Longitudinal analysis

The longitudinal stream of FreeSurfer is currently overhauled. Please stay tuned ...

This page will describe the new version of the Longitudinal stream, while LongitudinalProcessingOld describes the old version. It is advised not to use the old version due to its bias with respect to time point 1.


Compared with cross-sectional studies, a longitudinal design can significantly reduce the confounding effect of inter-individual morphological variability by using each subject as his or her own control. As a result, longitudinal imaging studies are getting increased interest and popularity in various aspects of neuroscience. The default FreeSurfer pipeline is designed for the processing of individual data sets (cross-sectionally), and thus not optimal for the processing of longitudinal data series. It is an active research area at the Martinos Center for Biomedical Imaging, how to obtain robust and more reliable cortical and subcortical morphological measurements by incorporating additional (temporal) information in a longitudinal data series.

The longitudinal scheme is designed to be unbiased wrt. any time point. Instead of initializing it with information from a specific time point, a base/template volume is created and run through FreeSurfer. This template can be seen as an initial guess for the segmentation and surface reconstruction. The FreeSurfer cortical and subcortical segmentation and parcellation procedure involves solving many complex nonlinear optimization problems, such as the deformable surface reconstruction, the nonlinear atlas-image registration, and the nonlinear spherical surface registration. These nonlinear optimization problems are usually solved using iterative methods, and the final results are known to be highly sensitive to the selection of a particular starting point (a.k.a. algorithm initialization). It's our belief that by initializing the processing of a new data set in a longitudinal series using the processed results from the unbiased template, we can reduce the random variation in the processing procedure and improve the robustness and sensitivity of the overall longitudinal analysis. Such an initialization scheme makes sense also because a longitudinal design is often targeted at detecting small or subtle changes. Aditionally to the base template new probabilistic methods (temporal fusion) were introduced to further reduce the variablility of results across timepoints. For these algorithms it becomes necessary to process all timepoints cross sectionally first.

The longitudinal processing scheme is coded in the recon-all script via the "-long" flag. Use the -help flag for help on its options.

Here is a summary of the longitudinal workflow:

1) cross-sectionally process all time points with the default workflow (tpN is one of the timepoints):

recon-all -all -s <tpNsubjid> -i path_to_tpN_dcm

2) create the base template and process it cross sectionally:

  recon-all -base <baseid> -base-insubj <tp1subjid> -base-insubj <tp2subjid> ... -all

3) longitudinally process all timepoints:

  recon-all -long <tpNsubjid> <baseid> -all

4) do comparisons on results from 3), e.g. calculate differences between <tp1subjid>.long.<baseid> and <tp2subjid>.long.<baseid>

The longitudinal processing stream always produces output subject data containing <tpNsubjid>.long.<baseid> in the name (to help distinguish from the default stream).

In the following, we list the major differences between the two processing streams. We assume that the longitudinal series have two time-points, tp1 and tp2. Both have already been processed cross sectionally using the standard FreeSurfer recon-all. First we discuss the construction of the base template and then the longitudinal processing of tpN (step 3).

1. Creation of Base

In step 2 the base template is created. This is done by mri_robust_template, which constructs a mean/median T1.mgz volume together with the transforms that align each tp's T1.mgz volume with the template. Thus, all timepoints are brought to an unbiased common space. This template volume is then processed cross sectionally with the standart FreeSurfer stream. The only difference is that the norm.mgz is not created in the usuall way, but again computed as the mean/median of the norm.mgz of all time points. The corresponding transforms are more accurate than the transforms obtained from the T1 image. Once the base is processed it can be used to initialize many steps in the longitudinal stream (step 3).

2. Difference 1: Register tpN with base

The longitudinal scheme requires aligning the image data of tpN to base, but this map has allready been construced in a robust manner during the base creation (see also mri_robust_register which is used by mri_robust_template to construct the maps). Since all data sets come from the same subject, these rigid registrations with 7 dof (translation,rotation,intensity scaling) are enough to get a good alignment between the images. The registration result (a coordinate transformation matrix) and its inverse have been saved during the template construction and will be used to transfer information between time points in later steps.

...

More later


Original Author: MartinReuter