woctezuma / playing-with-simpleitk

Playing with SimpleITK and nnU-Net to process data from the CHAOS challenge on Google Colab.

Playing with SimpleITK and nnU-Net

This repository contains Python code to play with SimpleITK and nnU-Net.

Requirements

• Install Python 3.X,
• Install the required packages:

pip install -r requirements.txt

Data

• Download data from the Combined (CT-MR) Healthy Abdominal Organ Segmentation (CHAOS) challenge,
• Extract the archive to data/.

Alternatively, data can be found in Releases on Github.

Usage

Prepare data

• Convert data from DICOM to NifTI with SimpleITK:

python convert_to_nii.py

• Follow instructions for the folder structure of data, as detailed for nnU-Net.

Train nnU-Net

• Run the playing_with_pytorch.ipynb notebook.

Training was performed with:

• a training dataset of 5 images downsampled to 4x4x4mm,
• no cross-validation (in which case the validation dataset is the same as the training dataset to find the best checkpoint and to decrease the learning rate).

Training for 1 epoch can require 2m30s, but up to 10m, depending on the machine which you obtained.

Results obtained after ~ 100 epochs are shown below:

where:

• training loss is in blue,
• validation loss is in red,
• evaluation metric (to maximize) is in green.

In terms of Dice scores, the average is:

• 99.1% for the 5 patients of the training data,
• 94.8% for the 15 patients of the validation data (training data not used).

References

• Github repository for SimpleITK,
• Notebooks for SimpleITK,
• Documentation for SimpleITK,
• nnU-Net,
• Combined (CT-MR) Healthy Abdominal Organ Segmentation (CHAOS) challenge,
• Evaluation of segmentation results with tools provided by organizers of CHAOS.

playing-with-simpleitk's Issues

Dependency Dashboard

This issue lists Renovate updates and detected dependencies. Read the Dependency Dashboard docs to learn more.

This repository currently has no open or pending branches.

Detected dependencies

github-actions

.github/workflows/pythonapp.yml

• actions/checkout v4
• actions/setup-python v5
• codecov/codecov-action v4

pip_requirements

requirements.txt

• SimpleITK >=1.1.0

---

• Check this box to trigger a request for Renovate to run again on this repository

Initial Update

The bot created this issue to inform you that pyup.io has been set up on this repo.
Once you have closed it, the bot will open pull requests for updates as soon as they are available.