FetMRQC SR

FetMRQC SR is the super-resolution extension of FetMRQC [paper1,paper2] is a tool for quality assessment (QA) and quality control (QC) of T2-weighted (T2w) fetal brain MR images.

It builds on top of the utilities developed in the FetMRQC repository, a tool for the QC of low-resolution T2w scans.

It contains the tools needed

It consists of two parts.

1. A rating interface (visual report) to standardize and facilitate quality annotations of T2w fetal brain MRI images, by creating interactive HTML-based visual reports from fetal brain scans. It uses a pair of low-resolution (LR) T2w images with corresponding brain masks to provide snapshots of the brain in the three orientations of the acquisition in the subject-space.
2. A QA/QC model that can predict the quality of given super-resolution reconstructed volumes.

Given a list of SRR images listed using qc_list_bids, it then uses srqc_segmentation to compute the segmentations using BOUNTI [1] and extracts image quality metrics (IQMs) using srqc_compute_iqms. These IQMs can then be transformed in FetMRQC SR predictions using srqc_inference.

If you have found this useful in your research, please cite

Thomas Sanchez, Vladyslav Zalevskyi, Angeline Mihailov, Gerard Martí-Juan, Elisenda Eixarch, Andras Jakab, Vincent Dunet, Mériam Koob, Guillaume Auzias, Meritxell Bach Cuadra. (2025) Automatic quality control in multi-centric fetal brain MRI super-resolution reconstruction. arXiv preprint arXiv:2503.10156

Installing FetMRQC_SR

To install FetMRQC SR, just create a new conda environment with python 3.9.0

conda create --name fetmrqc_sr python=3.9.0

Then, simply activate the environment and install fetmrqc_sr and its dependencies by running pip install -e .

Using FetMRQC_SR for inference

To run FetMRQC_SR on your data, please follow these steps:

1. Given a BIDS-formatted dataset, get a CSV list of the data with qc_list_bids (use --help to see the detail). You will need to use the option --skip_masks.
2. Compute brain segmentations using srqc_segmentation and IQMs using srqc_compute_iqms.
3. Run inference using the pre-trained FetMRQC_SR model using srqc_inference.

Custom model building using FetMRQC SR

1. Gathering manual annotations

The first step towards building a custom FetMRQC_SR model is gathering manual annotations. For this, we provide a code for building manual QC reports. After installing fetmrqc_sr, you will need to follow these steps to generate manual QC reports.

1. Given a BIDS-formatted dataset, get a CSV list of the data with qc_list_bids (use --help to see the detail). You will need to use the option --skip_masks.
2. Once you have your csv file, you can generate the visual reports for manual annotations using

qc_generate_reports --bids_csv <csv_path> --out_dir <output_directory> --sr

3. You can then run qc_generate_index to generate an index file to easily navigate the reports.

Note. We recommend following our protocol for manual quality rating. It is available on Zenodo.

If you find this useful or use it in your research, please cite our manual quality rating paper [2] as well as the Zenodo protocol [3].

2. Custom model training using FetMRQC SR

Once your manual ratings are done, you then train a custom QC model as follows.

1. Get back a CSV file using qc_ratings_to_csv in the folder where your ratings are stored.
2. Compute brain segmentations using srqc_segmentation and IQMs using srqc_compute_iqms.
3. Train your custom models using the manual ratings with automatically extracted IQMs using srqc_train_model.

References

[1] Uus, Alena U., et al. "BOUNTI: Brain vOlumetry and aUtomated parcellatioN for 3D feTal MRI." bioRxiv (2023).

[2] Sanchez, T., et al. "Assessing data quality on fetal brain MRI reconstruction: a multi-site and multi-rater study." International Workshop on Preterm, Perinatal and Paediatric Image Analysis. Cham: Springer Nature Switzerland, 2024.

[3] Bach Cuadra, M., et al. "Protocol for the Quality Rating of 3D Super-resolution Reconstruction in Fetal Brain MRI", 20 août 2025, p. 46‑56, https://doi.org/10.5281/zenodo.15696638.

License

Copyright 2025 Medical Image Analysis Laboratory.

Acknowledgements

This project was supported by the ERA-net Neuron MULTIFACT – SNSF grant 31NE30_203977.