RAMPS - Resections And Masks in Preoperative Space

An automated pipeline that generates 3D masks of pre-operative tissue resected, leveraging existing software including FreeSurfer, SynthStrip, Sythnseg and ANTS to aid in this task.

How to setup the code

1. Clone this repository.
2. Create a virtual python 3.8.20 enviroment (e.g via conda https://docs.conda.io/projects/conda/en/latest/user-guide/tasks/manage-environments.html) and install the requirements.txt. Example of code to do this is via the following.

conda create -n RAMPenv python=3.8.20
conda activate RAMPenv
pip install -r requirements.txt

4. Clone the https://github.com/BBillot/SynthSeg repository and place it inside "Place_SynthSeg_here". No additional python modules has to be installed. Make sure to follow SynthSeg step 3 and place the additional models in the SynthSeg models folder. (Link to these additional models can be found here Here
5. Ensure that freesurfer is installed (https://surfer.nmr.mgh.harvard.edu/) here is the link to the freesurfer set up MAC tutorial and Linux tutorial . Also check that the system enviroment 'FREESURFER_HOME' is set up and works correctly. To make this easier Its reccomened that you add somthing like the following to your shell startup (see freesurfer documentation for more information)

export FREESURFER_HOME=$HOME/freesurfer
$ source $FREESURFER_HOME/SetUpFreeSurfer.sh

Run the command

Once all the python packages are installed and Freesurfer and SynthSeg is installed then you can test RAMP with the following

python /Path_to/RAMP.py </Path_to/Pre-OP-Scan.nii.gz> </Path_to/Post-OP-Scan.nii.gz> </Path_to_Output_Folder_file_path/> <Output_Prefix> <Hemisphere> <Lobe>

where:

• <Pre-OP-Scan.nii.gz> is the absolute path to the pre operation nii.gz file (which is the space that the mask will be drawn in)
• <Post-OP-Scan.nii.gz> is the absolute path to the post operation nii.gz file
• <Output_Folder_file_path> This is the folder path to the place where you want to store the outputs of this script
• <Output_Prefix> an ID to use in the naming of the scripts
• L or R. Is the hemisphere in which the resection took place either L or R
• any combination of T F O P . this is to select the lobe of resection. Example T will just be the temporal lobe while TF will look at the frontal and temporal lobe. Note for ease of use Temporal inludes the Temporal, subcortical and Insula region.

Example of how it works

Lets say we have a patient X which we see a resection takes place in the Right Frontal lobe, the command to run this will be.

python /Path_to/RAMP.py patient_X-PRE-OP-Scan.nii.gz patient_X-POST-OP-Scan.nii.gz patient_X-Output_Folder_file_path patient_X R F

How this code works

This code can be broken down into 3 steps -

• PREPARING which is the steps taken to get the pre and post scans ready for registration
• REGISTRATION which is code needed to align the post-op image into the pre-operative space
• CREATION which is the code that creates the mask once the images are alligned in the same space

A - PREPARING

• First convert the pre and post op images into orig resolution via ants resample_image_to_target
• Use N4bias to correct the bias feild
• mri_synthstrip is applied to strip the image up to the pial surface
• SynthSeg is used to segmentate the brain into a parcellation scheme - to become a mask of the brain and the lobes
• Use the Synthseg mask to the pial surface
• The use SynthSeg parcellation to group regions into the lobe.
• Dilate each lobe to get the attached White matter for each lobe
• Create a mask of the resected and none resected lobes
• Extract the ventricles
• Remove the top 1% of voxel intensity and replace them with the median no CSF voxel value

B - REGISTRATION

• Transform the cleaned post-operative image into the pre-operative space via the use of antsRegistrationSyN[br]

C - CREATION

• Rescale the pre and post operative between 0 and 1
• Create a mask of each pre and post op images
• Get the difference between the two mask
• move the post op vents into pre-op space
• use atropos to find the resection in the post-op tissue (thats in pre-op space) by looking for voxels that are similar to the vents
• Subtract the post-op image from the pre-op image
• Use atropos to exapnd the post-op resection mask through the Subtraction space
• filter to resected lobe , get largest object and dilate by 1 voxel, add the difference the pre and post mask
• This get resection mask 1
• errode the resection mask by 1
• dilate the resection mask and check what has been added to the mask. If the expansion is only a few voxels then we highlight that area as a region to advoid further expansion
• Repate this until the diffrence between dilation is only 100 voxels
• Expand the mask into voxels that are between the voxels in the mask and CSF voxel that is only 3 voxels away
• add the difference the pre and post mask
• This creates mask 2

Here is an image outlining how it works (rough figure being worked on)