Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Automated Segmentation of the Pericardium Using a Feature Based Multi-atlas Approach

Norlén, Alexander LU (2014) In Master's Theses in Mathematical Sciences FMA820 20141
Mathematics (Faculty of Engineering)
Abstract
Multi-atlas segmentation is a widely used method that has proved to work well for the problem of segmenting organs in medical images. But standard methods are time consuming and the amount of data quickly grows to a point making use of these methods intractable. In this work we present a fully automatic method for segmentation of the pericardium in 3D CTA-images. We use a multi-atlas approach based on feature based registration (SURF) and use RANSAC to handle the large amount of outliers. The multi-atlas votes are fused by incorporating them into an MRF together with the intensity information of the target image and the optimal segmentation is found efficiently using graph cuts. We evaluate our method on a set of 10 CTA-volumes with manual... (More)
Multi-atlas segmentation is a widely used method that has proved to work well for the problem of segmenting organs in medical images. But standard methods are time consuming and the amount of data quickly grows to a point making use of these methods intractable. In this work we present a fully automatic method for segmentation of the pericardium in 3D CTA-images. We use a multi-atlas approach based on feature based registration (SURF) and use RANSAC to handle the large amount of outliers. The multi-atlas votes are fused by incorporating them into an MRF together with the intensity information of the target image and the optimal segmentation is found efficiently using graph cuts. We evaluate our method on a set of 10 CTA-volumes with manual expert delineation of the pericardium and we show that our method provides comparable results to a standard multi-atlas algorithm but at a large gain in computational efficiency. (Less)
Please use this url to cite or link to this publication:
author
Norlén, Alexander LU
supervisor
organization
course
FMA820 20141
year
type
H2 - Master's Degree (Two Years)
subject
keywords
computer vision, medical image analysis, multi-atlas segmentation, feature based registration, Markov Random Fields, pericardium segmentation
publication/series
Master's Theses in Mathematical Sciences
report number
LUTFMA-3267-2014
ISSN
1404-6342
other publication id
2014:E53
language
English
id
4695207
date added to LUP
2014-12-15 13:24:38
date last changed
2014-12-15 13:24:38
@misc{4695207,
  abstract     = {{Multi-atlas segmentation is a widely used method that has proved to work well for the problem of segmenting organs in medical images. But standard methods are time consuming and the amount of data quickly grows to a point making use of these methods intractable. In this work we present a fully automatic method for segmentation of the pericardium in 3D CTA-images. We use a multi-atlas approach based on feature based registration (SURF) and use RANSAC to handle the large amount of outliers. The multi-atlas votes are fused by incorporating them into an MRF together with the intensity information of the target image and the optimal segmentation is found efficiently using graph cuts. We evaluate our method on a set of 10 CTA-volumes with manual expert delineation of the pericardium and we show that our method provides comparable results to a standard multi-atlas algorithm but at a large gain in computational efficiency.}},
  author       = {{Norlén, Alexander}},
  issn         = {{1404-6342}},
  language     = {{eng}},
  note         = {{Student Paper}},
  series       = {{Master's Theses in Mathematical Sciences}},
  title        = {{Automated Segmentation of the Pericardium Using a Feature Based Multi-atlas Approach}},
  year         = {{2014}},
}