nnUNet EUCAIM Docker

Copyright (c) German Cancer Research Center (DKFZ). All rights reserved. Please make sure that your usage of this code is in compliance with its license. This project is supported by EUCAIM grant number is 101100633.

This repository provides Docker images for running nnUNet inference in a reproducible and containerized environment, tailored for the EUCAIM project.

---

Setup

To build all task-specific nnUNet images, run:

docker build  --build-arg USER_UID=2323  --build-arg USER_GID=2323 -t "nnunet:base" nnunet-base/
docker build -t harbor.eucaim.cancerimage.eu/processing-tools/nnunet-pancreas-tumour:latest nnunet-task-pancreas/
docker build -t harbor.eucaim.cancerimage.eu/processing-tools/ nnunet-abdominal-organ-segmentation:latest nnunet-task-abdominal/
docker build -t harbor.eucaim.cancerimage.eu/processing-tools/ nnunet-brain-tumour-pet:latest  nnunet-task-abdominal/
docker build -t harbor.eucaim.cancerimage.eu/processing-tools/ nnunet-colon-cancer:latest nnunet-task-abdominal/
docker build -t harbor.eucaim.cancerimage.eu/processing-tools/ nnunet-liver-tumour:latest nnunet-task-abdominal/

You may optionally specify the user UID and GID to avoid permission issues when running containers.

Image Hierarchy

• nnunet:base: Built first, this image includes all the necessary environment dependencies. Uses the official NVIDIA CUDA Ubuntu 22.04 runtime: nvidia/cuda:11.8.0-runtime-ubuntu22.04.
• nnunet-<organ task>:latest: Built on top of nnunet:base, each task-specific image includes the corresponding pretrained model checkpoint (downloaded at build time).

Usage

Each container expects two mounted directories for I/O:

• Input: /home/eucaim/nnUNet_input
• Output: /home/eucaim/nnUNet_output

Your DICOM series folders containing .dcm files should be placed inside the input directory. The container processes them sequentially and writes the segmentation outputs to the output directory using the same filenames.

Expected dataset structure:

dataset/
├── study_001/
│   ├── series_001/
│   │   ├── IMG0001.dcm
│   │   ├── IMG0002.dcm
│   │   └── ...
│   ├── series_002/
│   │   ├── IMG0001.dcm
│   │   ├── IMG0002.dcm
│   │   └── ...
│   └── ...
├── study_002/
│   ├── series_001/
│   │   ├── IMG0001.dcm
│   │   ├── IMG0002.dcm
│   │   └── ...
│   └── ...
└── ...

Example

To run inference using the image for Liver Tumor, execute:

docker run \
  -v /host/dataset/study_001:/home/eucaim/nnUNet_input \
  -v /host/data_out:/home/eucaim/nnUNet_output \
  --gpus all \
  nnunet-liver-tumour:latest

Data Format Conventions

• All input must as DICOM series folders.
• All files in the input folder will be processed in batch.
• Output files are written with identical filenames into the output folder.

Notes

A patched version nnUNet source is placed in the nnunet-base folder for installation purposes at build time. As soon as the issue MIC-DKFZ/nnUNet#2876 is fixed, direct git clone from official would be possible.