Seyedalireza Khoshsirat

Master's Thesis Abstract

Segmentation of the left ventricle (LV) in cardiac magnetic resonance images (MRI) is an essential step for calculation of clinical indices such as ventricular volume and ejection fraction. In this thesis, we first explain essential concepts, then review existing methods for segmentation of LV. We continue by implementing and evaluating a method which employs deep learning algorithms combined with a level-set method to fully automatically segment the LV in short-axis cardiac MRI datasets.
The method employs deep learning algorithms to learn the segmentation task from the ground truth data. Convolutional networks are employed to automatically detect the LV chamber in MRI dataset. Stacked autoencoders are utilized to infer the shape of the LV. The inferred shape is incorporated into a level-set method to improve the accuracy and robustness of the segmentation. We validated our method using 45 cardiac MRI datasets taken from the MICCAI 2009 LV segmentation challenge and compared the results to the state-of-the-art methods. Excellent agreement with the ground truth was achieved. We computed validation metrics such as the percentage of good contours, Dice metric, average perpendicular distance, and conformity as respectively 83%, 80%, 3.4mm and 70%.

Keywords: Deep-Learning, Level-set method, Left Ventricle, Cardiac MRI, Machine Learning