Machine Learning-Based Wearable Devices for Smart Healthcare Application With Risk Factor Monitoring

Introduction

Stroke has been considered as one of the most dangerous causes of all-cause mortality and morbidity throughout the world. Although the world has improved the vascular risk factor management, stroke systems of care, which has helped to mitigate the effects of stroke from number three cause to number five cause, but it is still a dangerous disease for human life (Fugate et al., 2014). In a survey and intensive investigation, 30% of stroke cases remains unsolvable and hidden due to lack of awareness (Bilal et al., 2020; Garkowski et al., 2015). To investigate and determine the correct disease on time, there requires a complex, critical, and detailed examination under micro-observation during clinical phase which can then detect the systematic symptoms of stroke (Bersano et al., 2020; Shulman & Cervantes-Arslanian, 2019). Such detailed observation become more complicated by the involvement phenotype heterogeneity as it is observed the stroke can also be primary cause for a particular body. When the blood supply disturbance happened in the brain then stroke occurs. It is a dangerous disease because it is hard to predict the precise time of occurrence. The result of brain stroke is oxygen decrease in brain cells that leads to death of brain cell (Hong et al., 2013). Some of the symptoms of stroke are bleeds, clots, and transient ischemic attack. The most common symptoms are arm weakness, difficult to speak, difficult to walk, blurred vision, tiredness, memory loss, stiffness in muscles and co-ordination issue (Subramaniyam et al., 2017) (World Health Organization, 2004). The success of treatment says that it should be treated within a few hours (4.5 hours) of occurrence to prevent the long-term damage from stroke. The development in imaging technique suggests that whether the stroke initiated within 4.5 hours. Further, the lesion data extracted from MRI image helps to predict the patients' speech production skills (Hope et al., 2013).

Wearable technology supports brain stroke prediction because it performs real-time monitoring of stroke-related physiological parameters. Based on the wearable characteristics, such as weight, accessibility, frequency of use, data continuity, and response time, the wearables have margins in reporting high-precision prediction outcomes. The trend of integration of wearables into the internet of things (IoT), electronic health records (EHRs) and machine learning (ML) algorithms to launch a stroke risk prediction system (Chen & Sawan, 2021; Usmani et al., 2021). Moreover, the risk factors that is measured with the help smart devices or wearable devices helped to observe the health of patient. While histories of other diseases with these wearable devices can also be maintained to further treatment, medication, and other medical purposes in different available EHRs (Sirsat et al., 2020; Tayyab, Marjani, Jhanjhi, & Hashem, 2021). Due to abundant real-time dataset that is recorded using machines and different EHRs, ML techniques (Tayyab, Marjani, Jhanjhi, Hashim et al, 2021) have been used to analyze the association and order them in specific risk factor to predict the correct brain stroke rate (Goldstein et al., 2017; Vashistha et al., 2019).