Desarrollo de un Clasificador de Normalidad Cardíaca Basado en las Técnicas de Monitorización Ambulatoria de Electrocardiografía y Presión Arterial a Partir de Señales ECG

thumbnail.default.alt
Share
Date
2023-11-24
Publisher
Universidad Antonio Nariño
Document type
COAR type
http://purl.org/coar/resource_type/c_7a1f
Citation
Bibliographic Managers
Document Viewer
Select a file to preview:
item.page.resume
Abstract
Heart and hypertensive diseases are among the leading causes of death in Colombia and the world. Because of this, they are included in the Sustainable Development Goals of the WHO’s 2030 agenda as a high priority issue. The study of cardiac dynamics is essential to understand heart disease and develop effective diagnostics and treatments. Ambulatory blood pressure monitoring (ABPM) and cardiac Holter are important techniques to assess cardiac activity under daily conditions. Cardiovascular disease and hypertension are significant health problems, and the lack of early detection is of concern. The aim of this work is to develop an algorithm that combines ECG and blood pressure estimation to classify cardiac dynamics as normal or abnormal. The algorithm is based on ECG signal analysis and intelligent blood pressure estimation through the DII derivative of this signal. The purpose is to improve early diagnosis and prevent cardiovascular diseases. Machine learning and quantitative methodology will be used for data extraction and classification.As a result of the study, two Gaussian process regression (GPR) models were derived to estimate blood pressure, specifically the systolic blood pressure model exhibited a coefficient of determination (R2) of 0.83 and a mean square error of 8.4, while the diastolic blood pressure model showed an R2 of 0.88 and an RMSE of 3.54. Together, these regression models yielded a cumulative RMSE of 11.94, thus meeting the standards established by the British Hypertension Society (BHS) and classifying as type C. Therefore, the blood pressure estimation model possesses the feasibility for implementation in clinical settings. Regarding the cardiac normality classifier, a Bagged Trees classification model was employed and demonstrated an accuracy of 96.6 %, allowing effective classification of normal ECG signals among five types of arrhythmias. Finally, through a binary classification that considers the estimated values of arterial pressure, heart rate and the results of the normality classification, it is determined whether the cardiac dynamics are within normal parameters or whether they present any mechanical or electrical abnormality
item.page.coverage.spatial
Colombia( Popayán, Cauca)
item.page.coverage.temporal