Modelos de classificação e prognóstico para avaliação do bem-estar fetal

Abstract

Fetal monitoring signals such as fetal heart rate (FHR) are critical indicators of fetal well-being. Computer-assisted analysis of FHR patterns alone or combined with patient’ clinical data has been used as a decision supporting-tool and prognostic models in the clinical environment. The biosignal-based systems comprise functionalities that varying from fetal wellbeing indicators exhibition to predictive and pattern classification capabilities; However, the prognostic model design suffers from the lack of gold standards for the building blocks decision-making. Thus impairing the direct comparison of proposals and the development of new solutions. In this thesis we propose an evaluation process for the building blocks of the decision supporting-tools and a prognostic model based on advanced signal processing techniques and machine learning algorithms for the fetal state assessment. First, we employed well-established guidelines to systematically map the literature in order to establish an overview of the state of the art related to the main parameters and building blocks used in the development of support systems for biosignal-based fetal diagnosis systems. Then, a process for evaluating these building blocks was developed, in which supervised machine learning algorithms were evaluated separately for time series and feature engineering within three signal segmentation schemes. The evaluation process also utilizes a semi-supervised machine learning algorithm to analyze the fetal state using annotated biosignal databases in conjunction with non-annotated databases. Finally, a prognostic model was developed based on the combination of parameters and the best-evaluated building blocks. The evaluation process provided a combination of techniques and methods that increased the final prognostic model performance, achieving relevant results with a smaller amount of data when compared to the state-of-the-art fetal status assessment solutions.