Clasificación de señales ECG para la detección de enfermedades cardíacas : un estudio comparativo

ECG signals play an important role for heart disease detection, ranging from various types of arrhythmia to AV block and heart attack. Although electrocardiogram is a relatively simple test, its correct analysis requires both time and capable personnel. These requirements increase proportionally to the number of exams to be analyzed. Therefore, doing these processes in an automatic and trustworthy fashion becomes a need in the medical field. This paper explores the implementation of Machine Learning models for ECG signals classification in four different categories: normal patient, atrial fibrillation patient, patient with abnormal rhythm that could have a different disease and noisy signal that cannot be studied. Tackling this problem leads to examine the models? capacity to recognize a specific disease, differentiate between normal and abnormal signals that need further analysis as well as determining errors in electrocardiogram taking. Six models were trained: five based on a feature engineering approach and one based on deep learning. Likewise, two additional models trained with techniques to handling class imbalance (oversampling and cost sensitive classification) are proposed. Models with good classification performance are obtained, with F1 average scores between 0.73 and 0.8.

• Autores: Mosquera Rojas, Gonzalo Esteban
• Tipo de recurso: Trabajo de grado de pregrado
• Fecha de publicación: 2020
• Institución: Universidad de los Andes
• Repositorio: Séneca: repositorio Uniandes
• Idioma: spa
• OAI Identifier: oai:repositorio.uniandes.edu.co:1992/51547
• Acceso en línea: http://hdl.handle.net/1992/51547
• Palabra clave: Electrocardiografía; Fibrilación auricular; Clasificación automática; Ingeniería de características; Ingeniería
• Rights: openAccess
• License: https://repositorio.uniandes.edu.co/static/pdf/aceptacion_uso_es.pdf