Skip to content

Latest commit

 

History

History
132 lines (81 loc) · 10.1 KB

README.md

File metadata and controls

132 lines (81 loc) · 10.1 KB

Docker BRATS!

All relevant information regarding the BraTS challenge and detailed instruction for the tasks in general can be found here: http://braintumorsegmentation.org

If you have any problems, please open an issue (preferred) or contact us via email.

TL;DR:

Head over to our template folder and download it to build from there. All instructions are in the files and the README.md there: https://github.com/BraTS/Instructions/tree/master/template

Keep reading for the extended version.

Introduction

In a nutshell: We would like to have your algorithms in a Docker container, as well as in their original source code. We intend to run your dockerized algorithm on the BraTS 2020 test dataset to compare segmentation results as part of the BraTS manuscript, and to make all contributed Docker containers available through the BraTS algorithmic repository. Your source code will not be distributed and will only be used internally by the BraTS organizers, as a proof of code ownership (contact us if you cannot share your source code).

Your algorithm should be able to generate a tumor segmentation on any multimodal brain scan that is preprocessed like a BraTS test subject. To allow for fair comparisons and assessing performance differences across algorithms, you are expected to indicate what training set you have used, during training and/or design of your algorithm, and in the case of private datasets, a description of those datasets.

Docker containers

To ensure maximum compatibility and distributability, we want to use a container approach with Docker. We therefore want you to put your runnable code in a Docker container in accordance with the requirements listed below.

In most cases, containerization of your code is a simple and straightforward procedure. We will provide an example container with functional code soon. Furthermore, many containers are available on the internet to be used as a basis, probably including your favourite programming environment and neuroimaging tools.

Containerization

The concept of containerization is to simplify the deployment of applications. Docker is a technique to do so and will be used for the MICCAI BRATS challenge.

Docker can be used to “wrap” your entire segmentation method (including all dependencies and the operating system) into a single container. This container can be run as if it would be a single standalone application, anywhere, on any platform. Because your method and all dependencies are included in the container, the method is guaranteed to run exactly the same all the time.

Dockerization is a very popular concept and has been used successfully in previous MICCAI challenges, e.g., in the MSSEG challenge (https://portal.fli-iam.irisa.fr/msseg-challenge/overview) and the WMH Segmentation Challenge (http://wmh.isi.uu.nl/methods/). The current text is also based on info from the WMH Segmentation Challenge. Docker Hub (https://hub.docker.com/) provides a large overview of existing Docker containers (base images), that can be used to build your own container. Furthermore, many popular programming environments and image analysis methods have Dockerfiles available.

Step for step instructions

  1. Create a script for your segmentation code that reads the 4 imaging sequences (Format: .nii.gz) from /data, performs the segmentation and writes the result back to /data/results in nii.gz format.
  2. Create a Dockerfile that uses a base image, installs all necessary dependencies and then copies your code and the script from step 1 to the container.
  3. Run docker build .< Dockerfile in the directory where your Dockerfile, your code and your script reside.
  4. After building, your container should be able to take input via the -v <host_directory>:/data mount command and perform your segmentation.

The full run command can be seen below. I suggest you read the introduction to Docker on their website: https://docs.docker.com/get-started/part2/

Further instructions can be found here:

Caffe and Matlab: If you are using Caffe in combination with Matlab, I suggest to build an image using the Caffe image as base image and then installing the Matlab Runtime inside the container.

Data access

All test sets will be identical to the 2020 training/validation/test data that you have already processed. They are co-registered, skull-stripped, resampled to 1mm^3 isotropic resolution, and aligned to the SRI space. Data will be in NIfTI GZIP Compressed Tar Archive (.nii.gz) format, with all header information except the spatial resolution removed, and the individual volumes will be named with the case ID followed by *_flair.nii.gz, *_t1ce.nii.gz, *_t1.nii.gz, *_t2.nii.gz. You can use any of the BRATS training or testing image volumes to check whether your Docker image runs as expected.

Because your container runs in an isolated environment, the data needs to be mapped into the container. The input data files, i.e., the *_flair.nii.gz, *_t1ce.nii.gz, *_t1.nii.gz, *_t2.nii.gz volumes will be linked to /data and your segmentations must be placed in /data/results. Results should be a NIfTI file with the same resolution as the input data. Please call the resulting file tumor_your_image_class.nii.gz, where your_image is an eight character identifier of your algorithm (id of your container, title of your code/project). If your algorithm returns probabilities as well, you can return them accordingly, and name them, e.g., tumor\_'your_image'\_prob_4.nii.gz for results of class 4. If your algorithm returns tissue classes, please use tissue_your_name_wm.nii.gz for white matter (_gm.nii.gzand_csf.nii.gz` for the other two).

There should be no interaction with the container required other than running the Docker command below.

Your code must accept the files exactly as shown here:

  • *_flair.nii.gz
  • *_t1ce.nii.gz
  • *_t1.nii.gz
  • *_t2.nii.gz

For the data format instructions, see this file here: https://github.com/BraTS/Instructions/blob/master/data_formats.md

Computing environment & Resources

We will run your container on a selection on test cases. Docker can set resource limits on containers. Please give us an indication how many CPU cores and how much RAM is needed for you method, and what the resulting computation time will be (so we can estimate a timeout).

GPU computation

If your model runs faster on a GPU, please use Nvidia-Docker as the basis for your container. When submitting, please also submit the minimum GPU requirements needed for inference. For Instructions see below:

NVIDIA-Docker

The procedure of creating a Docker container with support for GPU computations is similar to the approach for normal Docker containers. To create a NVIDIA-Docker container, please follow the instructions from NVIDIA themselves: https://devblogs.nvidia.com/nvidia-docker-gpu-server-application-deployment-made-easy/

Github Repo: https://github.com/NVIDIA/nvidia-docker

We will be using Nvidia-Docker 2 to test containers with GPU support, so please make sure that a call with docker run --runtime=nvidia ... works (even though the most recent version of the Docker engine includes native GPU support).

Your still need to follow the interface specifications provided above. Please also submit performance indicators (computation time for the GPU/CPU used in your own tests) and the appropriate nvidia-docker run command in addition to the other infos required for a standard Docker implementation (see above).

If you have additional questions or troubles using NVIDIA-Docker, please contact us via e-mail (see bottom of this file) or open an issue (preferred).

Examples

To help you containerize your segmentation method with Docker, we will provide some examples: see the BraTS Algorithmic Repository for last year's implementations and run them to see how they work.

Assistance

If you are unsure whether your method can be containerized or how to proceed, please contact us in advance. We will try to help you with Docker.

Also, be sure to check out the documentation on the Docker website.

What we need from you

  • Your version of the Docker run command below with the name of your container and the script call
  • Your requirements as far as resource usage is concerned
  • Additional notes regarding functionality
  • Links to Docker Hub or another platform where the image is available

Docker commands

Your container will be run with the following commands:

docker run --rm −v <directory>:/app/data −it <your image>

"directory" will be our test directory containing the four modalities and the empty folder for your results.
"your image" is the name of your Docker image.

Our test system

  • Ubuntu 16.04.6 LTS
  • Docker version 19.03.0, build aeac949
  • GPU: Nvidia P100 PCIe, 16 VRAM
  • CPU: Intel(R) Xeon(R) CPU E5-2680 v4 @ 2.40GHz
  • RAM: a lot, but please try to keep your usage in check

Please try to keep the resource usage to a minimum, we want to be able to use the Docker containers on as many systems as possible (with lower specs).

Notes regarding this repository

Purpose

This repository should provide some files and documents to aid in the containerization of BRATS methods.

Additional documents

This repository also contains the legacy Docker interface definition (PDF - OUTDATED) for Docker containers taking part in the MICCAI BRATS 2017 competition and a markdown explaining the details for the participants. Please only follow the instructions here.

Please contact me in case of any questions or if you have remarks regarding the documents provided.

c.berger [at] tum [dot] de