Using Artificial Neural Networks to Extract Features for Heart Failure Prediction

Abstract

Heart failure is one of the most serious medical conditions affecting humans and potentially leading to death. It occurs when the heart muscle fails to pump blood adequately and effectively. Therefore, due to the seriousness of this disease, early prediction of patient outcomes is essential for enabling timely and appropriate treatment, which may reduce symptoms and increase longevity. This study aims to predict the survival status of heart failure patients and to identify the most influential clinical features affecting patient outcomes. A dataset of 299 heart failure patients was used, and artificial intelligence techniqueswere applied, specifically Artificial Neural Networks (ANN). In order to make this prediction, each feature was tested individually by feeding it into the ANN model to assess its impact on patient survival. The experimental results show that two features—serum creatinine and ejection fraction — were the most influential features and can independently be used to predict whether the patients with heart failure will survive or not.