Difference between revisions of "2015 Summer Project Week:LungCAD"

From NAMIC Wiki
Jump to: navigation, search
Line 33: Line 33:
 
* Completed analysis for 248 GGOs and trained SVM with an accuracy of 89% to classify GGOs
 
* Completed analysis for 248 GGOs and trained SVM with an accuracy of 89% to classify GGOs
 
* Developed the framework for the LungCAD module in Slicer
 
* Developed the framework for the LungCAD module in Slicer
* Evaluate the Lesion segmentation algorithm as part of the Chest Imaging Platform
+
* Evaluated the Lesion segmentation algorithm as part of the Chest Imaging Platform
 
* Segmentation works well and is able to prevent the segmentation of vessels running through the lesion
 
* Segmentation works well and is able to prevent the segmentation of vessels running through the lesion
 
* LungCAD will call Lesion Segmentation CLI for segmentation and HeterogeneityCAD to evaluate features
 
* LungCAD will call Lesion Segmentation CLI for segmentation and HeterogeneityCAD to evaluate features

Revision as of 14:41, 24 June 2015

Home < 2015 Summer Project Week:LungCAD

Key Investigators

  • Jayender Jagadeesan
  • Tobias Penskofer
  • Sandy Wells
  • Clara Meiner
  • Raul San Jose Estepar

Project Description

Objective

  • Develop a module in 3D Slicer to segment the ground glass opacity (GGO) tumor, apply HeterogeneityCAD to obtain imaging metrics and classify the GGO.
  • Provide the module as an extension part of OpenCAD

Approach, Plan

  • Implement a simple region growing algorithm
  • Apply HeterogeneityCAD module
  • Use predetermined SVM classifier to decide the lesion type

Progress

  • Completed analysis for 248 GGOs and trained SVM with an accuracy of 89% to classify GGOs
  • Developed the framework for the LungCAD module in Slicer
  • Evaluated the Lesion segmentation algorithm as part of the Chest Imaging Platform
  • Segmentation works well and is able to prevent the segmentation of vessels running through the lesion
  • LungCAD will call Lesion Segmentation CLI for segmentation and HeterogeneityCAD to evaluate features
  • Pre-processed SVM will be utilized to classify the GGOs