Brain_MRI_Tumor

This project explores the possibility of removing (by warping) the pathology area of a tumor-presenting brain MR image so that an automatic sub-cortical segmentation method designed from healthy brain can handle it.

Working Environment

0 - Prerequisites This setup instruction assume you are working on an ubuntu 18.04 or 20.04 system. The simplest way to setup all dependencies for this project is using anaconda. If you haven't install conda in your system, please first do so by following the instruction in https://docs.anaconda.com/anaconda/install/linux/

1 - Conda Environmrnt If you have already installed conda, create a new environment for this project by

conda env create -f Working_Environment/volume_segmentation_environment.yml

volume_segmentation_environment.yml is a list of required packages that could be installed by conda. I will keep update this file through this project. After the environment is created, activate it by conda activate volume_segmentation.

2 - Pymesh Pymesh is a powerful python tool for processing mesh data. To make it working properly, one have to compile it and its dependencies in local. The download like and install instruction could be found here: https://pymesh.readthedocs.io/en/latest/installation.html This project use Pymesh to do many mesh operations. One of the most important is tetrahedralizing surface meshes.

3 - FreeSurfer FreeSufer is an essential software tool for processing medical Images. You can find its download and install instruction here: https://surfer.nmr.mgh.harvard.edu/fswiki/DownloadAndInstall This project use FreeSurfer to make affine registration (images from different datasets to a common brain space).

4 - TumorSim TumorSim is a cross-platform simulation software that generates pathological ground truth from a healthy ground truth. You can find its download and install instruction here: https://www.nitrc.org/projects/tumorsim/ We relay on this software to generate training data. Edit TumorSim_executable in Working_Environment/environment_variables.py to point to the binary executable of TumorSim.

5 - SamsegTumor The original SamsegTumor.py in freesurfer/python/packages/freesurfer/samseg should be replaced by our modified version at Modified_SamsegTumor/SamsegTumor.py.

6 - PCNN displacement network The code for training the PCNN displacement network is in another repo: https://github.com/jiameng1010/pcnn_brain_tumor.

Datasets

1 - BraTS BraTS 2020 dataset could be download from: https://www.med.upenn.edu/cbica/brats2020/data.html Once extracted, edit BraTS_dataset_dir in Working_Environment/environment_variables.py to point to where BraST is stored.

2 - Mindboggle Mindboggle-101 dataset could be download from: https://mindboggle.info/data Once extracted, edit Mindboggle_dataset_dir in Working_Environment/environment_variables.py to point to where Mindboggle-101 is stored.

3 - TumorSim Input data Some synthetic brain MR data is released together with TumorSim software package. You should download #1, #2, #3, #4, and #5 of the data from: https://www.nitrc.org/frs/?group_id=546 As the same, edit BrainSim_inputdata_dir in Working_Environment/environment_variables.py to point to where these files are stored.