E2MIP Challenge, MICCAI 2024

Starting point for E2MIP challenge, task 3:

2D Segmentation of an unknown dataset (fetal brain MRI)

This repository can be used as a starting point for the E2MIP challenge on the fetal brain MRI dataset with Monai and PyTorch.

This repository contains:

1. information about the submission for the task 3 of the challenge
2. the structure of the training and testing data folders that will be used for your submission
3. few examples of the data

1. Challenge submission:

The following information is preliminary. Final submission information will be available soon.

• Please provide your complete algorithm for training and predicting in a docker script. Further information about how the script should look like will be published here soon.
  • script takes as input the path to the "training_data" and "testing_data" folders
  • script outputs predicted segmentation in a newly created folder "testing_data_prediction". The predictions need to be filed in the folder in a certain folder structure (see 2. Data for Challenge)
• Besides the performance metric also the energy consumption during training and evaluation is being measured and both determine the Challenge ranking.
• Submissions for the E2MIP Challenge are now being accepted at https://cmt3.research.microsoft.com/E2MIP2023/.
• The deadline for final method submissions is September 17th. Late submissions might be possible, but we cannot guarantee the execution of the submission.

2. Data for Challenge, Task 3:

The folder structure of the training and testing data used for evaluating your code will look like the following:

*important: The data in "train_slice" folder are 2D slices with different resolutions and The data in "test_volume" folder are 3D stacks with different resolutions ***Your saved predictions in "test_volume_prediction" are also corresponding stack prediction of "test_volume" **

train_slice
├── images
│   ├── DWI_case0000_slice00.nii.gz
    ...
│   ├── FMRI_case0000_slice00.nii.gz
    ...
│   ├── T2W_case0000_slice00.nii.gz
    ...
│   
├── masks
│   ├── DWI_case0000_slice00_mask.nii.gz
    ...
│   ├── FMRI_case0000_slice00_mask.nii.gz
    ...
│   ├── T2W_case0000_slice00_mask.nii.gz
    ...

test_volume
├── DWI_case0008.nii.gz
...
├── FMRI_case0003.nii.gz
...
├── T2W_case0005.nii.gz
...

The folder structure of the segmentation predictions that your script should create from "testing_data" should have the following structure:

test_volume_prediction
├── DWI_case0008_maskpred.nii.gz
...
├── FMRI_case0003_maskpred.nii.gz
...
├── T2W_case0005_maskpred.nii.gz
...

3. Sample Data

Please see the "data" folder.

For further questions about this code, please contact razieh.faghihpirayesh@childrens.harvard.edu with subject "E2MIP Challenge, task 3"