Few Shot Learning for Classification of Heart Diseases from Electrocardiograms
Cardiovascular Diseases, Neural Networks, Few Shot Learning, Electrocardiogram, Classification of cardiovascular diseases
Cardiovascular diseases are the most common cause of death worldwide, and in Brazil, they represent the main cause of disability pensions and expenses with hospitalizations in the national territory. However, only 4.1% of doctors in Brazil are cardiologists, and this shortage compromises the analysis and preparation of reports based on simple exams, such as the electrocardiogram (ECG). In order to streamline the screening process in medical centers that make ECG reports from a distance, researchers have been developing a set of computational algorithms to automatically classify ECG signals as to the state of normality or changes in cardiac electrical activity. This work aims to develop a neural network model using Few Shot Learning to classify heart disease from digital tracings of ECG signals containing 12 leads.