Detecção de fibrilação atrial em ECG com o uso da decomposição tensorial BTD

Abstract

The Atrial Fibrillation (AF) is a common cardiac arrhythmia associated with various cardiovascular diseases and has a significant impact on mortality worldwide. This work focuses on AF detection using data collected from cardiac monitoring through Eletrocardiogram (ECG), proposing new attributes for AF prediction. In particular, the present work proposes the use of convergence and optimization indicators of the Block-term Decomposition (BTD). This information will be combined with the R-R Intervals (RRIs) from ECG, a feature widely explored in related works for AF and other arrhythmias detection. This set of features enabled the early detection and diagnosis of AF using machine learning algorithms that performed the binary classification task between a healthy signal, labeled as Normal Sinus Rhythm (NSR), or with AF. The study also discusses the data acquisition from three different ECG databases: Atrial Fibrillation Database (AFDB), Long-term Atrial Fibrillation Database (LTAFDB) e Normal Sinus Rhythim Database (NSRDB). This set of databases provided the investigation that aimed to validate the proposed method and indicate future directions for technique improvement. The results obtained demonstrate the effectiveness of the presented approach for detecting FA in the test base, reaching an accuracy of 97.04%, sensitivity of 97.44% and specificity of 96.59%. These results highlight the potential of the method for integration into monitoring devices intended for diagnosing cardiac conditions.