Andrés Vargas, Universidad de las Fuerzas Armadas, Sangolquí - Ecuador 2019

Official link: http://repositorio.espe.edu.ec/jspui/handle/21000/21202

Abstract

The computer-aid diagnosis of cardiac arrhythmias is performed through the digital processing of electrocardiographic (ECG) signals. In such vein, various techniques and tools have been developed. Nevertheless, there are still open issue regarding the computational cost, precision and size related to the embedded systems applied to the analysis of ECG signals.

This master's thesis presents a diagnostic support system for the detection of normal and pathological heartbeats in long-term (Holter) ECG records, through the use of embedded systems. Such system allows for analyzing the characteristics of the QRS complexes, which are suitable to classify ventricular-type cardiac arrhythmias.

The use of the supervised KNN algorithm together with a reduced selection of training data, as well as an adequate selection of characteristics, are the most significant contributions of this research. The experiments are performed over the cardiac arrhythmia database from the Massachusetts Institute of Technology (MIT), which contains records with different types of arrhythmias. In addition, this research proposes a low-computational cost system for implementation in embedded systems, which should be able to perform real-time processing for its subsequent classification and visualization of ECG signals. To validate the performance of the classification algorithm, performance measures recommended by the literature are applied. This study is carried out according to the standards recommended by the Association for the Advanced of Medical Instrumentation (AAMI).

ECG signal processing

Graphic interface

About us