Enhancing the research toward wearable solutions for continuous noninvasive blood pressure monitoring

Wearable devices are potential power instruments for facing major health-related challenges such as the aging population, chronic diseases, and hospital service delivery. While wearables are nowadays widely used for monitoring physical activity and active life, also through some physiological parameters, their use in the clinical environment remains limited and challenging. Among the many reasons for the limited use of wearable solutions in the clinical field is the lack of a complete and updated regulatory framework, including specific guidelines to establish the accuracy of wearables and the ownership of the cost burden. Not to mention the acceptability of such devices among patients and healthcare providers in terms of comfort and trust. The European Horizon 2020-funded SINTEC project (Soft Intelligence Epidermal Communication Platform) aims to develop soft, sticky, and stretchable devices for specific use in the clinical field, particularly for continuous blood pressure monitoring (BP). Monitoring BP is essential for hypertension diagnosis and monitoring both in a clinical and home environment, especially in a scenario where deaths due to hypertension are increasing in high-income countries, according to WHO (World Health Organization). Different methods have been developed for BP monitoring including, invasive, noninvasive, intermittent, continuous, and cuffless techniques but still, a noninvasive technique for continuous BP monitoring remains a challenge. This thesis works aims to add major robustness to motion artifacts (MA) to the algorithm developed for the SINTEC project, exploiting the electrocardiographic (ECG) and photoplethysmographic (PPG) signals to extrapolate BP measurements through the Machine Learning (ML) linear regression technique. The novelty proposed in this thesis work includes the acquisition of the accelerometer signals, adding an extra filtering possibility through a Least Mean Square (LMS) adaptive filter, and extra features cleaning through thresholding based on the accelerometer signals. Specifically, a combination of both techniques has allowed for an average of 25% reduction of MAE (Mean Absolute Error) in highly corrupted signals, allowing in some cases to obtain an acceptable low error according to the AAMI/ISO/ESH guidelines from acquisitions that gave nonacceptable MAE. For testing the validity of these techniques, signals have been acquired through SHIMMER (Sensing Health with Intelligence, Modularity, Mobility, and Experimental Reusability) devices in healthy subjects in different conditions, such as sitting still on a chair, standing up and down from the chair, doing small walking, and doing small hand gestures. Furthermore, this thesis work aims to enhance the useability of this new BP monitoring technique by proponing a GUI (Graphical User Interface) for better determinate input parameters necessary for a good BP estimation, to enhance the accessibility to this technology also to users without programming background and allowing the transferability of the system for signals that will be recorded with the intended SINTEC final devices. Finally, the GUI speeds up the research process for the best ML coefficients needed in the algorithm. In conclusion, this thesis work aims to contribute to the research for a wearable solution for BP monitoring that is accurate, highly comfortable for the patient, easy to set up, and use for both patients and physicians to be officially recognized in a clinical environment.