Not Logged In

Segmenting Brain Tumors using Alignment-Based Features

Full Text: 2005-ICMLA.pdf PDF

Detecting and segmenting brain tumors in Magnetic ResonanceImages (MRI) is an important but time-consumingtask performed by medical experts. Automating this processis a challenging task due to the often high degree ofintensity and textural similarity between normal areas andtumor areas. Several recent projects have explored ways touse an aligned spatial 'template' image to incorporate spatialanatomic information about the brain, but it is not obviouswhat types of aligned information should be used. Thiswork quantitatively evaluates the performance of 4 differenttypes of Alignment-Based (AB) features encoding spatialanatomic information for use in supervised pixel classification.This is the first work to (1) compare severaltypes of AB features, (2) explore ways to combine differenttypes of AB features, and (3) explore combining AB featureswith textural features in a learning framework. We consideredsituations where existing methods perform poorly,and found that combining textural and AB features allows asubstantial performance increase, achieving segmentationsthat very closely resemble expert annotations.

Citation

M. Schmidt, I. Levner, R. Greiner, A. Murtha, A. Bistritz. "Segmenting Brain Tumors using Alignment-Based Features". International Conference on Machine Learning and Applications (ICMLA), Los Angeles, December 2005.

Keywords: brain segmentation, tumor, alignment-based feature, Random Field, machine learning, medical informatics, BTAP
Category: In Conference

BibTeX

@incollection{Schmidt+al:ICMLA05,
  author = {Mark Schmidt and Ilya Levner and Russ Greiner and Albert Murtha and
    Aalo Bistritz},
  title = {Segmenting Brain Tumors using Alignment-Based Features},
  booktitle = {International Conference on Machine Learning and Applications
    (ICMLA)},
  year = 2005,
}

Last Updated: April 27, 2012
Submitted by Russ Greiner

University of Alberta Logo AICML Logo