FLAIR templates

FLAIR templates

In this page you will find a set of brain templates constructed using fluid-attenuated inverse recovery (FLAIR) images, which can be used, for instance, to help to automatically identify white matter hyperintensities (WMH; e.g. leukoaraiosis, multiple sclerosis) from magnetic resonance images (MRI), or as target for registration of other individual FLAIR images without the need of a matching T1-weighted image, using software such as FSL and SPM.

Two versions are available, based on the full or on a restricted sample, each of these in three different resolutions (0.7 mm, 1.0 mm and 2.0 mm). The difference between the full sample and the restricted sample is that in the full sample, all subjects available at the time of the construction of the templates were used (N=853), whereas the restricted sample only subjects that presented none or the least volumes of white matter hyperintensities were used (see more details below). In addition to FLAIR, templates for gray matter (GM), white matter (WM), cerebrospinal fluid (CSF) and T1-weighted (T1) are also provided for both samples.

Full sample (N=853) Restricted sample (N=366)
FLAIR GG-853-FLAIR-0.7mm.nii.gz GG-366-FLAIR-0.7mm.nii.gz
GG-853-FLAIR-1.0mm.nii.gz GG-366-FLAIR-1.0mm.nii.gz
GG-853-FLAIR-2.0mm.nii.gz GG-366-FLAIR-2.0mm.nii.gz
GM GG-853-GM-0.7mm.nii.gz GG-366-GM-0.7mm.nii.gz
GG-853-GM-1.0mm.nii.gz GG-366-GM-1.0mm.nii.gz
GG-853-GM-2.0mm.nii.gz GG-366-GM-2.0mm.nii.gz
WM GG-853-WM-0.7mm.nii.gz GG-366-WM-0.7mm.nii.gz
GG-853-WM-1.0mm.nii.gz GG-366-WM-1.0mm.nii.gz
GG-853-WM-2.0mm.nii.gz GG-366-WM-2.0mm.nii.gz
CSF GG-853-CSF-0.7mm.nii.gz GG-366-CSF-0.7mm.nii.gz
GG-853-CSF-1.0mm.nii.gz GG-366-CSF-1.0mm.nii.gz
GG-853-CSF-2.0mm.nii.gz GG-366-CSF-2.0mm.nii.gz
T1 GG-853-T1-0.7mm.nii.gz GG-366-T1-0.7mm.nii.gz
GG-853-T1-1.0mm.nii.gz GG-366-T1-1.0mm.nii.gz
GG-853-T1-2.0mm.nii.gz GG-366-T1-2.0mm.nii.gz

All templates can be downloaded in just one file, here: GG-all.tar (447 MB).

Information about the sample

Subjects are participants of the Genetics of Brain Structure and Function Study (GOBS), a collaborative effort involving the Texas Biomedical Research Institute, the Yale University School of Medicine and the University of Texas Health Science Center at San Antonio. The participants are of Mexican–American ascent, and members of extended families (pedigrees). For the full sample (N=853), the mean age was 43.85 years (std. dev.: 14.81, min.: 18, max.: 85), with 515 females. For the restricted sample (N=366), the mean age was 35.44 years (std. dev.: 11.11, min.: 18, max.: 69), with 205 females.

Information about the MRI acquisition protocol

The images were acquired in either of two Siemens magnetom Trio 3 T scanners, one of these being a also Tim system. The aquisition sequences were:

  • The FLAIR (Fluid-Attenuated Inverse Recovery, T2-weighted) sequence used echo time (TE): 1.8 ms, inversion time (TI): 353 ms, repetition time (TR): 5000 ms, flip angle 180°, echo train length: 221, voxel size: 0.5 x 0.5 x 1.0 mm.
  • The MPRAGE (Magnetisation Prepared Rapid Gradient Echo, T1-weighted) sequence used TE: 3.04 ms, TI: 785 ms, TR: 2100 ms, flip angle 13°, voxel size: 0.8 x 0.8 x 0.8 mm. A total of 7 acquisitions were performed and averaged to improve the signal-to-noise ratio (SNR; details in Kochunov et al., 2006, and here).

Overview of the image processing

The templates were constructed as part of the pipeline described in Winkler et al. (2009) for segmentation of white matter hyperintensities. In short, the FLAIR and the T1-weighted images for each subject were coregistered to each other, corrected for magnetic field inhomogeneities, and normalised to the MNI template (from the Montreal Neurological Institute).

After normalisation, voxels belonging to the white matter were selected and used as reference for a global intensity normalisation, to ensure that all subjects would have similarly ranged intensities. The normalisation used only the central region of the image histogram, such that the effect of eventual lesions would have minimal or no impact on this step. The full sample template (N=853 subjects) was produced by averaging all the images, whereas the template with a restricted sample (N=366 subjects) used only those that contained few or no white matter hyperintensities.

All the image processing steps used the FLAIR Segmentation Toolbox described in Winkler et al. (2009), which runs on SPM5 software. The normalisation uses the Unified Segmentation Algorithm (Ashburner et al., 2005).

Projects using these templates

  • FLAIR Segmentation Toolbox, by Anderson Winkler and colleagues, described here (no longer maintained).
  • Clinical Toolbox for SPM, by Chris Rorden and colleagues, available here.

How to cite

If you use these templates in your research, please cite it as:

Other references

Availability

These templates are available through the Glahn Group website and mirrored at the Brainder website.

Version history

  • 27.Nov.2012: Updated the center of the coordinate system to match the anterior commissure (AC), making registration easier in certain applications. Thanks to Torben E. Lund (Aarhus University, Denmark) for pointing out.
  • 25.Aug.2012: Templates updated to larger samples (N=853 and N=366), in three different resolutions (0.7, 1.0 and 2.0 mm), and available for FLAIR and T1-weighted modalities, as well as GM, WM and CSF partitions.
  • 09.May.2011: First public release of the FLAIR templates, using two samples sizes (N=462 and N=181), with resolution of 1 mm.