For n3/nu_correct, you should read the paper by Sled. For mri_normalize, it selects points that it strongly suspects are in 100% white matter. It is assumed that all WM is iso-intense, so any variation is bias field. It then uses soap bubble smoothing fill in the values between control points.
On 11/12/2015 10:45 AM, Guido Orlando Pascariello wrote:
Hi FS Experts
I'm bioengineering student and I'm working on my final project to get my degree. it's about process neuroimaging (IRM) in order to cuantific some parameter which could be helpful for medicine (clinical neurology).
I'm using FreeSurfer to process my data. It works realy good. But it will be so much helpful to me if you can answer me a few question about it!
*FIRST ONE*:
I've been using 'mri_normalize' and I read the paper which describe it. But:
- In step 4: Discard outliers from the array: You suppose variation in intensity due to magnetic field inhomogeneities is smooth across space. But, field inhomogeneities are corrected in 'NuCorrect' module using N3.
- I understood the 1D normalization process. However the 3D normalization process works estimating a bias field like that which is corrected by 'NuCorrect' module.
Could you tell me why did you that? and what are diferences with 'NuCorrect' module?
*SECOND ONE*:
It's abuou 'mri_nu_correct':
I've read the description but I'm not sure about it. Does it use the N3 algorithm to work?
In the description is written that "changing the number of iterations should be done carefully, because this can exacerbate artifacts". Could you tell me what are these artifact? and why is the default number of iteration four?
I'll be grateful if you can help Being all for now, I await a reply
Guido PASCARIELLO Facultad de Ingeniería - Universidad Nacional de Entre Ríos Argentina
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer