CMRxMotion Task 1

Code example for MICCAI2022 challenge CMRxMotion task 1 - Cardiac MRI motion artefacts classification

Paper: Motion-related Artefact Classification Using Patch-based Ensemble and Transfer Learning in Cardiac MRI

Challenge website: http://cmr.miccai.cloud/

Usage

Requirements

Python 3.9
PyTorch 1.11

pip install -r requirement.txt

Model Training

1. Train heart region segmentation model (train_seg.py) on CMRxMotion task 2 data.
2. Crop the 3D scans into 2D slices and segment the heart region assisted by the segmentation model (pred_crop.py).
3. Divide the slices into 5-fold for cross-validation for classification model training (split_slices.py).
4. Train classification models (train_seg.py, train_seg.py, train_seg.py) on 2D slices.

Model inference

Crop the validation data into 2D slices and segment the heart region (pred_crop.py), then run inference.py.