Thanks Andreas and Mark, I guess my question was whether it is recommended to normalize using the -inm argument in this case and where the 1000 value comes from? Cheers Reem On 20/11/2010, at 12:36, "Andreas Bartsch" <[log in to unmask]> wrote: > Hi Reem, > > don't worry about cal min! > And yes, 6 DoFs is good, assuming you are talking about intra-individual follow-up registrations without much pathology / surgery... > Cheers, > Andreas > ________________________________ > Von: FSL - FMRIB's Software Library [[log in to unmask]] im Auftrag von Reem Jan [[log in to unmask]] > Gesendet: Freitag, 19. November 2010 22:21 > An: [log in to unmask] > Betreff: Re: [FSL] Flirt average query > > Hi Mark > > Thanks alot for your reply, very reassuring indeed. > > Another query regarding this... When using flirt_average, do you recommend normalizing the input images first (as suggested by the post I mentioned). I.e. > > "A thing the flirt_average script is not doing is the normalization of the single scans before averaging. This step might be useful when there is a difference in image intensities. Just change the last line of the script for example to: > > fslmaths $output -inm 1000 -Tmean $output" > > When I tried adding the "-inm 1000" argument to my flirt_average script, the cal_min value became a lot more negative (-125). I'm not sure whether I should perform this step or not? > > Last question is can I double check that 6 DOF is a good value for flirt in this case? I can't see reason for using 12 DOF. > > Thank you in advance :) > > Cheers > > Reem > > On 19/11/2010, at 22:36, "Mark Jenkinson" <[log in to unmask]<mailto:[log in to unmask]>> wrote: > > Hi Reem, > > The cal_min and cal_max values are really unimportant. > They *only* control how the image looks in a viewer (the min and > max *displayed* range). However, they have absolutely no effect > on the stored intensity values. All they do is act as the initial values > in the FSLView display range boxes. > > So therefore it is completely unimportant what they are set to. > However, I would still recommend using flirt_average in general as > it produces slightly sharper images due to the sinc interpolation. > One downside of the sinc interpolation is the fact that it induces > negative values (due to ringing) near the strong edges. This is > almost certainly why the cal_min gets set to a negative value. > It really isn't important what cal_min is, but that is an indication > that there is some ringing in the output data. In general I would > say that this was fine and worth the improved sharpness in the > average, but it really is a judgement call. So have a look yourself > at the output images and go with whichever one you prefer. > > All the best, > Mark > > > On 19 Nov 2010, at 04:27, Reem Jan wrote: > > Dear Mark/Steve or anyone who is happy to answer my FLIRT query J > > I am in the process of averaging 2 xT1-weighted structural scans per subject to use in an FSL-VBM analysis. > > I have searched through the archives and found a very helpful post on flirt_average (https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0803&L=FSL&P=R24790&1=FSL&9=A&J=on&d=No+Match;Match;Matches&z=4) and hence I tried using flirt_average as follows: > > Flirt_average 2 input_1 input_2 output_average –dof 6 > > I noticed (using the fslinfo command) that the output_average file had a cal_min -16 and cal_max 876. When I opened the output file in fslview, these values of -16 and 876 where what I saw in the bricon min max tool bar. I compared these values to the input images which had cal_min and cal_max values of zero, however when viewed the input images in fslview I see a min value of 0 on the Bricon toolbar and a maximum value of 423. > > I got slightly concerned about the output (average T1) negative cal_min value (-16), so I decided to try other averaging methods to see if I get the same sort of output. I tried the following: > > 1. Flirt (where the reference is input_1, the input is input_2 and the output is input_2flirted) using 6 DOF > 2. Fslmaths input_1 –add input_2flirted –div 2 output_average > > The output_average from this method had a cal_min of zero and cal_max of 838 > > I then tried another method (I think this is what flirt_average script is based on) > > 1. Flirt (where reference is input_1, the input is input_2 and the output is input_2flirted) using 6 DOF > 2. fslmerge –t output_merged input_1 input_2flirted > 3. fslmaths output_merged –Tmean output_average > > The output_average from this method was exactly the same as the method above (cal_min of zero and cal_max of 838). Both these methods have resulted in a cal_min value of zero as opposed to the negative number I get from using the flirt_average command. > > My questions are > 1. what are cal_min and cal_max values? > 2. Should I not be using flirt_average because of the negative value I am getting for cal_min? > 3. Is it ok that the cal_max value is almost double of that of the original input files (although the output file has been averaged)? > > Sorry about the long explanations and I appreciate any advice you can provide. > > Many thanks > Reem > > > __________ Information from ESET NOD32 Antivirus, version of virus signature database 5631 (20101118) __________ > > The message was checked by ESET NOD32 Antivirus. > > http://www.eset.com