Opis:

Profil naukowy autora (MOST Wiedzy): Artur Poliński

Słowa kluczowe: blood pressure, photopletysmography (PPG), Artificial Neural Networks (ANN)

An individual’s blood pressure is a critical parameter for assessing overall health. It allows for the early detection of potential issues, such as hypertension or hypotension, which can lead to severe conditions like heart disease, stroke, or kidney failure. Hence, it is desirable to develop a method that will allow blood pressure to be measured continuously and noninvasively, which will not be burdensome for the tested person. There are different approaches to achieve this goal. Thus, in this work, we will focus on selected techniques and selected problems related to the issue of non-invasive blood pressure estimation.

The simulation of blood flow, performed using software developed independently by the Author, was employed to demonstrate the influence of various model parameters on the simulation outcomes. This analysis illustrates how different components of the model affect the results and, consequently, highlights the inherent limitations of such an approach.

An original contribution to the assessment of blood pressure estimation is also the demonstration of the impact of low-pass filtering on the accuracy of locating characteristic points in the photoplethysmographic signal, as well as in the blood pressure signal. The latter reduces the influence of indirect measurement based on photoplethysmographic techniques. Since the accuracy of locating characteristic points directly affects the estimation of pulse transit time and pulse arrival time, the analysis indirectly reveals the consequences for the accuracy of blood pressure estimation. As an alternative to low-pass filtering, the monograph presents results obtained using several approximation-based methods for reconstructing the photoplethysmographic and blood pressure signals, including a model developed by the Author.

The subsequent chapter examines which mathematical relationship between pulse transit time and blood pressure provides the best approximation of various theoretical models. The chapter also includes the Author’s considerations regarding the potential applicability of frequency-domain analysis, motivated by existing studies addressing this topic.

As an alternative to estimating blood pressure from the photoplethysmographic signal, the literature describes an approach based on measuring impedance changes. Accordingly, this work employs numerical models to present original research identifying the factors that generate the impedance-change signal for various electrode configurations and different model parameters.

The next chapter contains original analyses of blood pressure estimation models derived from physiological principles and incorporating information from pulse arrival time, the electrocardiographic signal, and the respiratory signal. The results obtained for these models are compared with those produced by neural network architectures that rely on approximation capabilities. To ensure a fair comparison, both approaches were evaluated using identical datasets and models with a comparable number of parameters.

The final chapter presents the Author’s original approach to two issues. The first concerns the locality of pulse wave velocity, a topic mentioned in the literature but not examined in depth. The second concerns the estimation of blood pressure using only the variable component of the photoplethysmographic signal, an aspect that has been largely overlooked in previous research.