Implementación de un sistema para análisis y clasificación de electrocardiogramas con algoritmos de inteligencia artificial

Abstract

CONDITION FOR PUBLICATION OF PROJECT. English to write the same as described before in the Resumen. Use cursive fonts in this section. Cardiovascular diseases are the leading cause of mortality worldwide. Arrthythmias require early and accurate detection, with the electrocardiogram (EKG) being the standard tool for analysis. However, manual interpretation requires time and clinical expertise, which limits its applicability. The project aims to implement a heartbeat classification system in EKG signals using artificial intelligence techniques, utilizing convolutional neural networks (CNNs) and the standardized AMMI classification. The project was developed using the MIT-BIH Arrhythmia database, processed using noise removal, normalization, and heartbeat segmentation filtering techniques. Data balancing was applied with SMOTE, and a CNN was subsequently trained and validated in a Python environment using specialized libraries such as WFDB, TensorFlow, and Keras. The results showed a model capable of classifying heartbeats into the five categories defined by the AAMI, achieving accuracy and sensitivity metrics for a reliable diagnosis. The system constitutes a complementary tool for medical personnel and has the potential for integration into portable devices and telemedicine, contributing to the early detection of arrhythmias and the optimizing medical care. Keywords: Arrhythmias, AI, Neural Networks, AAMI.