Quick Start Guide for TOF Segmentation with Docker

Prerequisites:

Ensure Docker is installed and properly configured on your system.

Step 1: Build the Docker Image

Run the following command to build the Docker image for TOF segmentation:

docker build -t tofsegmentation:latest .

Step 2: Run the Docker Container

Execute the container using the command below, adjusting the paths as needed for your environment:

docker run --gpus 0 -v /path/to/input/data:/in -v /path/to/output/data:/out -v /path/to/model:/model  tofsegmentation:latest \
    --src_path /in --dst_path /out --model_path /model \
    --filter_series csBTFE_M2D --num_workers 4

• /path/to/input/data: Local directory containing input MRI data.
• /path/to/output/data: Local directory where the results will be saved.
• /path/to/model/: Local directory to the trained model.

To view all available commands and options, run the following command:

docker run tofsegmentation:latest -h

Output Details:

The results will be organized in the output folder under the following structure:

Subfolder for each patient (e.g., patientID):

• Contour_images: Contains JPG images of all slices with contours.
• predicted_TOF.nii.gz: The 4D cine with predicted segmentation masks.

Results file:

In the root of the output folder, you will find a CSV file named CMR_quantification_UZL.csv, containing quantification data with the following columns for each patient ID:

• patient
• time
• serie
• LV EDV [ml] (Left Ventricular End-Diastolic Volume)
• LV ESV [ml] (Left Ventricular End-Systolic Volume)
• LV EF [%] (Left Ventricular Ejection Fraction)
• LV mass [g] (Left Ventricular Mass)
• RV EDV [ml] (Right Ventricular End-Diastolic Volume)
• RV ESV [ml] (Right Ventricular End-Systolic Volume)
• RV EF [%] (Right Ventricular Ejection Fraction)
• RV mass [g] (Right Ventricular Mass)

Adjust the command and paths according to your specific data and model setup.

Download and Prepare the Model

1. Download the model

Access the model using the following link: model

2. Extract the Model

Unzip the downloaded main_folder.zip file:

unzip main_folder.zip

3. Set the Docker Mount Path

Ensure the extracted folder is accessible by the Docker container. When running the container, mount the directory as follows:

-v /path/to/main_folder:/model

---

Reference

This work is described in: Automated biventricular quantification in patients with repaired tetralogy of Fallot using a 3D deep learning segmentation model Tilborghs, Sofie et al. Journal of Cardiovascular Magnetic Resonance, Volume 0, Issue 0, 101092.