Performance Evaluation of Cuffless Blood Pressure Estimation Methods Using Linear and Non-Linear Models with PTT and PR Parameters

Blood pressure (BP) variability is linked with heart diseases, stroke, kidney failure, diabetes, and sleep apnea which are the leading cause of deaths worldwide. Continuous measurement of BP plays a vital role in timely detecting hypertension and hypotension and enabling accurate and reliable diagnosis and treatment of life-threatening diseases. Compared to the traditional sphygmomanometer and automatic oscillometric cuff BP measurement methods, cuffless BP measurement device can enable continuous measurement of systolic BP (SBP) and diastolic BP (DBP) in daily life without any disturbance, that play a major role in effective monitoring and controlling of hypertension. In this paper, we study the performance of linear and non-linear BP estimation models based on the pulse transit time (PTT) and pulse rate (PR) for measuring SBP, DBP, and mean arterial pressure (MAP). We demonstrate the reliability of the estimated SBP and DBP values by finding the MAP error computed between the estimated MAP using the MAP model and the MAP calculated directly using the estimated SBP and DBP values. The performance of the BP estimation methods is evaluated using the MIMIC database in terms of mean absolute error (MAE), Bland-Altman ratio (BAR), and BP estimation error range metric. Evaluation results showed that the non-linear BP estimation model performs better than the linear model. Results further showed that BP estimation models highly demand accurate estimation of PTT and PR from electrocardiogram (ECG) and photoplethysmogram (PPG) signals which is challenging under different kinds of artifacts.