Detecção de hipertrofia de ventrículo esquerdo através do eletrocardiograma

Abstract

This work proposes a system for automatic analysis of Left Ventricular Hypertrophy (LVH) using the electrocardiogram (ECG). The ECG is automatically segmented using especific rules to each wave used in this work (QRS complex and T wave). The end of the T wave is one of the hardest points to segment in the ECG. For this reason several algorithms for this point are compared. The best algorithm among the compared ones has generated a set of errors with mean and standard deviation of -8.16 ms and 11.37 ms respectively, while the manual annotation of two experienced specialists generated mean and standard deviation of 8.94 ms and 11.28 ms. In addition, tests were performed in order to detect LVH through electrocardiographic and physiological variables. Besides comparing the capacity to separate groups of normal and sick patients for various characteristics and criteria indicated by the literature, this paper proposes an original score to identify patients with LVH, for wich the sensitivity is 78.3% and the specificity is 91.3%.