SW pro automatickou analýzu a modelování myokardu z MOLLI sekvencí

Abstract

This diploma thesis deals with design and implementation of algorithm for segmentation of heart structures from MOLLI sequences. The active contour method based on the local Gaussian distribution was used for this segmentation. This method was chosen based on the study of scientific papers contained in the theoretical part. Thanks to this segmentation, it is possible to measure the cardiac structure parameters in the resulting binary model. Furthermore, the theoretical part contains knowledge about heart sensing using MRI and MOLLI sequences. The practical part describes the individual steps of preprocessing, segmentation and measurement of parameters of the given structure. The measured parameters are length, width, cavity content in the incision and thickness of the left ventricular heart muscle. The conclusion of this work is design of graphical user interface, which allows simple and intuitive segmentation. Part of the thesis is also testing segmentation robustness by comparing segmentation results with manual contours created by an expert. Based on testing of similarity to manual contours using DICE coefficient, mean quadratic error and correlation coefficient, robustness and success of segmentation were evaluated.