Morphometric Analysis of Brain Structures for Improved Discrimination

  • Li Shen
  • James Ford
  • Fillia Makedon
  • Yuhang Wang
  • Tilmann Steinberg
  • Song Ye
  • Andrew Saykin
Conference paper

DOI: 10.1007/978-3-540-39903-2_63

Part of the Lecture Notes in Computer Science book series (LNCS, volume 2879)
Cite this paper as:
Shen L. et al. (2003) Morphometric Analysis of Brain Structures for Improved Discrimination. In: Ellis R.E., Peters T.M. (eds) Medical Image Computing and Computer-Assisted Intervention - MICCAI 2003. MICCAI 2003. Lecture Notes in Computer Science, vol 2879. Springer, Berlin, Heidelberg

Abstract

We perform discriminative analysis of brain structures using morphometric information. Spherical harmonics technique and point distribution model are used for shape description. Classification is performed using linear discriminants and support vector machines with several feature selection approaches. We consider both inclusion and exclusion of volume information in the discrimination. We perform extensive experimental studies by applying different combinations of techniques to hippocampal data in schizophrenia and achieve best jackknife classification accuracies of 95% (whole set) and 90% (right-handed males), respectively. Our results find that the left hippocampus is a better predictor than the right in the complete dataset, but that the right hippocampus is a stronger predictor than the left in the right-handed male subset. We also propose a new method for visualization of discriminative patterns.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Copyright information

© Springer-Verlag Berlin Heidelberg 2003

Authors and Affiliations

  • Li Shen
    • 1
  • James Ford
    • 1
  • Fillia Makedon
    • 1
  • Yuhang Wang
    • 1
  • Tilmann Steinberg
    • 1
  • Song Ye
    • 1
  • Andrew Saykin
    • 2
  1. 1.DEVLAB, Computer ScienceDartmouth CollegeHanoverUSA
  2. 2.Psychiatry and Radiology, Dartmouth Medical SchoolLebanonUSA

Personalised recommendations