Employing Visual Analytics to Aid the Design of White Matter Hyperintensity Classifiers.

Information

• Publication Type: Journal Paper (without talk)
• Workgroup(s)/Project(s):
  • Visualization Group
• Date: 2016
• Journal: Proceedings of International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI)

Abstract

Accurate segmentation of brain white matter hyperintensi-ties (WMHs) is important for prognosis and disease monitoring. To thisend, classiers are often trained { usually, using T1 and FLAIR weightedMR images. Incorporating additional features, derived from diusionweighted MRI, could improve classication. However, the multitude ofdiusion-derived features requires selecting the most adequate. For this,automated feature selection is commonly employed, which can often besub-optimal. In this work, we propose a dierent approach, introducing asemi-automated pipeline to select interactively features for WMH classi-cation. The advantage of this solution is the integration of the knowledgeand skills of experts in the process. In our pipeline, a Visual Analytics(VA) system is employed, to enable user-driven feature selection. Theresulting features are T1, FLAIR, Mean Diusivity (MD), and RadialDiusivity (RD) { and secondarily,CSand Fractional Anisotropy (FA).The next step in the pipeline is to train a classier with these features,and compare its results to a similar classier, used in previous work withautomated feature selection. Finally, VA is employed again, to analyzeand understand the classier performance and results.

Additional Files and Images

Weblinks

No further information available.

BibTeX

@article{raidou_miccai16,
  title =      "Employing Visual Analytics to Aid the Design of White Matter
               Hyperintensity Classifiers.",
  author =     "Renata Raidou and Hugo J. Kuijf and Neda Sepasian and Nicola
               Pezzotti and Willem H.  Bouvy and Marcel Breeuwer and Anna
               Vilanova i Bartroli",
  year =       "2016",
  abstract =   "Accurate segmentation of brain white matter
               hyperintensi-ties (WMHs) is important for prognosis and
               disease monitoring. To thisend, classiers are often trained
               { usually, using T1 and FLAIR weightedMR  images. 
               Incorporating  additional  features,  derived  from 
               diusionweighted MRI, could improve classication. However,
               the multitude ofdiusion-derived features requires selecting
               the most adequate. For this,automated feature selection is
               commonly employed, which can often besub-optimal. In this
               work, we propose a dierent approach, introducing
               asemi-automated pipeline to select interactively features
               for WMH classi-cation. The advantage of this solution is
               the integration of the knowledgeand skills of experts in the
               process. In our pipeline, a Visual Analytics(VA)  system  is
                employed,  to  enable  user-driven  feature  selection. 
               Theresulting  features  are  T1,  FLAIR,  Mean  Diusivity 
               (MD),  and  RadialDiusivity (RD) { and secondarily,CSand
               Fractional Anisotropy (FA).The next step in the pipeline is
               to train a classier with these features,and compare its
               results to a similar classier, used in previous work
               withautomated feature selection. Finally, VA is employed
               again, to analyzeand understand the classier performance
               and results.",
  journal =    "Proceedings of International Conference on Medical Image
               Computing and Computer Assisted Intervention (MICCAI)",
  URL =        "https://www.cg.tuwien.ac.at/research/publications/2016/raidou_miccai16/",
}