Unconstrained blood pressure monitoring based on a neural network–assisted multistage pressure textile sensor

Abstract

Non-invasive blood pressure (NIBP) assessment via constrained wearable devices aided by intelligent algorithms has led to tremendous advancements in healthcare. However, existing NIBP assessment methods generally impose hard constraints on human behavior, resulting in lower accuracy of blood pressure assessments. Therefore, for unconstrained ambulatory blood pressure monitoring, we propose a textile-based sensor with a multistage structure that acquires human ballistocardiogram (BCG) signals. Using nylon fabric as a substrate, a multistage structure consisting of sensitive micro-cilia and nylon supports ensures that the sensor maintains high sensitivity (0.533 V/kPa) even under static pressure (0–7 kPa) equal to the self-weight of the human body. The BCG waves acquired by the sensor have excellent agreement with the cardiac vibration wave obtained with laser vibrometry (displacement precision of 0.5 pm). Benefiting from the accurate acquisition of BCG, the time interval between each characteristic point of BCG waves is independently used as the characteristic parameter. A blood pressure assessment model via a back-propagation neural network is proposed to realize an unconstrained blood pressure assessment. For unconstrained measurements of diastolic pressure (DP) and systolic pressure (SP), the mean absolute error and standard deviation of error were 3.90 ± 4.79 mmHg and 4.62 ± 6.00 mmHg, respectively; this result meets the accuracy criteria of the American Association for the Advancement of Medical Instrumentation for Class II medical. We propose a new approach for unconstrained ambulatory blood pressure monitoring that shows great potential in home, hospital, and personalized medicine.