Analysis of Optimization Techniques In the Non-Invasive and Cuff-Less Method of Blood Pressure Estimation

Resumen

Regulation and control of high blood pressure, also known as hypertension, is complex,
necessitating the use of a continuous, accurate blood pressure monitoring system. All existing
continuous noninvasive approaches have their own set of problems, such as precise sensor
implantation, arterial pressure reconstruction from a finger cuff, and regular and subjectbased
calibration. The algorithm presented in this paper is based on new time-domain
features for continuous blood pressure monitoring in intensive care, which is critical. It can
be used to predict cardiovascular disease in units. The objective of this project is to propose
an optimized method for estimating blood pressure that uses Womersley number, QRS, QTc
interval, SDI, and regression algorithms to extract important characteristics from ECG and
PPG signals. These are then used to continually estimate the blood pressure. Performance
indicators such as Mean Absolute Error, Mean Square Error, Root Mean Square Error and
variance are then calculated. Significant characteristics such as alpha, QRS complex, QT
interval, SDI, and heart rate are obtained as features. For SBP, DBP, and MBP, the best
optimal feature set from GA lowered MAE from 13.20 to 9.54 mmHg, 9.91 to 5.48 mmHg,
and 7.71 to 3.37 mmHg, respectively. The relevance of ECG characteristics association with
BP is validated from the results. The importance of identifying the relationships between
ECG features and blood pressure is that it aids in the selection of important aspects that
reduce computational costs and errors in BP assessment. In this paper, we suggest different
regression models and optimization techniques employed on regression models to estimate
Blood Pressure with least error.