Template method to improve brain segmentation from inhomogeneous brain magnetic resonance images at high fields

  • Authors:

  • Marcelo A. Castro;Jianhua Yao;Yuxi Pang;Eva Bake;John Butman;David Thomasson

  • Affiliations:

  • National Institutes of Health, Department of Radiology and Imaging Sciences;National Institutes of Health, Department of Radiology and Imaging Sciences;National Institutes of Health, Department of Radiology and Imaging Sciences and Philips Healthcare;National Institutes of Health, Department of Radiology and Imaging Sciences;National Institutes of Health, Department of Radiology and Imaging Sciences;National Institutes of Health, Department of Radiology and Imaging Sciences

  • Venue:
  • ISBI'10 Proceedings of the 2010 IEEE international conference on Biomedical imaging: from nano to Macro
  • Year:
  • 2010

Quantified Score

Hi-index 0.00

Visualization

Abstract

Magnetic resonance imaging of the brain at high fields (e.g. 3T) provides high resolution and high signal to noise ratio images suitable for a wide range of clinical applications. However, radiofrequency (or B1) inhomogeneity increases with the magnetic field and produces undesired intensity variations responsible for inaccuracies in quantitative analyses. A method to perform brain segmentation using T1 maps whose inhomogeneity was corrected using previously acquired B1 maps is described. A library of B1 maps was created and a method to compensate the T1 inhomogeneity using a B1 map from another subject (template) was developed. The performance of the template-based method was evaluated in 19 healthy volunteers. Our method produced significantly better segmentations than the retrospective N3 method and the one without B1 inhomogeneity correction.