Bibliographic citations
Pereyra, L., Gomez, J. (2024). Desarrollo de un software de detección de arritmias cardiacas en señales ECG basado en algoritmos de Deep Learning [Trabajo de Suficiencia Profesional, Universidad Peruana de Ciencias Aplicadas (UPC)]. http://hdl.handle.net/10757/683776
Pereyra, L., Gomez, J. Desarrollo de un software de detección de arritmias cardiacas en señales ECG basado en algoritmos de Deep Learning [Trabajo de Suficiencia Profesional]. PE: Universidad Peruana de Ciencias Aplicadas (UPC); 2024. http://hdl.handle.net/10757/683776
@misc{renati/1300807,
title = "Desarrollo de un software de detección de arritmias cardiacas en señales ECG basado en algoritmos de Deep Learning",
author = "Gomez Hurtado, John Roland",
publisher = "Universidad Peruana de Ciencias Aplicadas (UPC)",
year = "2024"
}
This study introduces a software designed for detecting cardiac arrhythmias from ECG signals. For this purpose, a comparison of the performance of different deep learning techniques (LSTM, Dense and CNN) applied to the detection and classification of arrhythmias in an ECG signal was performed. The primary goal is to identify the optimal predictive model for cardiac arrhythmias through each Deep Learning technique and to analyze their respective performances. The Physionet MIT-BIH database was used for training, providing a dataset of 108,854 samples derived from 48 patients, including those with and without diagnosed arrhythmias. The three specified neural networks were trained using this dataset, and their performance was assessed through the ROC curve analysis. Among the models, the one-dimensional convolutional neural network (1D-CNN) demonstrated the highest performance. This network was then applied to construct an arrhythmia detection algorithm, featuring one hidden convolutional layer and processing 3,600 samples at a time (equivalent to 10-second signal segments). The results were highly satisfactory, with the 1D-CNN achieving a precision rate of 92.07%, LSTM 82.10%, and Dense 75%.
This item is licensed under a Creative Commons License