-
Notifications
You must be signed in to change notification settings - Fork 157
recon all QC tips
We will start with the outputs from "MNIfsaverage32k" qcmode. In these views, the surfaces have been transformed into MNI space using FSL fnirt.
This is actually quite useful for visual QC because:
- All subjects are oriented the same way - which makes outliers easier to spot.
- Some errors in tissue classification will get exaggerated by the MNI transform - making them easier to spot here.
If the qc pages are on your local system you can view them using your browser
firefox /path/to/hcp/data/qc_MNIfsaverage32k/index.html
My QC workflow is:
- Scroll through the CombineView Index Pages
- if any image looks odd, click on that subject to view the single subject view and investigate further
- Scroll through the Sagittal Surface Outline View
- if any image looks odd, click on that subject to view the single subject view and investigate further
Throughout this process. Make notes about any poor scans in a separate document.
The single subject views shows all snapshots taken from this subject in one page. The views are (from top to bottom).
- aparc: A surface reconstruction (medial and lateral views)
- SurfOutlineAxial: Axial anatomical slices with the surfaces shown (white surface - blue, pial surface - lime green)
- SurfOutlineCoronal: Coronal anatomical slices with the surfaces shown (white surface - blue, pial surface - lime green)
- SurfOutlineSagittal: Sagittal anatomical slices with the surfaces shown (white surface - blue, pial surface - lime green)
- CombinedView: the surface reconstruction on top of the anatomical. Lateral (L and R) and Dorsal and Vental views.
This index page shows the surface reconstruction of the brain (labeled using the aparc atlas) with the T1w Image. You see the brain from the 1) Left Side, 2) Right Side 3) Top View 4) Bottom View
Every line is a separate subject. The freesurfer subject ID is printed below the image. To look at any subject in more detail - click on the image.
In the image above all subjects pass visual QC.
If recon-all did not finish. The surface files might not exist... so they are not plotted.
A completely black image may be seen is recon-all failed very early in the pipeline.. For a very poor quality anatomical. The surface will look shrivelled up.
Sometimes when brain masking fails during recon-all in a part of the brain, than area of the brain will be stretched strangely in the underlying image.
If the gray matter is missing part of the occipital lobe, the back of the brain will look split apart on the bottom view (far right)
The is the second place to look. Like the Combined View - index page, here we see one line per subject, with the the Freesurfer subject id printed below. Also, like all index pages, you can go to the single subject view by clicking on any subject's image.
In the above image, all of the participants pass visual QC.
A key place to look are the two temporal poles. Surface reconstruction in these area can fail.
For this participant, the surface reconstruction in the aparc view looks jagged (especially in the orbital frontal cortex)
The temporal pole is not ideal.