Difference between revisions of "AHM2009:PNL"

From NAMIC Wiki
Jump to: navigation, search
 
(19 intermediate revisions by 4 users not shown)
Line 3: Line 3:
 
__NOTOC__
 
__NOTOC__
 
==PNL Roadmap Project==
 
==PNL Roadmap Project==
 
 
{|
 
{|
|[[Image:Helix.png|thumb|280x150px|Stochastic Tractography on helix phantom]]
+
|[[Image:Helix.png|thumb|Stochastic Tractography on helix phantom]]
|[[Image:Wmm.png|thumb|280x150px|White Matter Mask generated from phantom]]
+
|[[Image:Wmm.png|thumb|White Matter Mask generated from phantom]]
|[[Image:STArcuate.jpg|thumb|280x150px|Stochastic Tractography of Arcuate Fasciculus]]
+
|[[Image:STArcuate.png|thumb|Stochastic Tractography of Arcuate Fasciculus]]
 +
|[[Image:Step1.png|thumb|56px|Stochastic Tractography Module]]
 +
|[[Image:Step2.png|thumb|152px|Stochastic Tractography Module]]
 
|}
 
|}
  
 
== Overview ==
 
== Overview ==
* What problem does the pipeline solve?
+
;*What problem does the pipeline solve?
Most tractography methods (Deterministic/Principal Diffusion Direction) estimate fibers by tracing the maximum direction of diffusion. A limitation of this approach is that, in practice, several factors introduce uncertainty in the tracking procedure, including, noise, splitting and crossing fibers, head motion and image artifacts.  
+
Most tractography methods (Deterministic/Principal Diffusion Direction) estimate fibers by tracing the maximum direction of diffusion.  
Stochastic tractography attempts to quantify the uncertainty associated with estimated fibers, by using a propagation model based on stochastics and regularization, which allows paths originating at one point to branch and return a probability distribution of possible paths.  
+
 
 +
A limitation of this approach is that, in practice, several factors introduce uncertainty in the tracking procedure, including, noise, splitting and crossing fibers, head motion and image artifacts.  
 +
 
 +
Stochastic tractography attempts to quantify the uncertainty associated with estimated fibers, by using a propagation model based on stochastics and regularization, which allows paths originating at one point to branch and return a probability distribution of possible paths.
 +
 
Based on probability functions, using a sequential importance sampling technique ([http://lmi.bwh.harvard.edu/papers/pdfs/2002/bjornemoMICCAI02.pdf Bjornemo et al., 2002]), one can generate thousands of fibers starting in the same point by sequentially drawing random step directions. This gives a very rich model of the fiber distribution, as contrasted with single fibers produced by conventional tractography methods. Moreover, from a large number of sampled paths, probability maps can be generated, providing better estimates of connectivity between several anatomical locations.  
 
Based on probability functions, using a sequential importance sampling technique ([http://lmi.bwh.harvard.edu/papers/pdfs/2002/bjornemoMICCAI02.pdf Bjornemo et al., 2002]), one can generate thousands of fibers starting in the same point by sequentially drawing random step directions. This gives a very rich model of the fiber distribution, as contrasted with single fibers produced by conventional tractography methods. Moreover, from a large number of sampled paths, probability maps can be generated, providing better estimates of connectivity between several anatomical locations.  
* Who is the targeted user?
+
;* Who is the targeted user?
 
Since diffusion direction uncertainty within the gray matter is quite significant; principal diffusion direction approaches usually do not work for tracking between two gray matter regions. Thus if one requires finding connections between a priori selected anatomical gray matter regions, defined either by anatomical segmentations (in case of using structural ROI data), or functional activations (in case of megring DTI with fMRI), stochastic tractography seems to be the method of choice.  
 
Since diffusion direction uncertainty within the gray matter is quite significant; principal diffusion direction approaches usually do not work for tracking between two gray matter regions. Thus if one requires finding connections between a priori selected anatomical gray matter regions, defined either by anatomical segmentations (in case of using structural ROI data), or functional activations (in case of megring DTI with fMRI), stochastic tractography seems to be the method of choice.  
 
Stochastic Tractography is also comparable, if not better, in defining large white matter fiber bundles, especially those traveling through white matter regions characterized by increased diffusion uncertainty (fiber crossings).  
 
Stochastic Tractography is also comparable, if not better, in defining large white matter fiber bundles, especially those traveling through white matter regions characterized by increased diffusion uncertainty (fiber crossings).  
* How does the pipeline compare to state of the art?
+
;* How does the pipeline compare to state of the art?
To date, there are two popular toolboxes that include probabilistic/stochastic tractography modules. Functionality of both modules (FSL and PICO) is limited, and software is not open source.  
+
To date, there are two popular toolboxes that include probabilistic/stochastic tractography modules. Functionality of both modules (FSL and PICO) is limited, and software is not open source.
  
 
==Detailed Information about the Pipeline==
 
==Detailed Information about the Pipeline==
Can be found in:
+
Stochastic Tractography pipeline has been written in Python, and is part of current Slicer3 release.
MODULES > PYTHON MODULES > PYTHON STOCHASTIC TRACTOGRAPHY
+
 
Functionality of Python Stochastic Tractography module in Slicer 3.0
+
Functionality of Python Stochastic Tractography module in Slicer 3.0 includes:
* Loading volumes:
+
;*Preprocessing
Module reads files (DWI and ROIs) in nhdr format.
+
 
* Smoothing:
+
Reading DWI and ROI files in nhdr format
One can smooth the DWI data (only Gausian smoothing is supported at this time). We recommend it if the data is noisy.
+
 
* Masks:
+
Smoothing DWI data
Masks are removing some of the unwanted tracts, same way principal diffusion direction tractography uses stopping criteria for tractography termination. Three different methods for masking are provided.
+
 
:1. Otsu Mask- a thresholding method that is based on T2 intensity
+
Creating brain and white mater masks
:2. WM Mask- a thresholding method based on FA intensity
+
 
:3. Artifact removal- one can use its own WM mask, obtained from T1 or T2 segmentation, and registered to DWI space, then artifact removal compares WM mask with FA map, and patches WM mask with voxels that have high FA values.
+
Removing artifacts in WM masks by comparing this mask with FA map
*Tensor estimation:
+
 
Module uses two methods for estimating tensors, and can output anisotropy indices (FA/Mode/Trace), if requested
+
Producing diffusion indices (FA, Mode, Trace)
*Stochastic Tractography:
+
 
Parameters that need to be adjusted:
+
;*Creating probability maps (parameters involve number of tracts per seed, tract length, step size, stopping criteria)
:1. The amount of tracts that will be seeded from each voxel (we recommend between 500 and 1000 tracts, depending on the workstation power- 1000 tracts per voxel seeded within the large ROI for high resolution DWI can take a long time to compute).
 
:2. maximum tract length (in mm), this can eliminate long, unwanted tracts if the regions for which connection is measured are located close to each other
 
:3. Step Size: distance between each re-estimation of tensors, usually between 0.5 and 1 mm
 
:4. Spacing???
 
:5. Stopping criteria. This can be used on the top of WM mask, to terminate tracts (in case they really want to travel through the CSF, for example).
 
*Probability Map:
 
This step creates output probability maps.
 
:1. rough: each voxel is counted only once if at least one fiber pass through it
 
:2. cumulative:Tracts are summed by voxel independently
 
:3. discriminative: tracts are summed by voxel depending on their length
 
  
Then, probability maps can be saved as ROIs, and either used directly, or thresholded (at certain probability, step claimed by few publications to remove noise) in slicer to mask and compute average FA, Mode, Trace for entire connection. Diffusion indices can be also weighted by the probability of connection for each voxel.
+
Probability maps can be saved as ROIs, and either used directly, or thresholded (at certain probability, step claimed by few publications to remove noise) in slicer to mask and compute average FA, Mode, Trace for entire connection. Diffusion indices can be also weighted by the probability of connection for each voxel.
  
* Demonstrate parameters/steps that need to be adjusted using someone else's data
+
;* Parameters/steps that need to be adjusted using someone else's data
 +
Software was tested on 3 sets of data (1.5T anisotropic Siemens, 3T anisotropic GE, 3T isotropic GE), and works for all of them without major modifications.
  
 
==Software & documentation==
 
==Software & documentation==
 
* Download Slicer 3.0 [http://www.slicer.org/pages/Downloads here].
 
* Download Slicer 3.0 [http://www.slicer.org/pages/Downloads here].
*[[Media:IJdata.tar.gz|Training Dataset]]
 
*[[Media:Slicer3STModule.ppt|Training Presentation]]
 
*[[Media:SlicerSTtutorial.ogg|Training Webcast]] ([http://www.theora.org Theora format])
 
  
==Team==
+
:* Module documentation can be found here:
[[Image:PnllogoworkNEW.png|left|50px]]
+
:**[[Media:IJdata.tar.gz|Training Dataset]]
* DBP: Marek Kubicki, MD, PhD, Sylvain Bouix, PhD, Julien Von Siebenthal, PhD, Doug Terry, BS
+
:**[[Media:PNLStochasictNew.ppt|Training Presentation]]
Psychiatry Neuroimaging Laboratory, Department of Psychiatry, Brigham and Women's Hospital, Harvard Medical School
+
:**[[Media:Helix.zip|Sample Helix Dataset]]
Boston, MA
 
  
 +
:* Software that you will also need to launch:
 +
:**[http://www.python.org/download/releases/2.5.4/ Python 2.5 Software]
 +
:**[http://sourceforge.net/project/showfiles.php?group_id=1369&package_id=175103 NumPy Software]
 +
:**[http://pnl.bwh.harvard.edu/NAMIC/Slicer3-build.zip Slicer 3 Build for Windows]
 +
:**[http://pnl.bwh.harvard.edu/NAMIC/Slicer3-lib.zip Slicer 3 Lib for Windows]
  
[[Image:csail.jpg|left|50px]]
+
==Team==
* Core 1: Polina Gollard, MIT
 
 
 
 
 
[[Image: Kitware.png|left|80px]]
 
* Core 2: Brad Davis, Steve Pieper, Kitware
 
  
 +
{|
 +
|[[Image:Logo_pnl2.png|75px]]
 +
|valign="top"| DBP: Marek Kubicki, MD, PhD, Sylvain Bouix, PhD, Julien De Siebenthal, PhD, Doug Terry, BS,
 +
[http://pnl.bwh.harvard.edu/ Psychiatry Neuroimaging Laboratory]
 +
|-
 +
|[[Image:csail.jpg|75px]]
 +
|valign="top"|Core 1: Polina Golland, [http://www.csail.mit.edu/ CSAIL MIT]
 +
|-
 +
|[[Image: Kitware.png|75px]]
 +
|valign="top"| Core 2: Brad Davis, Steve Pieper, [http://www.kitware.com/ Kitware]
 +
|}
  
 
==Outreach==
 
==Outreach==
* Visit our [http://pnl.bwh.harvard.edu/pub/all.html publication database.]
+
;* Visit our [http://pnl.bwh.harvard.edu/pub/all.html publication database.]
* Planned outreach activities (including presentations, tutorials/workshops) at conferences
+
;* Planned outreach activities (including presentations, tutorials/workshops) at conferences
 +
:1. Method of stochastic tractography, along with the module functionality will be presented at the DTI symposium during World Biological Psychiatry Symposium in Florence, Italy in April 2009.
 +
:2. Clinical applications, including results of arcuate fasciculus study in schizophrenia will be presented at the Biological Psychiatry Congress in Vancouver, Canada in May 2009.

Latest revision as of 17:33, 17 February 2009

Home < AHM2009:PNL

Back to AHM 2009 Agenda


PNL Roadmap Project

Stochastic Tractography on helix phantom
White Matter Mask generated from phantom
Stochastic Tractography of Arcuate Fasciculus
Stochastic Tractography Module
Stochastic Tractography Module

Overview

  • What problem does the pipeline solve?

Most tractography methods (Deterministic/Principal Diffusion Direction) estimate fibers by tracing the maximum direction of diffusion.

A limitation of this approach is that, in practice, several factors introduce uncertainty in the tracking procedure, including, noise, splitting and crossing fibers, head motion and image artifacts.

Stochastic tractography attempts to quantify the uncertainty associated with estimated fibers, by using a propagation model based on stochastics and regularization, which allows paths originating at one point to branch and return a probability distribution of possible paths.

Based on probability functions, using a sequential importance sampling technique (Bjornemo et al., 2002), one can generate thousands of fibers starting in the same point by sequentially drawing random step directions. This gives a very rich model of the fiber distribution, as contrasted with single fibers produced by conventional tractography methods. Moreover, from a large number of sampled paths, probability maps can be generated, providing better estimates of connectivity between several anatomical locations.

  • Who is the targeted user?

Since diffusion direction uncertainty within the gray matter is quite significant; principal diffusion direction approaches usually do not work for tracking between two gray matter regions. Thus if one requires finding connections between a priori selected anatomical gray matter regions, defined either by anatomical segmentations (in case of using structural ROI data), or functional activations (in case of megring DTI with fMRI), stochastic tractography seems to be the method of choice. Stochastic Tractography is also comparable, if not better, in defining large white matter fiber bundles, especially those traveling through white matter regions characterized by increased diffusion uncertainty (fiber crossings).

  • How does the pipeline compare to state of the art?

To date, there are two popular toolboxes that include probabilistic/stochastic tractography modules. Functionality of both modules (FSL and PICO) is limited, and software is not open source.

Detailed Information about the Pipeline

Stochastic Tractography pipeline has been written in Python, and is part of current Slicer3 release.

Functionality of Python Stochastic Tractography module in Slicer 3.0 includes:

  • Preprocessing

Reading DWI and ROI files in nhdr format

Smoothing DWI data

Creating brain and white mater masks

Removing artifacts in WM masks by comparing this mask with FA map

Producing diffusion indices (FA, Mode, Trace)

  • Creating probability maps (parameters involve number of tracts per seed, tract length, step size, stopping criteria)

Probability maps can be saved as ROIs, and either used directly, or thresholded (at certain probability, step claimed by few publications to remove noise) in slicer to mask and compute average FA, Mode, Trace for entire connection. Diffusion indices can be also weighted by the probability of connection for each voxel.

  • Parameters/steps that need to be adjusted using someone else's data

Software was tested on 3 sets of data (1.5T anisotropic Siemens, 3T anisotropic GE, 3T isotropic GE), and works for all of them without major modifications.

Software & documentation

  • Download Slicer 3.0 here.

Team

Logo pnl2.png DBP: Marek Kubicki, MD, PhD, Sylvain Bouix, PhD, Julien De Siebenthal, PhD, Doug Terry, BS,

Psychiatry Neuroimaging Laboratory

Csail.jpg Core 1: Polina Golland, CSAIL MIT
Kitware.png Core 2: Brad Davis, Steve Pieper, Kitware

Outreach

  • Visit our publication database.
  • Planned outreach activities (including presentations, tutorials/workshops) at conferences
1. Method of stochastic tractography, along with the module functionality will be presented at the DTI symposium during World Biological Psychiatry Symposium in Florence, Italy in April 2009.
2. Clinical applications, including results of arcuate fasciculus study in schizophrenia will be presented at the Biological Psychiatry Congress in Vancouver, Canada in May 2009.