 |
 |
Thank you Mark ________________________________ From: FSL - FMRIB's Software Library [[log in to unmask]] On Behalf Of Mark Jenkinson [[log in to unmask]] Sent: Tuesday, April 23, 2013 6:38 PM To: [log in to unmask] Subject: Re: [FSL] MRF beta value in FAST Hi, It is _very_ unlikely that you'll get a good registration of the functional to MNI space if you cannot get a good registration to your structural. I strongly recommend that you upgrade your FSL version and use the new tools. All the best, Mark On 23 Apr 2013, at 16:21, Helen Sawaya <[log in to unmask]<mailto:[log in to unmask]>> wrote: Hi Mark, I'm afraid I'm using an older version of FSL that doesn't recognise the epi_reg command. I was wondering if as an alternative I could register my functional images to MNI and then use the FSL tissue priors to extract csf and white matter signal from the functional image in MNI space. Would that be an accurate method or is it preferable to use functional images in their anatomical space? Thanks a lot Helen ________________________________ From: FSL - FMRIB's Software Library [[log in to unmask]<mailto:[log in to unmask]>] On Behalf Of Mark Jenkinson [[log in to unmask]<mailto:[log in to unmask]>] Sent: Tuesday, April 23, 2013 5:27 PM To: [log in to unmask]<mailto:[log in to unmask]> Subject: Re: [FSL] MRF beta value in FAST Hi, Try using BBR via the epi_reg command. This generally gives a very good EPI (fMRI) to T1 registration. All the best, Mark On 23 Apr 2013, at 15:06, Helen Sawaya <[log in to unmask]<mailto:[log in to unmask]>> wrote: Thank you Mark. Quick follow up: I used flirt with 6 dof to register the functional image to its T1 image then InvertXFM to get the inverse matrix which I applied to the pve segments. The csf and white matter pves however appear not to overlap very well with the functional image. Am I correct to think that this looks like a problem with registration (although trying with 7 dof instead did not improve the results)? I therefore tried directly changing from T1 to functional space using flirt, with T1 as the input image and the functional image as reference. I got the same results as the first method I used (example corpus callosum in functional image appears below that of the white matter pve). Are the two methods above identical? Although they gave very similar results I'm guessing not because the matrices outputted were not identical.. Any suggestions on how I can get the functional and T1 images to overlap more closely? Thanks a lot for your help Helen ________________________________ From: FSL - FMRIB's Software Library [[log in to unmask]<mailto:[log in to unmask]>] On Behalf Of Mark Jenkinson [[log in to unmask]<mailto:[log in to unmask]>] Sent: Tuesday, April 23, 2013 2:27 PM To: [log in to unmask]<mailto:[log in to unmask]> Subject: Re: [FSL] MRF beta value in FAST Dear Helen, There is no fixed rule for what values to use in the MRF as it depends on image resolution, type of sequence, SNR, subjects of interest, artifacts present, etc, etc. The defaults are fairly general and give reasonable results most of the time but we strongly recommend that people try different options the first they the use a new scanner/sequence/type-of-subject/etc and evaluate the results by looking at them carefully from the point of view of anatomical accuracy. The other important thing to say is that we do not recommend segmenting EPI/functional images, as they rarely have good anatomical contrast, normally have poor resolution, and suffer from intensity-related artifacts. What you should do is segment your structural (T1-weighted) image and then transform the segmentation into the functional space by using registration. Once you have the white-matter mask in the functional space then you can extract your average signal. All the best, Mark On 23 Apr 2013, at 08:21, Helen Sawaya <[log in to unmask]<mailto:[log in to unmask]>> wrote: Dear experts, I just want to make sure I understand the difference between the -f function in FAST: "MRF beta value during initial bias field removal phase" and the -H function: "Main MRF beta value, initial segmentation calculation". Is it correct that the -H value determines how much spatial neighbourhood information is used? If yes is it advisable to use a large beta value (example 0.5) as opposed to a smaller one, or would that lead to too much smoothing that the separation between tissue types would be less accurate? I am getting a significant amount of white matter activation in my ICA components so I would like to regress as much white matter signal as possible. The default of -f is 0.02; is it advisable to leave it as such? Would it be affected by smoothing of the bias field (-l function which at default is 20mm). I was also wondering whether the following order of steps is correct: 1- FAST segmentation using preprocessed functional image 2- regression of motion, white matter and csf signal from the preprocessed functional image (using fsl_regfilt) 3- registration of functional image to T1 then MNI space 4- input functional image in MNI space into melodic for postprocessing Thank you for your help Helen