Hi Reem,
The mask should be a mask of the non-zero probability voxels
(so typically the brain mask).
A threshold of 0 means that nothing passes FDR.
It will hopefully change when you use the right mask.
All the best,
Mark
On 2 Jun 2011, at 07:15, Reem Jan wrote:
> Hi Steve, Gwenaelle and Hannah
>
> I’ve followed this thread and I had a similar question to Hannah which has been answered very well by Steve and Gwenaelle.
>
> As a consequence of this thread, I have applied FDR correction to my grot_vox_p_tstat images (after inverting them and obtaining the 1_minus_grot_vox_p_tstat images), and I had a simple question regarding the fdr command:
>
> Fdr –i 1_minus_grot_vox_p_tstat1 –m mask –q 0.05
>
> What kind of mask can we input here?
> I tried running the fdr command without a mask and got the output “Probability threshold is 0”. What does this mean? Will this change if include a mask in the command?
>
> Your advice on this is much appreciated.
>
> Kind regards
> Reem
>
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Stephen Smith
> Sent: Wednesday, June 01, 2011 8:13 PM
> To: [log in to unmask]
> Subject: Re: [FSL] fsl VBM -clusters
>
> Hi
>
> On 31 May 2011, at 12:01, Hannah Bruehl wrote:
>
>
> Hi to All,
>
> I ran an fsl VBM analysis in order to bolster my cortical thickness analysis. Uncorrected (tfce-p-tstat file), I see clusters in roughly the same area as in the CT analysis. I'd now like to extract all the relevant information about those clusters, which I understand I need to do using "cluster" and have 2 questions.
> 1: is there a rule of thumb as to how many voxels one reports for an uncorrected p, say 300 and up or sth like this?
>
> We would not put it like this - in general you should use corrected p-values, whether you are doing voxelwise, clusterwise or TFCE-based thresholding - i.e. you should be using corrected p-values output by randomise, or else taking uncorrected p-values and feeding them into a further correction process such as FDR. Simply limiting the number of voxels in a cluster does not directly let you control what the corrected p-value is.
>
>
> 2: apparently, the coordinates I get are in aligned anatomical space, since that is what it says when I open the relevant contrast in fslview. How to I get MNI or Talairach coordinates then?
>
> randomise works in the space of the data you feed into it. If you have used FSL-VBM then your group testing in randomise will almost certainly be in MNI152 space already.
>
>
> and , related, how do i get a mean percentage about the structure the cluster is located in, since that is changing when clicking around in a given cluster within fslview?
>
> Do you mean wrt a particular atlas?
>
>
>
>
>
> Thanks,
>
> hannah
>
>
>
> ---------------------------------------------------------------------------
> Stephen M. Smith, Professor of Biomedical Engineering
> Associate Director, Oxford University FMRIB Centre
>
> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
> +44 (0) 1865 222726 (fax 222717)
> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
> ---------------------------------------------------------------------------
>
>
>
>
>
> __________ Information from ESET NOD32 Antivirus, version of virus signature database 6172 (20110601) __________
>
> The message was checked by ESET NOD32 Antivirus.
>
> http://www.eset.com
>
>
> __________ Information from ESET NOD32 Antivirus, version of virus signature database 6172 (20110601) __________
>
> The message was checked by ESET NOD32 Antivirus.
>
> http://www.eset.com
|