Refinement of the 2D-segmentation of the left cardiac ventricle in magnetic resonance images using simulated annealing / Refinamento da segmentação 2D do ventrículo esquerdo em imagens de ressonância magnética utilizando "simulated annealing"

AUTOR(ES)

Griselda Esther Jara de Garrido

DATA DE PUBLICAÇÃO

1998

RESUMO

Diseases in complex organs like the heart require suitable analysis tools for diagnosis, therapy and patient attendance. An emerging area of interest in image processing are quantitative techniques that provide accurate data about the size, position and other parameters concerning these organs. It is not possible to apply these techniques if no accurate identification of the object under analysis is available. The main objective of this work is the proposal of a technique for the correct identification of the object. This work describes the proposal, implementation and assessment of a methodology to obtain an accurate segmentation of the cardiac left ventricle in magnetic resonance images. The method begins with an initial segmentation, not necessarily accurate, obtained through the balloon expansion technique and accomplishes subsequent refinements based on simulated annealing, looking for the minimization of an energy function that takes into consideration the local intensity of the voxel, the contour/surface relation and the edge contrast. The evaluation of the proposed technique is accomplished over numerical phantoms designed with the dimensions and typical contrasts found in magnetic resonance images of the heart. The robustness of the method is measured with respect to the initial segmentation and comparisons are made with other techniques, through the assessment of segmentation errors. Results over real magnetic resonance images of the heart are also presented.

ASSUNTO(S)

cardiac ventricles ventrículos cardíacos ressonância magnética simulated annealing recozimento simulado markov chains cadeias de markov segments magnetic resonance computer science segmentos

Documentos Relacionados

• Refinamento da segmentação 2D do ventrículo esquerdo em imagens de ressonância magnética utilizando "simulated annealing"
• Segmentação interativa do ventriculo esquerdo em sequencias de imagens de ressonancia magnetica (Cine MR)
• Segmentação de tumores de encéfalo em imagens por ressonância magnética baseada em informações texturais.
• Segmentação de displasias corticais focais em imagens de ressonancia magnetica do cerebro humano
• Segmentação do pulmão em sequências de imagens de ressonância magnética utilizando  a transformada de Hough.