Estudo sobre a estimativa da pressão arterial utilizando dados de fotopletismografia da base MIMIC e uma rede neural artificial

Abstract

This paper addresses the development of an artificial neural network with the purpose of estimating continuously blood pressure value. The types of methods currently used for measuring blood pressure in clinics, hospitals and residences are discussed. Afterwards, considering that this paper proposes a continuous measuring method, the pros and cons of each current continuous measuring method is discussed in details. This proposes the use of photoplethysmography (PPG) signals to estimate blood pressure from an artificial neural network (ANN). It will receive the fast Fourier transform of the PPG signal obtained from the MIMIC database -III and give out a value of blood pressure. For ANN’s training and testing there were used 108 and 45 blocks, respectively. At the end, the results obtained from the neural network will be analyzed statistically in conjunction with the actual blood pressure values to validate whether this method could be developed for a possible replacement of the currently used methods. The analyses showed that, with an improvement in the filtering and separation of the input data, an ANN can measure a person’s blood pressure.