Artificial neural networks ensemble applied to the electrical impedance tomography problem to determine the cardiac ejection fraction

Abstract

Cardiac Ejection Fraction (EF) is a parameter that indicates how much blood the heart is pumping to the body. It is a very important clinical parameter since it is highly correlated to the functional status of the heart. To measure the EF, diverse non-invasive techniques have been applied such as Magnetic Resonance. The method studied in this work is the Electrical Impedance Tomography (EIT) which consists in generate an image of the inner body using measures of electrical potentials - some electrodes are attached to the body boundary and small currents are applied in the body, the potentials are then measured in these electrodes. This technique presents lower costs and a high portability compared to others. It can be done in the patient bed and does not use ionizing radiation. The EIT problem consists in define the electrical distribution of the inner parts that results in the potentials measured. Therefore, it is considered as a non-linear inverse problem. To solve that, this work propose the application of an Artificial Neural network (ANN) Ensemble since it is simple to understand and implement. Our results show that the ANN Ensemble presents fast and good results, which are crucial for the continuous monitoring of the heart.