==== SynthStrip dataset ======================================================= The SynthStrip dataset is a collection of full-head images, brain masks, and label maps from 622 MRI, PET, and CT scans gathered across an array of public sources. We created it to evaluate the SynthStrip brain-extraction tool: Hoopes A, Mora JS, Dalca AV, Fischl B*, Hoffmann M* (*equal contribution) SynthStrip: skull-stripping for any brain image NeuroImage, 260, 119474, 2022 https://doi.org/10.1016/j.neuroimage.2022.119474 Website: https://w3id.org/synthstrip ==== Dataset breakdown ======================================================== The scans in this data collection span across imaging modalities, resolutions, and subject populations, ranging from infants to patients with glioblastoma. Their distribution is as follows (see reference above for details). Dataset Modality Voxel size (mm^3) Images ========================================================= IXI T1w MRI 0.9 x 0.9 x 1.2 50 T2w MRI 0.9 x 0.9 x 1.2 50 PDw MRI 0.9 x 0.9 x 1.2 50 MRA 0.5 x 0.5 x 0.8 50 DWI 1.8 x 1.8 x 2.0 32 --------------------------------------------------------- FSM T1w MPRAGE 1.0 x 1.0 x 1.0 38 T2w 3D-SPACE 1.0 x 1.0 x 1.0 36 PDw 3D-FLASH 1.0 x 1.0 x 1.0 32 qT1 MP2RAGE 1.0 x 1.0 x 1.0 32 --------------------------------------------------------- ASL T1w MPRAGE 1.0 x 1.0 x 1.0 43 PCASL 2D-EPI 3.4 x 3.4 x 5.0 43 --------------------------------------------------------- QIN T1w 2D-FLASH 0.4 x 0.4 x 6.0 54 T2-FLAIR 2D-FSE 0.4 x 0.4 x 6.0 17 T2w 2D-FSE 1.0 x 1.0 x 5.0 39 --------------------------------------------------------- Infant T1w MPRAGE 1.0 x 1.0 x 1.0 16 --------------------------------------------------------- CIM (*) FDG PET 2.0 x 2.0 x 2.0 20 CT 0.6 x 0.6 x 1.5 20 --------------------------------------------------------- Total: 622 (*) As we cannot redistribute the PET and CT images from the CERMEP-IDB-MRXFDG dataset (CIM), we only include the corresponding brain masks in the native space of these scans. To obtain the images, please contact the original authors or fill out their request form: Ines Merida (merida@cermep.fr) https://framaforms.org/request-form-for-cermep-idb-mrxfdg-database-1637844711 ==== Dataset structure ======================================================== Each subdirectory represents an individual scan. Subdirectory names follow a SUBSET_MODALITY_SUBJECT format. We distribute the unmodified image from the source dataset and a manually refined binary brain mask. For adult MPRAGE scans, the dataset also includes a label map. All image files have the native shape, resolution, and orientation of the respective scan: Filename Scans Content ==================================================================== image.nii.gz all except CIM unmodified image mask.nii.gz all binary brain mask labels.nii.gz adult MPRAGE only automatically fitted label map The label maps include standard anatomical SynthSeg/FreeSurfer labels and six non-brain labels of value 100-105. These non-brain labels are not necessarily anatomically meaningful or consistent across scans: their purpose was to generate variability outside the brain for SynthStrip training. The specific labels are defined in the FreeSurfer colormap seg_labels.txt, which you can use to visualize the label maps in FreeView. For example: cd ixi_t1_002 freeview image.nii.gz labels.nii.gz:colormap=lut:lut=../seg_labels.txt ==== 2D subset ================================================================ The 2D subset comprises sagittal slices with approximate IA axis orientation (inferior-anterior), extracted from the above images for all scans except CIM. Since image geometries vary across scans, we extract sagittal slices 16 mm to the left of the brain-mask center, without interpolating. Because the source images have not been registered, some of the corresponding 2D label maps can include right-hemispheric labels. ==== Validation split ========================================================= In the SynthStrip publication, we use the following scans for validation, holding out all other images for testing (see files scans_*.txt). ixi_t1_016 ixi_t1_017 ixi_t2_016 ixi_t2_017 ixi_pd_016 ixi_pd_017 ixi_mra_016 ixi_mra_017 fsm_t2_02cd fsm_t2_24dz fsm_pd_02cd fsm_pd_24dz fsm_qt1_02cd fsm_qt1_24dz asl_t1_110 asl_t1_120 asl_epi_110 asl_epi_120 qin_t1_01 qin_t1_02 qin_flair_01 qin_flair_02 qin_t2_01 qin_t2_02 ==== License ================================================================== You can use the IXI images under a Creative Commons Attribution-ShareAlike 4.0 International Licence (CC BY-SA 4.0). We distribute all other content of the SynthStrip dataset under a Creative Commons Attribution 4.0 International License (CC BY 4.0). https://creativecommons.org/licenses/by-sa/4.0/ https://creativecommons.org/licenses/by/4.0/ ==== References =============================================================== If you find this dataset useful, please cite the SynthStrip reference at the top and any relevant references below to credit the original authors of the source datasets we obtained the images from. Information eXtraction from Images (IXI): IXI Dataset Biomedical Image Analysis Group, Imperial College London https://brain-development.org/ixi-dataset FreeSurfer Maintenance (FSM): A deep learning toolbox for automatic segmentation of subcortical limbic structures from MRI images Greve DN, Billot B, Cordero D, Hoopes A, Hoffmann M, Dalca AV, Fischl B, Iglesias JE, Augustinack JC NeuroImage, 244, 118610, 2021 https://doi.org/10.1016/j.neuroimage.2021.118610 In-house HCP-A pseudo-continuous arterial spin labeling (ASL): Extending the Human Connectome Project across ages: Imaging protocols for the Lifespan Development and Aging projects Harms MP, Somerville LH, Ances BM, Andersson J, Barch DM, et al. NeuroImage, 183, 972-84, 2018 https://doi.org/10.1016/j.neuroimage.2018.09.060 Characterizing cerebral hemodynamics across the adult lifespan with arterial spin labeling MRI data from the Human Connectome Project-Aging Juttukonda MR, Li B, Almaktoum R, Stephens KA, Yochim KM, Yacoub E, Buckner RL, Salat DH NeuroImage, 230, 117807, 2021 https://doi.org/10.1016/j.neuroimage.2021.117807 QIN GBM Treatment Response from The Cancer Imaging Archive (QIN): Mamonov AB, Kalpathy-Cramer J Data From QIN GBM Treatment Response The Cancer Imaging Archive https://doi.org/10.7937/K9/TCIA.2016.nQF4gpn2 Prah MA, Stufflebeam SM, Paulson ES, Kalpathy-Cramer J, Gerstner ER, Batchelor TT, Barboriak DP, Rosen BR, Schmainda KM Repeatability of Standardized and Normalized Relative CBV in Patients with Newly Diagnosed Glioblastoma American Journal of Neuroradiology, 36 (9), 1654-61, 2015 https://doi.org/10.3174/ajnr.A4374 Clark K, Vendt B, Smith K, Freymann J, Kirby J, Koppel P, Moore S, Phillips S, Maffitt D, Pringle M, Tarbox L, Prior F The Cancer Imaging Archive (TCIA): Maintaining and Operating a Public Information Repository Journal of Digital Imaging, 26 (6), 1045-57, 2013 https://doi.org/10.1007/s10278-013-9622-7 Infant subjects from Boston Children's Hospital (Infant): De Macedo Rodrigues K, Ben-Avi E, Sliva DD, Choe MS, Drottar M, Wang R, Fischl B, Grant PE, Zollei L A FreeSurfer-compliant consistent manual segmentation of infant brains spanning the 0-2 year age range Frontiers in Human Neuroscience, 9, 21, 2015 https://doi.org/10.3389/fnhum.2015.00021 CERMEP-IDB-MRXFDG (CIM): Merida I, Jung J, Bouvard S, Le Bars D, Lancelot S, Lavenne F, Bouillot C, Redoute J, Hammers A, Costes N CERMEP-IDB-MRXFDG: a database of 37 normal adult human brain [18F]FDG PET, T1 and FLAIR MRI, and CT images available for research EJNMMI Research, 11, 91, 2021 https://doi.org/10.1186/s13550-021-00830-6 ==== Changes ================================================================== 2022-03-16 (v1.0): * Released initial dataset. 2022-03-17 (v1.1): * Updated README and data subsets included. 2022-07-26 (v1.2): * Added the Infant data. * Listed the validation scans used in the publication. 2022-12-16 (v1.3): * Added the FSM data and CIM masks. * Added missing asl_epi_115 image. * Specified license. 2023-09-07 (v1.4): * Added label maps for adult MPRAGE scans. * Released a 2D version of the dataset. * Included FreeView colormap and subject lists.