Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

Appearance settings

matt-baugh/many-tasks-make-light-work

Open more actions menu

Repository files navigation

Many tasks make light work: Learning to localise medical anomalies from multiple synthetic tasks

Environment installation:

  • Once that's done, create a virtual environment with make_virtual_env.sh
  • Activate the environment with source multitask_method_env/bin/activate
  • Set paths for input/output files in multitask_method/paths.py

Alternatively, you can use the .devcontainer to run the code in a docker container which creates a virtual environment from the requirements.txt file.

Data

The HCP dataset is available at https://www.humanconnectome.org/study/hcp-young-adult

The BraTS 2017 dataset is available at https://www.med.upenn.edu/sbia/brats2017/registration.html

The ISLES 2015 dataset is available at https://www.smir.ch/ISLES/Start2015

The VinDr-CXR dataset is available at https://physionet.org/content/vindr-cxr/1.0.0/

Inter-dataset blending datasets

The MOOD dataset (for 3D inter-dataset blending) is available at https://www.synapse.org/#!Synapse:syn21343101/wiki/599515

The ImageNET dataset (for 2D inter-dataset blending) is available at https://www.kaggle.com/c/imagenet-object-localization-challenge/overview

Preprocessing

Preprocessing scripts are available in the multitask_method/preprocessing folder. Don't forget to set the paths for preprocessed data and predictions to be saved in multitask_method/paths.py

Training

With the environment activated, run python train.py ABSOLUTE_PATH_TO_EXPERIMENT_CONFIG FOLD_NUMBER

Experiment configs used for the paper are in the experiments folder.

Prediction and Evaluation

To generate predictions on the test set, run python predict.py ABSOLUTE_PATH_TO_EXPERIMENT_CONFIG.

To evaluate the predictions at the normal resolution, run python eval.py ABSOLUTE_PATH_TO_EXPERIMENT_CONFIG.

Result metrics are saved in the predictions folder as results.json.

To evaluate the predictions at CRADL's resolution, run python cradl_eval.py ABSOLUTE_PATH_TO_EXPERIMENT_CONFIG.

Reproducibility

Brain

Download the datasets from the above links Run the brain preprocessing script multitask_method/preprocessing/brain_preproc.py

When experimenting with training on T tasks, run:

For F in range(0, 5CT) run python train.py <path_to_repo>/experiments/exp_HCP_low_res_T_train.py F To produce predictions run: python predict.py <path_to_repo>/experiments/exp_HCP_low_res_T_train.py F To evaluate the predictions run: python eval.py <path_to_repo>/experiments/exp_HCP_low_res_T_train.py F

VinDr-CXR

Download the datasets from the above links Run the brain preprocessing script multitask_method/preprocessing/vindr_cxr_preproc.py When experimenting with training on T tasks, run: For F in range(0, 5CT) run python train.py <path_to_repo>/experiments/exp_VINDR_low_res_T_train.py F To produce predictions run: python predict.py <path_to_repo>/experiments/exp_VINDR_low_res_T_train.py F To evaluate the predictions run: python eval.py <path_to_repo>/experiments/exp_VINDR_low_res_T_train.py F

Cite this work:

@InProceedings{baugh2023manytasks,
    author="Baugh, Matthew
    and Tan, Jeremy
    and M{\"u}ller, Johanna P.
    and Dombrowski, Mischa
    and Batten, James
    and Kainz, Bernhard",
    title="Many Tasks Make Light Work: Learning to Localise Medical Anomalies from Multiple Synthetic Tasks",
    booktitle="Medical Image Computing and Computer Assisted Intervention -- MICCAI 2023",
    editor="Greenspan, Hayit
    and Madabhushi, Anant
    and Mousavi, Parvin
    and Salcudean, Septimiu
    and Duncan, James
    and Syeda-Mahmood, Tanveer
    and Taylor, Russell",
    year="2023",
    publisher="Springer Nature Switzerland",
    address="Cham",
    pages="162--172",
    isbn="978-3-031-43907-0"
}

Acknowledgements

(Some) HPC resources were provided by the Erlangen National High Performance Computing Center (NHR@FAU) of the Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU) under the NHR projects b143dc and b180dc. NHR funding is provided by federal and Bavarian state authorities. NHR@FAU hardware is partially funded by the German Research Foundation (DFG) – 440719683.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published
Morty Proxy This is a proxified and sanitized view of the page, visit original site.