External Email - Use Caution
Hi Freesurfers,
We are collaborating with a group at Michigan that planss to use Freesurfer's longitudinal pipeline to process T1 data collected from children (see brief project summary below) at multiple time points. Our role is to oversee recon QA and application of a custom cortical parcellation labeling system. As we move to longitudinal processing, we have a couple questions about when/how/if to apply various custom templates that we're using.
In preparing the cross-sectional recons, we used a custom tal atlas generated from our sample with help from Avi Snyder.
Question 1: When creating the base or running longitudinal processing, should this atlas be specified in some manner?
We also generated a custom pediatric surface template and cortical classifier; we co-register each subject to this template and apply the classifier (we actually just use surf2surf to map the labeled template to the individual subject surfaces) to generate cortical labels.
Question 2: how should we do this in the longitudinal stream? would we register the subject's base to the pediatric surface template then use the the resulting sphere.reg in the call to mris_ca_label?
In the longitudinal work flow, it looks like cortical labeling is done in the longitudinal stream.
Question 3: If we notice a labeling error and correct it, is there a means of propagating that correction across all the time points?
Question 4: If we've labeled and reviewed all the cross-sectional time-point-one data, is there a way to map those labels to the longitudinal stream time points that doesn't compromise the resulting labels?
Question 5: Given the project summary below, are there any other concerns that we should be aware of when applying the longitudinal pipeline?
*Brief summary of the longitudinal study of childhood stuttering at University of Michigan (Soo-Eun Chang)*
Over the last few years, we have collected yearly structural and functional MRI scans from around 300 children who do and do not stutter. These are children between the ages of 3-10 at initial scan. Scans were acquired each year for up to 4 years (some children do not have all 4 years, but many do). Data from year 1 has been published (Garnett et al., 2018) using a custom labeling system and atlas. We have just finished manual editing of the scans from years 2 – 4 and are preparing to conduct a longitudinal analysis with all 4 years of scans in FreeSurfer.
We plan to extract cortical and subcortical measures of thickness, area, volume, LGI, etc. and compare developmental trajectories of these measures in the stuttering group vs. the control group. Also, because of the longitudinal design, we have information on whether children in the stuttering group went on to recover or if their stuttering persisted. All children were considered stuttering at their initial visit, and based on evaluations each year, we are able to retrospectively assign them to the persistent group or the recovered group. In this way, in addition to comparing morphometric measures between children who stutter and controls, we also hope to find neural markers that help predict persistence vs. recovery in childhood stuttering.
Many thanks for your input.
Jason