Supervised Fissure Enhancement

From NAMIC Wiki
Revision as of 22:16, 8 January 2015 by Jcross186 (talk | contribs) (→‎Project Description)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
Home < Supervised Fissure Enhancement

Key Investigators

  • James Ross (Brigham and Women's Hospital, Boston)
  • German Gonzalez (Brigham and Women's Hospital, Boston)
  • Rola Harmouche (Brigham and Women's Hospital, Boston)

Project Description

Namic2015 supervisedFissureEnhancement.jpg Jcr namic 2015 results.jpg

The Chest Imaging Platform provides automatic tools for lung lobe segmentation. These rely in part on detecting pulmonary fissures, which can be difficult due to poor signal-to-noise ratios and/or lung pathologies.

Objective

  • Use positive and negative examples of fissures to train a classifier that can be used to detect/enhance fissures

Approach, Plan

  • Guided by approach described in van Rikxoort et al: 'Supervised enhancement filters: application to fissure detection in chest CT scans'
  • Use itkDiscreteHessianGaussianImageFunction and itkDiscreteGaussianDerivativeImageFunction for point-wise computation of Hessian and gradient features
  • Use logistic regression as classifier. Investigate different feature subsets for best classification results.

Progress

  • Early experiments are promising. Would like to speed up gradient and Hessian computations.
  • Results: Increased itkDiscreteHessianGaussianImageFunction and itkDiscreteGaussianDerivativeImageFunction max error parameter from 0.01 to 0.5. Noticed a factor of ten speed-up with negligible impact on logistic regression classifier