Esquema de compensación de movimiento mediante registrado grupal aplicado a imagen cardiaca dinámica

Tipo de Tesis: 
Proyecto Fin de Carrera
Santiago Sanz Estébanez
Carlos Alberola López; Lucilio Cordero Grande
Matrícula de Honor (Honors)
ETSI Telecomunicación, Universidad de Valladolid

This project focuses on the analysis of cardiac magnetic resonance images and, specifically, on the characterization of the dynamic properties of the different structures responsible of blood pumping to the rest of the body. The objective pursued consists on facilitating the interpretation of the information contained in the images using a registration algorithm. Image registration is a particular interest in numerous medical imaging applications, in which a geometrical transformation that aligns corresponding points in a group of images is determined. In this project, a new elastic registration method is proposed, which will be refered as groupwise registration. In order to validate this method, a segmentation procedure will be executed. For this purpose, a versatile framework of image analysis that tries to overcome some of the limitations in the clinical application of cardiac magnetic resonance images will be introduced. Primarily, the difficulty in achieving images with both good spatial and temporal resolution, due to the limitations of acquisition in apnea and the presence of image artifacts, which are mainly derived from the existence of motion artifacts.
The normal procedure for the marking is the manual delination on the images by a medical expert, which causes errors due to the randomness and variability in the human behaviour. That is the reason why establishing a framework to automatically obtain segmentation in the whole registered sequence from manual initial segmentation, by using an overlap indicator like the Dice coefficient. The experiments that have been carried out in this project on magnetic resonance images from real patients validate the model here proposed, and also highlight the satisfactory performance of the method as far as segmentation is concern, even though the poor quality or the presence of artifacts in the images. Finally, it is included a number of appendices where the analysis carried out are theoretically justified.