+44 (0)1509 263171
Please use this identifier to cite or link to this item:
|Title: ||Automatic segmentation of adipose tissue from thigh magnetic resonance images|
|Authors: ||Purushwalkam, Senthil|
McPhee, Jamie S.
|Issue Date: ||2013|
|Citation: ||PURUSHWALKAM, S. ... et al, 2013. Automatic segmentation of adipose tissue from thigh magnetic resonance images. IN: Kamel, M. and Campilho, A. (eds). Image Analysis and Recognition: 10th International Conference, ICIAR 2013, Póvoa do Varzim, Portugal, June 26-28, 2013, Proceedings. Lecture Notes in Computer Science, 7950, pp.451-458|
|Series/Report no.: ||Lecture Notes in Computer Science;7950|
|Abstract: ||Automatic segmentation of adipose tissue in thigh magnetic resonance imaging (MRI) scans is challenging and rarely reported in the literature. To address this problem, we propose a fully automated unsupervised segmentation method involving the use of spatial intensity constraints to guide the segmentation process. The novelty of this method lies in two aspects: firstly, an adaptive distance classifier, incorporating intra-slice spatial continuity, is used for robust region growing and segmentation estimation; secondly, polynomial based intensity inhomogeneity maps are generated to model inter- and intra-slice intensity variation of each pixel class and thus refine the initial classification. Our experimental results have demonstrated the effectiveness of imposing 3D intensity constraints to successfully classify the adipose tissue from muscles in the presence of image noise and considerable amounts of non-uniform MRI intensity. © 2013 Springer-Verlag.|
|Description: ||The final publication is available at Springer via http://dx.doi.org/10.1007/978-3-642-39094-4_51|
|Version: ||Accepted for publication|
|Publisher Link: ||http://dx.doi.org/10.1007/978-3-642-39094-4_51|
|Appears in Collections:||Conference Papers and Presentations (Computer Science)|
Files associated with this item:
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.