Difference between revisions of "2012 Summer Project Week:Interactive Needle Segmentation"

From NAMIC Wiki
Jump to: navigation, search
Line 4: Line 4:
  
 
Image:PW-MIT2012.png|[[2012_Summer_Project_Week#Projects|Projects List]]
 
Image:PW-MIT2012.png|[[2012_Summer_Project_Week#Projects|Projects List]]
Image:gyn-needle-segmentation-1.png|Screenshot of Manual Needle Segmentation (20 minutes)
+
Image:gyn-needle-segmentation-1.png | Screenshot of Manual Needle Segmentation (20 minutes)
 +
Image:Needle-TractographyFiducialSeeding.png | Screenshot of Single Needle Modeled With Tractography Fiducial Seeding Module
 +
Image:Needle-TractographyFiducialSeeding2.png | Screenshot of Multiple Needles Modeled With Tractography Fiducial Seeding Module (Hassian Image Visible)
 
</gallery>
 
</gallery>
  
Line 32: Line 34:
  
 
<h3>Progress</h3>
 
<h3>Progress</h3>
 +
 +
Needles were effectively modeled from pelvic MRI scans using the Tractography Fiducial Seeding module:
 +
*Scans were cropped to the region of interest
 +
*Hessian images were generated from this data
 +
*Fiducials were placed in regions known to be part of a needle
 +
*The Tractography Fiducial Seeding module was used to generate “tracts” that pass through the fiducial – in this case, the needles
 +
Refinements need to be made to optimize the modeling for brachytherapy needles rather than fiber bundles:
 +
*A sense of “momentum” is needed, so that needles do not abruptly end or change direction in signal voids
 +
*Existing needle and template models (eg, from iGyne) may also be incorporated
  
  

Revision as of 04:23, 22 June 2012

Home < 2012 Summer Project Week:Interactive Needle Segmentation

Key Investigators

  • Rutgers: Nabgha Farhat
  • BWH: Yi Gao, Neha Agrawal, Nabgha Farhat, Jan Egger, Tina Kapur
  • Isomics: Steve Pieper

Objective

The goal of this project is to achieve fast (< 2 minute) interactive segmentation of 5-10 needles from MRI images using Slicer.


Approach, Plan

We will explore different segmentation/editing tools in Slicer to determine which ones can most efficiently segment 2mm needles (10-15 cm long), from MRI scans in which voxels are about 2mm in each dimension. (These needles are not straight lines, and are often placed within a few mm of each other.)

Progress

Needles were effectively modeled from pelvic MRI scans using the Tractography Fiducial Seeding module:

  • Scans were cropped to the region of interest
  • Hessian images were generated from this data
  • Fiducials were placed in regions known to be part of a needle
  • The Tractography Fiducial Seeding module was used to generate “tracts” that pass through the fiducial – in this case, the needles

Refinements need to be made to optimize the modeling for brachytherapy needles rather than fiber bundles:

  • A sense of “momentum” is needed, so that needles do not abruptly end or change direction in signal voids
  • Existing needle and template models (eg, from iGyne) may also be incorporated


Delivery Mechanism

N/A

References