Hi,

The principle of fugue is explain in the paper:
Jezzard, P. and Balaban, R.S.; Correction for geometric distortion in echo planar images from B0 field variations; Magnetic Resonance in Medicine, 34(1):65-73; 1995.

If you only acquire 116 points in each line of k-space, then I think
that is the right number to use in the calculation, but this is quite
an unusual setup for me so I am not sure.  Please check with your
local physicist/operator who should understand the Philips setup
better than me.  It is quite unusual for us to have the scanner
interpolate the data - in general we do not want this - and I suspect
that this will alter the relationship that FUGUE assumes between
the image distortion and the other parameters.  Hence you are
likely to need to "adjust" the dwell time (effective echo spacing) to
account for this.  Essentially the distortion in the image space is
proportional to this "dwell time", so you can always adjust this to
do the right thing.  I suspect that in this case you will need to do
just that as the intrinsic voxel size that is related to the k-space is
different from the actual reconstructed voxel size due to interpolation,
but FUGUE will not deal with this and assume that the voxel size
in the image is the intrinsic, not interpolated, voxel size.

In general I would advise against interpolation on the scanner.  It
never improves analysis and quite often makes things worse.  If
you can, just turn it off.

All the best,
	Mark


On 26 May 2011, at 21:07, Shugao Xia wrote:

> Dear Mark
> Thank you for your reply
>  
> You are right,  I didn't run brain extraction on the fieldmap magnitude scan.
>  
>  
> The purpose of this experiment is simply to test the EPI distortion correction using the fieldmap.
>  
> in-plate image size:  
> Size = 240 x 240  ,Dims = 0.9583333135 x 0.9583333135,  which is corresponding to REC voxel MPS(mm): 0.96/0.96/3.00,
>  
> but k-space data size: ACQ matrix M x P:116 x 114, ACQ voxel MPS (mm): 1.98/2.02/3.00. 
>  
> The discrepancy is because of the image interpolation in the image reconstruction. Here we approximate the dwell time as
> 1/(BW * 240) using image size or 1/(BW*114) in ACQ matrix ?
>  
> Could you please briefly explain how the fugue works on distortion correction or could you give me some references? Thank you very much 
>  
> all the bests
>  
> Shugao
>  
>  
> On Thu, May 26, 2011 at 6:50 AM, Mark Jenkinson <[log in to unmask]> wrote:
> Dear Shugao,
> 
> I have tried running your images through FEAT and they seemed
> to register fine to me (fieldmap to EPI).  It is necessary to run brain
> extraction on the fieldmap magnitude scan, as is indicated in the
> documentation and bubble-help.  So that is one possible cause of
> problems if you did not do this.
> 
> I wasn't sure of what to use for the Phase Encode direction in
> your scans (you seem to have more signal loss than distortion
> so I didn't find it easy to tell by eye).  I also am not familiar with
> Philips scanners and am not a pulse sequence programmer, so
> the list of parameters didn't help me much (as terms like "phase
> encode direction" were not there - so I assume they are using
> Philips-speak).  I was a little concerned/confused by the following:
> SENSE =                         "yes";
>    P reduction (AP) =          2;
>    P os factor =               1;
> Fast Imaging mode =             "EPI";
>    shot mode =                 "single-shot";
> Echoes =                        1;
>    partial echo =              "no";
>    shifted echo =              "no";
> Halfscan =                      "yes";
>    factor =                    0.600000024;
>    reversed order =            "no";
> EPI factor =                    61;
> FOV          RL (mm) =          230;
>             AP (mm) =          230;
>             FH (mm) =          114;
> Voxel size   RL (mm) =          2;
>             AP (mm) =          2;
> Slice thickness (mm) =          3;
> 
> As it looks like SENSE acceleration may have been used, however
> I can't make all the numbers be consistent for me, especially as the
> image you sent me was of size:
>  Size = 240 x 240 x 38 x 1 : Dims = 0.9583333135 x 0.9583333135 x 3.0000038147
> so that the slice-thickness seems right but the in-plane resolution
> is not that specified in the parameter file.  It would make more sense
> to me if there were 115 voxels of 2mm each, and in that case I guess
> than an EPI factor of 61 and a SENSE acceleration of 2 would be
> sensible.  I'm not sure what Halfscan=yes means though.
> 
> As for the "effective echo spacing" or "dwell time" that is needed
> for the distortion correction, this is not about how long it is between
> RF pulses (which is influenced by the EPI factor), but about how
> long between each crossing of the k-space axes (and hence the
> formation of gradient echos).  This number is needed to work out
> the phase accrual between different lines of k-space, and so it doesn't
> matter how many lines were acquired per RF pulse (which is what
> I believe EPI factor is) but only about how long it takes to move from
> one line to another.  It is also important to know if lines are "skipped"
> by acceleration, and if so the actual time taken between the acquired
> lines needs to be divided by the effective distance between the lines
> (as measured in units of the line spacing of an unaccelerated sequence,
> so that skipping one line means a line spacing of 2 and hence requires
> a division of the time taken per acquired line by the factor 2, which is
> also equal to the acceleration factor).  Hence EPI factor should not
> enter into the calculation of the "dwell time" unless you are working
> backwards from some total time to acquire all of these lines.  I would
> expect the value to be closer to the read out bandwidth multiplied
> by the number of samples in the readout direction (although, even this
> is unclear to me given the discrepancy between the image dimensions
> and those stated in the parameter file).  However, this calculation does
> not take into account ramp sampling or the time for the phase encoding
> blips, and so will be less than the true time, though unlikely to be out
> by more than 50%.
> 
> Sorry that I cannot help more.  If you take this email and the parameters
> to your local physicist or scanner operator, then they should be able
> to tell you the correct phase encode direction and how to calculate the
> dwell time.
> 
> All the best,
>        Mark
> 
> 
> 
> On 24 May 2011, at 14:56, Shugao Xia wrote:
> 
> > Dear  Mark
> >
> > The reference number is 929658.
> >
> > "EPI factor" on a Philips scanner means the number of lines acquired per shot. The right echo spacing is 1/(BW * EPI factor), but  my collegue think that due to interpolation in the image reconstruction, ACQ matrix is different to image size, and image size is the number of lines acquired per shot when calculating the dwell time for EPI distortion correction.
> > I attached the text file for the EPI scanning parameter.
> >
> > Thank you very much
> >
> > Shugao
> >
> >
> >
> >
> > On Tue, May 24, 2011 at 3:36 AM, Mark Jenkinson <[log in to unmask]> wrote:
> > Dear Shugao,
> >
> > I cannot really diagnose the problem from this image, so please
> > upload your data to:
> >   http://www.fmrib.ox.ac.uk/cgi-bin/upload.cgi
> > and send us the reference number.
> >
> > As for the dwell time (or echo spacing) calculation, neither option
> > is perfect because scanners don't instantaneously move from
> > one line to another, and the time for the phase encoding blips
> > and any ramp sampling needs to be taken into account.  You
> > will get something roughly right by using the bandwidth and the
> > number of readout points (your first calculation) but it will be
> > somewhat off.  As for the second calculation, I do not know
> > exactly what "EPI factor" means on a Philips scanner but it seems
> > unlikely to me that this is the correct thing to do.  It may refer to
> > the number of lines acquired per shot, but then I would expect it
> > to be a factor of 240. However, the number of lines acquired does
> > not affect the dwell time which is the time taken per line.  It may
> > indicate that you have a segmented acquisition or an accelerated
> > acquisition. If you have an accelerated or segmented acquisition then
> > you also need to divide by the acceleration/segment factor (the ratio
> > of the line spacing in the phase-encode direction of k-space to what it
> > would be in a single-shot, unaccelerated acquisition) in order to get the
> > "effective" echo spacing (or dwell time, the two terms are interchangeable).
> >
> > All the best,
> >        Mark
> >
> >
> > On 23 May 2011, at 23:07, Shugao Xia wrote:
> >
> > > Hi,
> > >
> > > I used the Feat GUI to correct EPI distortion and the results were wrong, the small portions of the fieldmap only appeared in space of EPI. I think this is because of the wrong registration between the magnitude image and EPI.  So I tried to use flirt as following:
> > >
> > > flirt -in mag   -ref EPI  -dof 6 -omat rot.mat -o mag_in_EPI  [ -usesqform ]
> > >
> > > the attached fig shows the image mag_in_EPI,  does any one advise me to correct it?  if you need the data, I will upload it.
> > >
> > > Another question:  how to calculate the dwell time?  I used 3T Philiphs to obtain EPI data wiht  BW(Hz)=24.1, image resize=240*240,  EPI factor = 61.  So the dwell time = 1/(24.1*240) = 0.17ms or the dwell time  = 1/(24.1*61) = 0.68ms.  My colleague thinks we use the former one here, but I am confused.  Can anyone help me clearify?  Thank you very much
> > >
> > > Take care
> > >
> > > Shugao
> > >
> > >
> > > <1.jpg>
> >
> > <FE_EPI SENSE.txt>
>