ECoG Recons

From OpenWetWare
Jump to navigationJump to search

Accessing Reconstructions

There are several types of reconstructions, and they are each trustworthy to a different degree. HWNI Cluster documents the entire file structure, which is standardized across patients, but we will discuss here the main places you can look for images that localize the electrodes on the brain.

  1. Reference files from the hospital. These files are all in the 2D_Images folder.
    • Intraop_Photos - These are photos from surgery of the electrodes as viewed through the craniotomy. These images can be used to make Intraop Recons for patients who don't have a post-implantation CT.
    • OR sketches - We often get hand-drawn sketches that are used in the OR to indicate electrode location. If they include the numbering orientation of the electrode components, they will be in Electrode_Maps, and otherwise they are stored in Off_Site_Recons.
    • Off_Site_Recons - These are electrode images, usually from the CT and MR (but not the full 3D model), done by our collaborators. They are located in the 2D_Images folder.
  2. 2D images of in-house reconstructions.
    • Rough_Recons - We always save screenshots from our first attempt at making reconstructions. These files are also in the 2D_Images folder.
    • Final recon images - Once we have a 3D recon that we deem "good enough" (we aim for a margin of error no more than 1cm), snapshots are saved in Recon/Patient_2D.
    • 2D Recons - We make a layered image with several different features (gyri color coding, removable numbers, colorable electrodes) that researchers can use to make images for publications or presentations. These images are also saved in Recon/Patient_2D. We also have documentation for how to use them.
  3. 3D Recons
    • Recon working files - The process of making 3D reconstructions involves creating a lot of files. Each session (i.e. attempt) to make a 3D recon keeps its working files in a Recon_date folder in 3D_Images.
    • Finalized 3D models - If you want to look at the 3D model of the patient's brain and electrode locations, you can find the final version in Recon/Patient_3D. We also have documentation for how to use them.
    • As per requested a How To Video has been made for how to open/view 3D Recons in BioImageSuite
    • MNI Recons - 3D models can be used to map several subjects into an average brain space. If the subject in question has already been transformed into MNI space, you can find the transform and MNI electrode coordinates in the Recon/MNI_3D folder.

How we do reconstructions

We have a basic reconstruction workflow that requires an anatomical MR scan, preferably T1, and a post-implantation CT scan showing the electrodes. Following is an outline of that workflow. If we have no CT, but have intra-operative photos, those can be used for a reconstruction as well. BioImagesuite has a wealth of information as well, here are some courses that they gave in pdf format on their webpage [1]

The first three steps can be done in any order:

  1. MR brain extraction is the process of extracting the skull from the brain. It is usually the first step since it allows several possible shortcuts to the full recon process. See Rough Recons for examples.
  2. CT MR registration creates a transform from the MR space to the CT space.
  3. CT mgrid placement creates a metadata file with the electrode locations in the coordinate space of the original CT.
  4. Mgrid transformation: Once there is a CT mgrid and a CT MR registration, we use the transform matrix from the registration to transform the mgrid into the MR space.
  5. 3D Recons: An mgrid file transformed into the MR space and a segmented MR brain are the two basic pieces that constitute a 3D reconstruction. Once we have these, we can load the 3D model and evaluate the reconstruction quality.
    • If the 3D model looks good, we will proceed to 2D reconstructions and 3D models in MNI space.
    • If the reconstruction is still visibly off, the nature of the inaccuracy should be documented in Recon Notes, and a screenshot of the model should be saved in 2D_Images/Rough_Recons
    • A flawed 3D reconstruction may require redo-ing one of the earlier steps to fix. See Troubleshooting Recons for examples of common inaccuracies.
    • If the reconstruction accuracy is substantially compromised by post-surgical fluid accumulation compressing the brain in the CT, we do a CT MR brain registration to correct for this.
  6. 2D Recons: Working with 3D models is overkill for what most researchers need. Since we usually want recon images from specific views, we make 2D Recons from the relevant standard views. These are layered images that are easy to customize during the course of analysis and for publications.
  7. MNI Recons: Many researchers want to generalize across subjects. MNI reconstructions are 3D Recons in an average brain space, so that researchers can load several patients' electrodes on one brain.

Data Tree

In Bioimagesuite3 there is an option to use a workflow manager called the Data Tree. This option is not available in bioimagesuite. The current release of Bioimagesuite3 is not very stable and crashes quite frequently, but upon the release of a more stable version of bioimagesuite3 the Data Tree option affords a consolidated workspace from which all files can be worked. Most likely the Data Tree will become the standard in future releases for working in between CT and MR files, therefore one should familiarize themselves with the Data Tree. The Data Tree is extensively documented in the Bioimagesuite documentation, [2] and a less technical rundown is offered on our own Data Tree wikipage.

Misc Documentation

Workflow Diagram

File:Knight Lab Recon Workflow.jpg

Additional Resources

Retrieved from "https://openwetware.org/mediawiki/index.php?title=ECoG_Recons&oldid=1047050"