Deep Learning (DL) RF-2016-02364081 dataset for the study titled: ‘Optimizing performance of transformer-based models for fetal brain MR image segmentation’.
Description
The dataset includes 172 subjects for a total of 519 fetal rs-fMRI scans with respective brain segmentations. Brain segmentations were obtained with the RS-FetMRI package (https://github.com/NicoloPecco/RS-FetMRI). For each subject the gestational week at scan is reported as the last two digits of the file name. The RF-2016-02364081 DL dataset has been used to train and test deep learning architectures (i.e. Swin-UNETR, UNTER, CNN, GAN) on a fetal brain extraction task for the study titled ‘Optimizing performance of transformer-based models for fetal brain MR image segmentation’. Pretrain weights of Swin-UNETR best model can be found on the ‘weight’ folder (Fetal_pretrain.pth).
Files
Steps to reproduce
The DL RF-2016-02364081 dataset can be used with the Swin-UNETR transformer model and other deep learning models in python using MONAI and Pytorch. All the steps and codes for the usage of the dataset are reported in https://github.com/NicoloPecco/Swin-Functional-Fetal-Brain-Segmentation.
Institutions
Categories
Funding
Ministero della Salute
RF-2016-02364081