typically followed by IoT-based healthcare systems. Sensor-based approaches utilize wearable and environmental sensors to monitor patients' vital signs.
For instance, a cost-effective and wearable galvanic skin response system detects the level of human physiological activities of an individual by processing, acquiring, and amplifying galvanic skin response data in smart e-healthcare applications. Resource management issues are one of the challenging problems in IoT-based healthcare systems due to the resource limitations, unpredictability, resource heterogeneity, and dynamic nature of the HIoT environment.
Resource-based approaches tackle HIoT's resource limitations by employing techniques like provisioning, load balancing, offloading, allocation, and scheduling. Communication-based approaches deal with communication infrastructures responsible for managing communications.
Application-based approaches provide systems like recommender systems and monitoring systems that can perform one or multiple specific services. Recommender and monitoring systems are introduced to detect and predict anomaly situations or observe patients for a long duration and recommend proper medicine to them. Security-based approaches address aspects like trust, confidentiality, access control, and privacy to ensure a secure HIoT system and maintain quality of service (QoS).
ML in HIoT
In HIoT, ML algorithms are important in diagnosis, prognosis and spread control, assistive systems, patient monitoring, and healthcare logistics. For instance, KNN, optimal SVM, CNN, LDA, and SVM+RF are utilized for heart disease diagnosis, lung cancer diagnosis, EEG-based pathology detection, automated spike detection in MEG signals, and type-2 diabetes diagnosis, respectively.
ANN, SVM, KNN, DT, and RF are used to diagnose chronic kidney disease, while SVM and KNN are employed to develop a diagnosis chatbot. Similarly, NB, NLP, KNN, logistic regression, CNN, and DT are utilized for heart disease onset prediction in early stages, influenza virus detection, epilepsy risk level classification, hemorrhagic shock recovery prediction, COVID-19 identification, and Ebola spread control, respectively.
In patient monitoring, linear regression, NB and RF, ANN, NB, and gradient boosting are used for cuff-less blood pressure monitoring systems using smartphone cameras, continuous patient monitoring for stroke prediction, future glucose level prediction through present monitoring data, fall prediction system, and infant health monitoring system, respectively. Portable and non-invasive blood pressure monitoring, better performance, and effective prediction of emergency cases are the major advantages of using these ML algorithms for patient monitoring.
Gradient boosting predicts whether preterm infants are suffering from bradycardia to allow early intervening measures. However, these algorithms also have many drawbacks, including the need for multiple high-end smartphone cameras, the need for improved IoT-based sensor systems for autonomous data collection, and inaccurate results for sudden data changes. Moreover, the effectiveness of NB for fall prediction and gradient boosting for infant health monitoring has not been verified in a large number of patients.
Deep learning algorithms are also utilized for patient monitoring systems. For instance, a deep neural network was proposed to detect the qualities of the signal in the sensor array to understand the patient’s body conditions. While neural networks offer high accuracy and low cost, their dependence on substantial data can be a drawback.
Navigating Challenges
Scalability is a big challenge in the implementation of IoT in healthcare systems. The technology must possess the ability to meet the changing requirements and adapt to the changes on a larger scale in the future. However, many approaches in HIoT systems operate on a small scale and their effectiveness has been confirmed only by some devices. Thus, scaling up these approaches that have been mostly utilized in limited scenarios is difficult. Interoperability plays a big role in exchanging data and resources between patients and smart objects in HIoT.
The key challenge of interoperability is the development of open-source frameworks with a steady connection. A collection of standards must be established to ensure that horizontal platforms are capable of programmability, operability, and communicability among operating systems, applications, and devices. Lastly, the other major challenges are implementing adaptive and dynamic architectures that are interoperable for large-scale IoT applications and developing scalable architectures that interact with non-homogeneous data centers and smart objects.
Overall, IoT is revolutionizing healthcare by enabling remote patient monitoring, personalized medicine, and improved care delivery through interconnected medical devices and data analysis. ML algorithms further enhance HIoT applications for disease diagnosis and prediction.
References and Further Reading
Bolhasani, H., Mohseni, M., Rahmani, A. M. (2020). Deep learning applications for IoT in health care: A systematic review. Informatics in Medicine Unlocked, 23, 100550. https://doi.org/10.1016/j.imu.2021.100550
Bharadwaj, H. K., Agarwal, A., Chamola, V., Lakkaniga, N. R., Hassija, V., Guizani, M., Sikdar, B. (2021). A review on the role of machine learning in enabling IoT based healthcare applications. IEEE Access, 9, 38859-38890. https://doi.org/10.1109/ACCESS.2021.3059858
Rejeb, A., Rejeb, K., Treiblmaier, H., Appolloni, A., Alghamdi, S., Alhasawi, Y., & Iranmanesh, M. (2023). The Internet of Things (IoT) in healthcare: Taking stock and moving forward. Internet of Things, 22, 100721. https://doi.org/10.1016/j.iot.2023.100721
Haghi Kashani, M., Madanipour, M., Nikravan, M., Asghari, P., & Mahdipour, E. (2021). A systematic review of IoT in healthcare: Applications, techniques, and trends. Journal of Network and Computer Applications, 192, 103164. https://doi.org/10.1016/j.jnca.2021.103164