The main function of the heart is to pump blood to all areas of the body and the proper occurrence of this process is essential for health. The ejection fraction is an important clinical parameter to determine the amount of blood pumped by the heart in each cardiac cycle. A value outside the normal range indicates that the heart is contracting abnormally. Combining low cost and high portability, Electrical Impedance Tomography with Regression Models is an alternative to obtain continuous estimations of cardiac ejection fraction, allowing a quick diagnosis on the health of the heart. This paper presents four computational intelligence models to estimate the ejection fraction using the electrical measures from Electrical Impedance Tomography: Multilayer Perceptron, Extreme Learning Machine, Random Forests and Elastic Net Regressor. A simulated dataset is used to train and assess the performance of all models. In order to improve the model's performance, an exhaustive search procedure on the parameters of each model is implemented. The overall evaluation of the results show that all models achieved percentage errors below 2% and the neural networks have produced better averaged predictions.