Print

Print
Hey Mark,

Thanks very much for your response. The settings on the Siemens scanner were Contrast --> Reconstruction --> Phase (as opposed to Magn./Phase), so it did not save the Mag images during acquisition. Are you suggesting they could somehow be retrieved anyway?

Assuming they are not, I had a couple questions about your suggestions. I have good masks based on the structural image, so it should not be a problem to register the structural mask to phasemap in the way that you suggest (I'm also aware that this needs to be very tight), then treat this as the skullstripped mag image, and then unwrap with the steps listed in fsl_prepare_fieldmap.

At this point, typically I would just feed my files through feat, and let it perform the distortion corrections and registrations. I'm unclear about what I would use epi_reg for and what you mean by 'wm images' - do you mean whole brain masks?

Also, I am curious about this new feature that incorporates bbr registrations. Does it matter if I register from epi --> T2 or epi --> T1? I have both T2 and T1 structural images for each subject (this is a monkey study) and T2 and T1 atlas images, so I have many options for atlas registration. What process would you suggest would be most robust: epi --> T2 --> T1 --> T1 atlas, epi --> T2 --> T2 atlas, or epi --> T1  --> T1 atlas?

Thanks VERY much!

David


On Fri, Apr 12, 2013 at 4:03 PM, Mark Jenkinson <[log in to unmask]> wrote:
Dear David,

It is a pity that you do not have a magnitude image for your fieldmap.

Before you do anything you need to convert the images into units of rad/s.  If your range is currently +/- 4096 then you should only need to divide by 4096, multiply by 3.14159 and then divide by the echo time difference of your fieldmap sequence (usually something like 2.5ms, which should be expressed in seconds, like 0.0025).  This will give you a fieldmap in rad/s (if the original mapping of +4096 = +pi and -4096 = -pi is correct; but you should check on this if possible).

After this you could rely on the scanner coordinates to align your images with the structural image, providing they were acquired in the same scanning session.  This can be done using the -usesqform and -applyxfm options in flirt (do not use the -init option).  Check the output carefully though, to see if you think it aligned well (you can normally see the noisy part of the fieldmap at the edge of the brain, so check that this lines up with the edge of the brain in the structural image).  Once you've done that then you can use flirt to do the bbr registration (you'll just need to segment the structural yourself and create the appropriate wm images - see the epi_reg script for how to make this and the appropriate options for flirt).

It would be a lot simpler if you could get a magnitude image, which is normally possible, so I would double check on that.

All the best,
Mark





On 11 Apr 2013, at 22:42, David Grayson <[log in to unmask]> wrote:

Also, just to follow up on that, the scans were acquired on a Siemens 3T scanner. The range of intensity values in the PHASE image are -4096 to 4096. The instructions on the fugue website indicate that the mag images are required, but I'm wondering if anyone has managed a viable workaround.

Thanks,

David


On Thu, Apr 11, 2013 at 1:52 PM, David Grayson <[log in to unmask]> wrote:
Hi fsl gurus,

I would like to carry out distortion correction of our functional EPI sequence, but the scanner only provided the phase image from the fieldmap scan, not any magnitude images. Our MR expert here does not seem to think that the mag images can be reconstructed from the phase dicoms that were provided.


Has anybody been able to carry out distortion correction with just the phase images? Do you have any advice on how to proceed? Thank you very much in advance,

David