Robust segmentation of brain structures in MRI

Authors:
Maria Murgasova;Daniel Rueckert;David Edwards;Jo Hajnal
Affiliations:
Department of Computing, Imperial College London;Department of Computing, Imperial College London;Department of Imaging Sciences, Faculty of Medicine, Imperial College London;Department of Imaging Sciences, Faculty of Medicine, Imperial College London
Venue:
ISBI'09 Proceedings of the Sixth IEEE international conference on Symposium on Biomedical Imaging: From Nano to Macro
Year:
2009

Citing 0
Cited 2

SPARC: unified framework for automatic segmentation, probabilistic atlas construction, registration and clustering of brain MR images

ISBI'10 Proceedings of the 2010 IEEE international conference on Biomedical imaging: from nano to Macro
Simultaneous reconstruction of 40-D myocardial motion from both tagged and untagged MR images using nonrigid image registration

MIAR'10 Proceedings of the 5th international conference on Medical imaging and augmented reality

Quantified Score

Hi-index	0.00

Visualization

Abstract

A novel method for the segmentation of brain structures combining registration-based and EM-based approaches is proposed. To address the issue of intensity variation within brain structures, we propose a method for creating spatial prior corresponding very well with the new subject anatomy, referred to as subject-specific atlas. The subject-specific atlas is created using multiple non-rigid registrations with the new subject, and therefore a robust non-rigid alignment of the prior with the image is achieved. The final EM segmentation step ensures a good quality classification of the complicated cortical structures based on intensity. The evaluation of the proposed method shows consistent results superior to both, registration-based and EM-based approaches.