https://atlas.rsna.org/schemas/2025-11/dataset.json

Ospedale San Raffaele fetal resting-state fMRI brain extraction dataset (internal) and external OpenNeuro sample

https://ordr.hsr.it/datasets/dyg9dpmgvs/1

1

Italian Ministry of Health’s Ricerca Finalizzata 2016 grant (RF-2016-02364081).

Conducted in accordance with the Declaration of Helsinki and approved by the Ospedale San Raffaele Ethics Committee (registration no. EK Nr.2174/2016). All women provided written informed consent prior to MRI examination.

Internal retrospective dataset used to train and evaluate transformer-based segmentation models for fetal brain extraction from resting-state fMRI; external dataset from OpenNeuro used for generalization and cross-validation.

Provide data to evaluate transformer-based methods for fetal brain extraction from rs-fMRI and to study effects of pretraining, dataset size, and input size, including generalization across scanners and gestational weeks.

Internal retrospective cohort collected 2018–2022; external OpenNeuro dataset independently sampled.

Internal dataset split at fetus level: 90:10 train:test; training further split 80:20 into train:validation with balanced gestational weeks and scanner distribution. Training downsampled to 66% and 33% subsets for experiments. External dataset split analogously (train/validation/test).

Exact imaging file formats, detailed acquisition parameters, and preprocessing specifics are provided only in supplemental Appendix S1 and not fully detailed in the article text.

Multiple scans per fetus (519 scans from 172 fetuses internally; 561 scans from 131 fetuses externally).

An independent external sample from OpenNeuro (131 fetuses; 561 scans) was used for testing and cross-validation.

Retrospective clinical imaging data with informed consent; institutional ethics approval obtained.

Fetal MRI data.