Computational intelligence for abdominal aortic aneurysm imaging analysis

Abstract

The main material on which this Thesis works are Computed Tomography (CT) abdominal images, specifically Computed Tomography Angiography (CTA) of patients of Abdominal Aortic Aneurysm (AAA) which underwent Endovascular Aneurysm Repair (EVAR). It is expected of the deployment of CT imaging devices to improve our ability for detecting and monitoring anatomical structures in the body, as well as tracking their evolution in time. However, these kind of images pose new kinds of challenges. This Thesis has grown along two main lines of work. First, we have worked on the classification based segmentation of challenging structures in the images (3D CTA) applying an Active Learning approach to the training of classifiers. Second, we have worked on the visual assessment of AAA's thrombus evolution, and the development of a Computer Aided Diagnosis (CAD) system predicting the evolution of patients who underwent EVAR. The CAD input features are the aortic deformation values measured by image registration techniques. Prediction has to be achieved by Machine Learning approaches. Some of the works in this Thesis have been done in close relation with the clinicians actually treating the AAA patients, so that some of the contents of the Thesis are directly aimed to provide them with useful tools for their daily work.