| Abstract [eng] |
Nocturnal non-dipping and morning surge in blood pressure for hypertensive patients are strong risk factors for cardiovascular events if untreated for prolonged time. Even though means for continuous and non-continuous monitoring of blood pressure parameters are available, there is a lack of methods for unobtrusive long-term monitoring. In this thesis, three different methods of pulse arrival time estimation from synchronously recorded electrocardiogram and photoplethysmogram are compared and later correlated with blood pressure parameters, such as systolic blood pressure, diastolic blood pressure, mean blood pressure and pulse pressure. The selected methods are middle amplitude of pulse slope, peaks of slope sum function and peaks of matched filtered photoplethysmogram. These methods are tested on thermal stress test database and breathing test database. The results were tested with two models such as Moens-Korteweg exponential arterial elasticity model and derivative of Moens-Korteweg linear model. The results show that volunteers from breathing test database have higher correlation than thermal stress test database. Also, the volunteers from breathing test database have moderate (0.4–0.59) to strong (0.6–0.79) downhill correlation for systolic blood pressure, mean blood pressure and pulse pressure, and very weak (0.0–0.19) to weak (0.2–0.39) correlation for diastolic blood pressure especially for middle amplitude of pulse slope method. When blood pressure estimation models based on pulse arrival time were created and tested, the correlation of real and predicted systolic blood pressure is moderate (0.4–0.59) to strong (0.6–0.79). Moreover, the models have good response such that their mean difference is <5 mmHg and standard deviation of difference is <8 mmHg when tested on systolic blood pressure and the results are similar for all three evaluated methods. |
