BundleMAP: Anatomically Localized Features from dMRI for Detection of Disease

Mohammad Khatami, Tobias Schmidt-Wilcke, Pia C. Sundgren, Amin Abbasloo, Bernhard Schölkopf und Thomas Schultz
In proceedings of Int'l Workshop on Machine Learning in Medical Imaging (MLMI), pages 52-60, Springer, 2015
 

Abstract

We present BundleMAP, a novel method for extracting features from diffusion MRI (dMRI), which can be used to detect disease with supervised classification. BundleMAP uses ISOMAP dimensionality reduction to aggregate measurements over localized segments of nerve fiber bundles, which are natural anatomical units in this data. We obtain a fully integrated machine learning pipeline by combining this idea with mechanisms for outlier removal and feature selection. We demonstrate that it increases accuracy on a clinical dataset for which classification results have been reported previously, and that it pinpoints the anatomical locations relevant to the classification.

Bilder

Zusätzliches Material

Bibtex

@INPROCEEDINGS{Khatami:MLMI2015,
     author = {Khatami, Mohammad and Schmidt-Wilcke, Tobias and Sundgren, Pia C. and Abbasloo, Amin and Sch{\"o}lkopf,
               Bernhard and Schultz, Thomas},
      pages = {52--60},
      title = {BundleMAP: Anatomically Localized Features from dMRI for Detection of Disease},
  booktitle = {Int'l Workshop on Machine Learning in Medical Imaging (MLMI)},
     series = {LNCS},
     volume = {9352},
       year = {2015},
  publisher = {Springer},
   abstract = {We present BundleMAP, a novel method for extracting features from   diffusion MRI (dMRI), which can
               be used to detect disease with supervised classification. BundleMAP uses ISOMAP dimensionality
               reduction to aggregate measurements over localized segments of nerve fiber bundles, which are
               natural anatomical units in this data. We obtain a fully integrated machine learning pipeline by
               combining this idea with mechanisms for outlier removal and feature selection.   We demonstrate that
               it increases accuracy on a clinical dataset for which classification results have been reported
               previously, and that it pinpoints the anatomical locations relevant to the classification.}
}