Automatic cardiac segmentation by fusing deep learning models

Abstract

Nowadays machine learning models can be used to automate the process of cardiac segmentation, a tedious task usually done by cardiologists and radiologists to diagnose heart diseases and get insights of a certain patient’s heart. In this work, we propose combining state-of-the-art deep learning based models to automatically delineate cardiac MRI slices. By combining existing successful models—using both a stacking ensemble and a majority voting algorithm—we get similar or better results than the existing individual methods. In the experiments carried out, the ensemble methods outperform the original baseline models.