Navegando por Assunto "Aprendizagem de máquina"

Filtrar resultados informando o último nome do autor
Agora exibindo 1 - 1 de 1
  • Imagem de Miniatura
    Item
    PyCBPE: umframework open­source para estudo de desempenho de algoritmos de aprendizagem de máquina aplicados para estimar a pressão arterial através de sinais de fotopletismografia
    (2021-07-15) LUZ, Luigi Fernando Marques da; Almeida Neto, Fernando Gonçalves de; Silva Neto, Eronides Felisberto da; http://lattes.cnpq.br/4364517952689670; http://lattes.cnpq.br/0473869396914603; http://lattes.cnpq.br/5085706500819680
    Heart diseases are a major cause of death nowadays. A typical sign of a serious heart conditions is the high blood preassure, which can be employed for helping the diagnosis of several heart diseases. However, the proper diagnosis of many heart illness may require a continuous monitoring of blood preassure, which can be difficult to achieve, since the standard sphygmomanometer method does not allow a continuous measurement of blood pressure, and the invasive cannula method is costly, unconfortable for the patient and requires access to hospital facilities. As an alternative, techniques employing photopletismography signals (PPG) have been proposed recently to estimate the blood pressure. These techniques use PPG signals, which can be acquired with a noninvasive sensor that can applied to the patient’s finger, for instance. Recent advances in the literature show that these signals can be used as input for machine learning models to provide an adequate estimation of blood pressure, according to some international standards for medical equipments. Besides an increasing number of studies in this field of knowledge, the authors usually do not make their simulations avaialable, what makes it difficult to compare different approaches. This work proposes an open­source framework to help researchers to compare the performance of machine learning algorithms, employed to estimate the blood pressure using PPG signals. The methodology to develop the framework is presented in detail in this document and a performace comparison of the models generated by four machnie learning algorithms (Linear Regression, Decision Tree, Random Forest and AdaBoost) is presented. The comparison show that using the proposed framework, the models achieve perfomance results similar to those obtained in the literature.