Marta,
The values that you are getting sound a bit high. A whole brain mean CBF is normally lower than 60 due to the mixture of grey and white matter: 30-40 seems quite common in the literature.
In principle ASL_GUI will try to do all the necessary registration and calibration for you, but it is not guaranteed that using the default settings that everything will work perfectly first time of course. It is always worth checking that the registration of ASL data to the T1 looks sensible. An important thing to bear in mind is that the mask generated for calibration purposes is of the CSF (by default) and thus should be what you see ref_mask. This is purely to be used as a basis for estimating M0, a further separate mask should be created in which CBF is calculated. By the looks of your image the brian mask in which CBF is calculated looks fine, but the CSF reference mask doesn't look right. This might be an issue with registration or the segmentation of the T1. You might need to create your own manual CSF mask - at least in one case to see if that makes a difference - and supply that to the GUI.
Michael
On 14 Aug 2014, at 15:31, Marta Boteller <[log in to unmask]> wrote:
> Dear all, I'm using the ASL gui in FSL 5.0 to calibrate perfusion ASL images (control and tag) obtained with a Siemens Tim Trio 3T with a PASL sequence. I've several short questions that maybe some of you have already dealt with, and I would really apreciate your help: (1) For healthy subjects, I'm expecting to have a mean CBF value of 60 ml/100g/min (absolute units) and a reference M0 value arround 1000. I've opened the resulting calibrated image (perfusion_calib.nii.gz) with FSLview, ImageJ and MriCron and looked at the histogram. Appart from seeing different intensity values with the different viewers, I obtain very high values (i.e , 400) and the mean value seems to be far from the expected 60. Is that normal? I understand I'm doing something wrong in the calibration process. They've given me the following values (I don't have access to the scanner adquisition panel): -PASL sequence with QUIPS II -BAT=1.3s, bolus=800ms, TI1 (blood)=1.6s and TI(tissue)=1.3s, 20 slices, TE=11ms TR=3500ms -50 tag-control pairs + a first control image (for M0 calibration purposes) I'm using the 'Background suppresion tissue' option with a co-registered ASL 4D volume containing the control/tags and a structural T1 (both brain extracted), fixed the bolus to be 800ms and provide the M0 calibration image. (2) If I want to have a unique CBF value for each individual, I was thinking on masking the 'perfusion_calib.nii.gz' with the generated mask in 'reff_mask.nii.gz' and calculate the mean intensity on the resulting image. But I've found that the reff_mask.nii.gz does not mach with what should be the CBF region (please see attached image 'Mask.jpg'). Do the ASL volume and the T1 have to be normalized before the calibration process? I've also tried to provide a co-registered CBF mask obtained from T1 segmentation without success. (3) Should I do first all the pre-processing steps and calibrate the ASL images at the end? that means for example transforming all the 50 pairs of control-label images into a 4D volume and normalize it together with the T1. This way I could also provide the '...mat' file for registration purposes. Thank you for your help! Kind regards, Marta Boteller Master in Medical Physics Hospital Clinic Barcelona, Spain <histogram.jpg><Mask.jpg>
|