FreeSurfer 6.0: Longitudinal Workflows

Summary

The longitudinal FreeSurfer workflows are designed with the intention of reducing the confounding effects of inter-individual and inter-time point variability as compared with the default cross-sectional recon-all pipeline. The user may input any number of timepoints from a longitudinal data series, and through incorporating additional (temporal) information obtain robust and reliable cortical and subcortical morphological measurements. The longitudinal scheme is designed to be unbiased with respect to any time point.

Please note that for these workflows it is necessary to process all time points cross-sectionally beforehand.

The longitudinal workflow(s) involve:

  • Algorithm initialization through creation of an unbiased template volume.
  • Cross-sectional processing of template.
  • Longitudinal processing of all timepoints, including:
    • Estimated Total Intracranial Volume
    • Automated whole-brain volumetry and cortical thickness calculation.
    • Full cortex parcellation according to Desikan-Killiany-Tournville atlas.
    • Subcortical volumetry
    • Optional full-head T2 -or- T2-FLAIR input to improve pial surface estimation.
    • Automated segmentation of four additional brainstem structures: medulla oblongata, pons, midbrain and superior cerebellar peduncle.
  • Comprehensive volumetric report.

For more information on FreeSurfer's longitudinal stream, see here.

Volumetric report output examples


Figure 1 - Freesurfer recon-all segmentation over time point 1

Figure 2 - Freesurfer recon-all segmentation over time point 2

Required inputs:

  • Previous cross-sectional recon-all analysis executed using at least one time point of the same individual.
  • T1
    • Anatomical 3D image
    • Isotropic resolution recommended
    • Must be labeled as 'T1' modality

Optional inputs:

  • T2
    • Anatomical 3D image
    • Isotropic resolution recommended
    • Must be labeled as 'T2' modality
  • T2 FLAIR
    • Anatomical 3D image
    • Isotropic resolution recommended
    • Must be labeled as 'T2' modality and have a ‘flair’ tag

Minimum input requirements:

  • Image must contain full skull (not brain extracted).
  • For optimal results reliability, isotropic resolution is recommended.
    • Recommended resolution: 1 mm isotropic.
    • Minimum reliable resolution: 2 mm isotropic.

Outputs:

  • Workflow main output:
    • report.pdf: report file with results summary.
    • Aseg_stats CSV files: Freesurfer volumetry and thickness complete results for each time point.
    • aparc+aseg/aseg.nii.gz files for each time point.
    • Volumetric.csv: CSV file with gray matter parcellated regions volume and cortical thickness information for each time point.
  • Template generation step:
    • data_file.json: details of which time point corresponds to which SSID
    • SUBJECT TEMPLATE folder: analogous to main FREESURFER recon-all script output folder
  • Longitudinal processing steps:
    • input_tpX.long.SUBJECT_TEMPLATE: analogous to main FREESURFER recon-all script output folder, where X corresponds to the timepoint being processed.
  • Brainstem Substructures output:
    • T1.nii.gz
    • brainstemSs_voxelspace.nii.gz
    • brainstem_substructures_volume.csv
    • FREESURFER output folder

Typical execution time:

  • ~ 20 hours for a workflow with two timepoints.

References:


License information:

  • Users of this tool should be in possession of a FreeSurfer license. Obtaining a license is free and comes in the form of a license.txt file.
  • More information on the FreeSurfer license can be found here.
  • Follow this link to obtain a license key.