Desenvolvimento de métodos de processamento e inteligência computacional no ECG ambulatorial

Abstract

Ambulatory monitoring of the electrocardiogram (ECG) allows to follow the patient's daily activities during periods of 24 hours (or more) making possible the study of cases with potentially fatal arrhythmic episodes. However, the major technological challenge that this type of monitoring faces is the loss of information due to the presence movement-related noise and artefacts. The analysis of the QT interval of the surface electrocardiogram or ventricular depolarization and repolarization interval is a non-invasive technique with a high value for the diagnosis and prognostics of cardiopathies and neuropathies, as well as for the prediction of sudden cardiac death. The analysis of the QT-interval standard deviation provides information about the dispersion (time or spatial) of ventricular repolarization. Hovever the presence of noise leads to errors in the detection of the T-wave end, which are non negligible due to small values of QT standard deviation in both pathological and healthy subjects. The main aim of this PhD thesis is to improve ambulatory ECG processing methods using computational intelligence, and specifically those involved in the detection of the T wave end and morphologic recognition of heartbeats, which could invalidate the QT interval variability analysis. A new approach and algorithm was proposed to identify the T-wave end, based on the computation of Trapezium’s areas. The method was validated (in terms of accuracy and repeatability), using signals from the Physionet QT Database. The performance of the proposed method in noisy conditions has been tested and compared with one of the most used approaches for estimating the T-wave end point, based on the threshold on the first derivative. The suggested computational intelligence method combines the features extraction using non-linear principal components analysis method and the multilayer perceptron neural network. The trapezium-based approach showed a good performance in noisy conditions and does not rely on any empirical threshold. It is adequate for use in scenarios where the levels of broadband noise are significant. The beats morphologic recognition methods were evaluated using ambulatory signals with and without artefacts from international prestigious databases, showing a good performance.