- ATLAS v3.0 manuscript (in preparation)
- ATLAS v3.0 is comprised of a training dataset (n=1453), and a generalizability dataset with completely hidden test data (n=362).
- The training dataset (n=1453) has been released publicly.
- Both the minimally preprocessed data in native space (containing brain extracted images only) and preprocessed dataset (in MNI space) are archived in INDI (https://fcon_1000.projects.nitrc.org/indi/retro/atlas_download.html).
- Hidden lesion segmentation masks from the generalizability dataset are only available to lesion segmentation challenges upon request.
- The ATLAS v2.0 public release of the Training (n=655) and Validation Datasets (n=300) have been expanded upon and replaced by the ATLAS v3.0 dataset.
- More information on our most recent ISLES '26 lesion segmentation challenge can be found here: https://isles-26.grand-challenge.org/
- A narrated walkthrough of the training material can be accessed on our YouTube channel: https://www.youtube.com/playlist?list=PLPqz6WxiWPVE
- Our complete lesion tracing guidance document along with the neuroanatomy training slides can be accessed here: https://github.com/npnl/ATLAS/tree/master/Lesion%20Tracing%20Training
- The corresponding imaging and mask data are divided into two sets, one for each stage of training, download them here:
- [Set 1] : https://fcp-indi.s3.amazonaws.com/data/Projects/INDI/ATLAS/R3.0/training_set_1.zip
- [Set 2] : https://fcp-indi.s3.amazonaws.com/data/Projects/INDI/ATLAS/R3.0/training_set_2.zip
- Scientific Data paper (Liew et al., 2022): https://www.nature.com/articles/s41597-022-01401-7
- ATLAS v2.0 is comprised of a Training Dataset (n=655), Testing Dataset (n=300), and a Generalizability Dataset (completely hidden test dataset; n=316)
- The public release (n=955) includes the Training Dataset (T1w MRIs + lesion masks) and Testing Dataset (T1s MRIs only)
- Preprocessed dataset (n=955) archived on INDI: http://fcon_1000.projects.nitrc.org/indi/retro/atlas.html
- Raw, native space dataset (n=955) archived on ICPSR: https://www.icpsr.umich.edu/web/ADDEP/studies/36684/versions/V4
- Hidden lesion masks from the Testing Dataset, and the full Generalizability Dataset, are only available to lesion segmentation challenges upon request
- Check out our current segmentation challenge: https://atlas.grand-challenge.org/
- Read about ATLAS 1.0 in our Scientific Data paper (Liew et al., 2018): https://www.nature.com/articles/sdata201811
- Full dataset (native n=304 | subset in standard space n=239): http://www.icpsr.umich.edu/icpsrweb/ICPSR/studies/36684
- The standardized, defaced dataset (n=229) has now been replaced by the ATLAS v2.0 dataset