Dear Mark, Thank you very much for your reply. Now I understand how the FSL estimate the motion pramrameters. Basically it's just make it as problem with the 3D matrix, if the voxel size is the same. I think I could probably try to move the phantom to see how the estimated parameters changed. Based on my understanding, may I confirm that the bvec should be the same space as the 3D matrix, inplan + slice direction? so if there is a slice angulation, let's say it's 45 degree from axial direction. Then the FSL assume the x-y of bvecs is inplan and z is the same with slice direction? Many many thanks! Best regards, On Fri, Aug 27, 2010 at 3:01 AM, Mark Jenkinson <[log in to unmask]> wrote: > Dear Xiaowei, > > I'm not familiar with MPS or LHP, but will briefly describe the coordinate > system used by FLIRT and MCFLIRT. I suggest that you only ever > work with matrices and never the parameters (rotations, translations) > as these are *much* harder to work out, and you can always get > the matrices too. > > The matrices map between coordinates in a mm coordinate space. > They are related to the nifti coordinates as follows: > - if the qform or sform (whichever has a non-zero code) has a > negative determinant then: > - x, y, z mm coordinates are just the voxel coordinates > multiplied by the voxel sizes (in mm) > - if the qform or sform has a positive determinant, then it is the > same as above *except* that the x-axis is reversed > (that is, if there were 64 voxels, then voxel number 0 would > become number 63 and vice-versa, 1 would become 62, etc) > - note that the qform and sform relate to each image, so you may > need to reverse either the initial or final x (voxel) coordinates > > The bvecs are in the same space as the mm coordinates, as all > FSL programs will do the above swapping in x when they read in > images. > > All the best, > Mark > > > > > > > On 26 Aug 2010, at 20:52, vivian wrote: > > Hi all, >> >> Sorry that I didn't specify the coordinates I mostly concerned. The main >> coordinate system we are using now is XYZ, MPS, or LHP. It would be truely >> helpful if the answer is related to these three systems. >> >> Thank you again >> >> >> >> On Thu, Aug 26, 2010 at 2:03 PM, Xiaowei Zou <[log in to unmask]> >> wrote: >> Greetings, >> >> I'm a new researcher in DTI field and my project needs to match the >> estimated motion parameter (6-DOF: x,y,z,roll,yaw,pitch) to the real motion >> in XYZ coordinate of scanner. I've read the lecture about FLIRT but still it >> didn't solve my questions. I'd be very gratful if someone could help me out. >> >> 1, What's the coordinate system in which the estimated 6-DOF motion >> parameter defined? >> >> The lecture mentioned voxel space coordinate which can make the problem >> become a pure 3D matrix. However, it requires voxel size so I think it might >> jump between voxel space coordinate and standard space coordinate. >> >> 2, How does this coordinate system related to the input NFITI 3D data? >> >> In either coordinate system I found in the lecture, the definition of >> coordinate is corelated to the image data coordinate. For example, the voxel >> space coordinate defines origin at lower-left corner. Then what's the >> lower-left cornner in the NIFTI 3D data? Is that the same lower-left corner >> if I read the 3D NFITI by spm_read_vols? >> >> 3, This is a little bit extension about FDT. What's the assumption of >> coordinate of the input bvecs? Is it XYZ ? >> >> >> FSL does a wonderful job about fiber tracking so I believe there must be >> some understandable transformation from real world to estimated motion >> parameters. >> >> Thank you veryone who read this message. Any help from you is highly >> appreciated. >> >> Best regards, >> >> Xiaowei Zou >> >> >> >> >> >> >> >> -- >> Xiaowei.Zou >> >> Department of Biomedical Engineering >> Columbia University >> New York, NY, 10027 >> Cell: 203-506-6130 >> > -- Xiaowei.Zou Department of Biomedical Engineering Columbia University New York, NY, 10027 Cell: 203-506-6130