Hi John,
I have a job which I feed to spm_preproc. However, as the bias field correction is part of the segmentation and I can't separate them easily. Though, I'd be perfect to run the bias field correction only as both is a bit time consuming (on the order of a few hours for both using my data).
Is it theoretically possible to run the bias field correction only? Or is the segmentation used in any way to guide the bias field correction (e.g. masking the image)? As far as I have seen the bias field corrected image is written a lot earlier than the segmented images. Therefore, I was wondering if the segmentation could be avoided (to some extent).
Best,
Falk
-----Ursprüngliche Nachricht-----
Von: John Ashburner [mailto:[log in to unmask]]
Gesendet: Freitag, 1. April 2016 13:40
An: Falk Lüsebrink; SPM@jiscmail ac. uk
Betreff: Re: [SPM] Bias field correction without segmentation possible
I'd suggest creating a job via the ui, saving it as a .m file, editing this to change the fwhm and running the modified job file.
The segmentation just fits a model to the data. It doesn't change the data first. I don't know if denoising helps, but I guess you could try running it on the data beforehand (or afterwards on the bias corrected version).
Best regards,
John
Falk Lüsebrink <[log in to unmask]> wrote:
>Dear all,
>
>I am working on high resolution 7T anatomical data and utilize SPM's bias field correction as part of my pre-processing. I have two questions regarding the bias field correction.
>
>First, to change the FWHM to lower values than can be specified using the GUI, I wrote a short script to create a job and feed it to spm_preproc_run.m However, as I am not interested in the segmentation during pre-processing I would like to disable it. Is this somehow possible? It would speed up the pre-processing greatly.
>
>Second, does the bias field correction involve any transformation of the input data? In case of some datasets I want to apply a denoising algorithm and I am trying to figure out at which stage in the pipeline I should apply it for best results. As transformation would result in a change of noise distribution, I would want to avoid it.
>
>Thank you very much!
>
>Best regards,
>Falk
>
>----------------------------------------
>Falk Lüsebrink, M. Sc.
>Otto-von-Guericke University, Magdeburg, Germany Forschungscampus
>STIMULATE http://www.forschungscampus-stimulate.de/
>
>Universitätsplatz 2
>39106 Magdeburg
>
>Room: ExFa - 4.06
>Phone: +49 391-67-19366
>Fax: +49 391-67-19347
|