FastSurfer-sMRI-preprocessing

Bash script to batch process structural MRI (sMRI) brain scans using FastSurfer's segmentation pipeline

---

Setting up FastSurfer

FastSurfer can be installed using several different ways, but the following script was primarily developed, tested, and run using FastSurfer on Docker on WSL 2 on a Windows 10 host machine. However, this script will also work with FastSurfer on Docker on a native Linux installation as well. The script was developed, tested, and run on a machine with the following specifications:

Operating System (OS)	Processor (CPU)	RAM	Graphics Card (GPU)	Storage
Windows 10	Intel Core i9-8950 HK @ 2.90 GHz	16.0 GB	NVIDIA GeForce GTX 1070 (8 GB)	500 GB SSD + 1.82 TB HDD

Steps to run the run_fastsurfer_custom_sample.sh bash script from an installation from Docker:

1. Install Docker for the respective host Operating System that you are running on by following the instructions on the Docker website.

2. Run the following command in an Ubuntu terminal: docker pull deepmi/fastsurfer:gpu-latest. Make sure that the Docker engine is first running, otherwise the command may not work. Refer to DockerDocs to help debug any Operating System-specific issues. This command will pull the latest Docker image of FastSurfer with GPU support. If running on CPU, utilize the following command: docker pull deepmi/fastsurfer:latest.

3. Download the ADNI dataset and place all of the images inside of a folder called ExtractedAll.

4. List the absolute file paths of all the images that you would like to run through the pipeline within the images_paths.txt file with the file path of each image on a separate line. Since this script was run on ADNI T1-weighted sMRI images, this script handles the images best in the folder/file structure that ADNI organized them with all of the DICOM slices for one image in a directory named with the image ID.

   • Note: This script assumes that the paths in the images_paths.txt file are all in the Windows-style path format. The script will convert these Windows-style paths to Unix-style ones as one of the first steps. In order to run this on Unix-style paths, there are two options: either convert all of the paths to Windows-style paths with the wslpath -w <path-to-folder-with-dcm-files> command or change line 24 of the script.

5. Run the following command to run the script: ./run_fastsurfer_custom_sample.sh

6. The output is stored in a folder called ExtractedAllOutput which is created at the same directory level as ExtractedAll.