Using Deep Learning to Estimate Systolic and Diastolic volumes from MRI-images

Deep learning with convolutional neural networks has become a widely used tool for computer vision tasks. In this paper, we focus on the problem of automatically annotating the volumes of the left ventricle in the heart, based on DICOM MRI-images. We discuss the solution, which was the second best in an international data mining competition. We show that it is possible to achieve near-human performance using a deep learning approach when using task-specific model architectures.