Optimized Blood Pressure Monitoring Using a Deep Learning-Driven Biosensor System
Amit SharmaSchool of Computer Applications Lovely Professional University, Phagwara, Punjab, India. profamitsharma@gmail.com0000-0003-1451-5892
Dr. Arun KhatriMittal School of Business, Lovely Professional University, India. arun.31886@lpu.co.in0000-0002-5895-7951
Danish KundraCentre of Research Impact and Outcome, Chitkara University, Rajpura, Punjab, India. danish.kundra.orp@chitkara.edu.in0009-0004-1739-8219
Gunveen AhluwaliaChitkara Centre for Research and Development, Chitkara University, Himachal Pradesh, India. gunveen.ahluwalia.orp@chitkara.edu.in0009-0006-6315-6450
Dr.R. MuruganDepartment of Computer Science and Information Technology, JAIN (Deemed-to-be University), Bangalore, Karnataka, India. murugan@jainuniversity.ac.in0000-0003-0903-5982
Dr. Pooja VarmaDepartment of Psychology, Jain deemed to be University, Bangalore, India. pvarma179@gmail.com0000-0002-4866-0177
Long-term blood pressure (BP) monitoring is necessary for the early diagnosis and treatment of hypertension and other cardiovascular diseases. But they are invasive and intermittent, and they are a real pain in the ass for the patient. The most recent advancements in wearables have transformed the smartwatch into a physical embodiment of a service that, in theory, could enable all-day blood pressure monitoring. In this paper, It propose a deep learning model for CMT (Continuous Monitoring of Blood Pressure), which can be measured using a smartwatch's sensors. For this, it utilizes optical and HR sensors, as well as an accelerometer, on the smartwatch to measure the physiological information of the wearer, such as heart rate (HR) and motion data. The behavior indicators are then input into a deep CNN for feature learning and an LSTM for analyzing temporal characteristics. The features from CNN and LSTM are extracted individually and then integrated and input into the regression model to estimate blood pressure. In this experiment, It acquired physiological signals from 50 unique subjects, as well as the reference blood pressure values, which were used as the dataset for training and testing the framework. It developed out the automatic pipeline, presented in this blog post, on a split dataset of 3 (train, validation, test). Results: It demonstrates the effectiveness of the framework, outperforming most state-of-the-art methods with a Mean Absolute Error of 3.26 mmHg and 2.12 mmHg in SBP and DBP, respectively. The presented framework enables non-intrusive and continuous 24/7 monitoring of blood pressure using widely used personal smartwatches. It’s not implausible to envision wearing a device like this 24/7, keeping track of blood pressure all the time and alerting to possible problems long before they even begin. This one, focusing on hypertension and other cardiovascular diseases (CVDs), may potentially contribute to what is needed to achieve better control and, ultimately, better overall health. (2) Its approach is extensible and applicable to other wearable devices, making it suitable for a broader range of users. In this context, the issue is raised regarding a (smart) watch that purportedly measures blood pressure daily and its significance for medical care.