Search academic texts

Registration Number

0224U002814, 0123U101143 , R & D reports

Title

A universal method of cuffless blood pressure measurement using neural networks with deep learning

Head

Rubel Andrii S., Доктор філософії

Registration Date

04-03-2024

Organization

National Aerospace University "Kharkiv Aviation Institute"

Description

The object of the study is electrocardiogram and photoplethysmogram signals obtained using sensors placed on the patient's hands. The subject of the study is methods of statistical and correlational analysis; methods of feature extraction in a spectral and bispectral domains, regression analysis methods; methods of machine learning, in particular, using convolutional and recurrent neural networks. The aim of the study is to create a universal method of cuffless blood pressure measurement which will allow measurement in difficult conditions (stress or physical exertion) and will automatically adjust to the individual characteristics of each patient without involving additional equipment for calibration. Today, hypertension is one of the leading causes of death in the world both among civilian and military persons, and the only way to detect and control this disease is to monitor blood pressure (BP). Therefore, the task of developing cuffless tonometers suitable for patient constant wearing is extremely important. Existing cuffless solutions are not accurate enough, in addition, the design and principle of operation of most of them requires the use of two hands at the same time, which makes these devices unsuitable for measuring BP in dynamics and during exercise. Another problem is the need to calibrate the device for each patient. This project proposes a method of cuffless blood pressure measurement, devoid of these shortcomings. The method is based on determining the pulse transit time (PTT) between two points of the blood vessel and the shape of the pulse wave (PW). The novelty of the method is the use of only one signal received from the sensor on the person's arm, as well as the use of deep learning neural networks and a refined signal model taking into account noise of different nature, which will significantly improve the accuracy and stability of blood pressure measurement and will eliminate the need to calibrate the device using additional equipment.

Product Description

popup.authors

Abramova Viktoriia V.
Viunytskyi Oleh H.
Kovalenko Bohdan V.
Kozhemiakina Nadiia V.
Rebrov Volodymyr S.
Rubel Andrii S.
Rubel Oleksii S.
Chumachenko Dmytro I.