NORMALISATION: HOW DOES IT WORK?


In Summary

Normalisation is based on ANTs software package. Please visit http://stnava.github.io/ANTs/ for a detailed description of their great work.


A good registration of your T1 to the MNI template is key if you wish to have optimal results. Therefore we added the new tool “normalisation” to our software package, which combined what we think being the best way to “normalise” your brain images.


Normalisation will calculate affine and diffeormorphic deformation to register your T1 to the MNI152 space. Adding an optional lesion directory will limit diffeormorphic deformation to areas located outside the lesioned tissue. This is important as the result might be biased by the presence of abnormal tissue.











 


1With the Classic way, the lesioned area will only be masked, but with the enantiomorphic (Nachev et al. 2008), the software will take the healthy tissue of the spared hemisphere to fill the damaged area and calculate a better transformation on an entirely “spared” image to apply it to your T1. Be careful, the enantiomorphic transformation cannot be used in case of a lesion in the left and the right hemispheres.


2You can apply transformations on 4D images, the 4D images will be splitted across the 4th dimension and transformations will be apply on each 3D images and then merged again.


*be careful to be consistent between your images and the template, for instance : if you choose to not use the skull stripping on T1 images with skull you have to select a template with skull (You can find a MNI152 with skull in BCBToolKit/Tools/extraFiles/MNI152_wskull.nii.gz). If you want to normalise lesioned T1 without the skull, you may want to use classic lesion masking.


**exact same name as the T1 images used above



PAPER METHOD SECTION (feel free to edit or copy and paste)


Registrations of Patient’s T1 MRI are performed using BCBtoolkit (Foulon et al. 2018) that implemented the following steps: As spatial normalization can be affected by the presence of a brain lesion, each lesion or signal abnormalities due to the lesion (manually segmented) can be used as a mask during the normalization procedure to optimize the brain normalisation (Ripoles et al., 2012; Volle et al. 2012) :

- In the case of the classic approach (choose this one if you have bilateral lesions or if you want to normalise images without skull), this masking procedure was used to weight the normalization to brain rather than non-brain tissue or lesions (Brett et al., 2001).

- In the case of the enantiomorphic approach (Nachev et al. 2008): Each patient lesions or signal abnormalities due to the lesion is replaced symmetrically by the healthy tissue of the contralateral hemisphere.


The skull stripping (if selected) is performed using the ANTs brain extraction algorithm (http://stnava.github.io/ANTs/)


T1 images are registered to the template (MNI152) using affine and diffeomorphic deformations (Klein et al., 2009; Avants et al., 2011).