Hi - got the data ok now.
The two timepoint images are both quite nice and very well aligned (in
mid-space) by SIENA.
The brain extractions are both good, and slight differences in brain
extraction won't be a factor here because the two brain masks are
combined before being re-applied to the original images.
The second timepoint image is more blurred than the first - if these
images are generally representative of what you're scanner is doing
then it seems that maybe you are getting slightly more blurred data
now, after the upgrade. I thought that this might be the cause of the
positive-bias in your results, so I smoothed the first timepoint image
with
fslmaths A -s 0.53 As
to get matching smoothness and it only made matters worse!
So then having looked even more carefully at the images e.g.
fslview A_halfwayto_B B_halfwayto_A
when you turn B on and off you can see that there is a thin layer of
bright tissue (possibly marrow from the centre of the skull) appearing
differently in the two timepoints. In A you have a nice clean dark
area surrounding the brain almost everywhere, whereas in B some bright
band is shifting around - at the front of the head it is 'moving
forwards' away from the brain, but at the back it is moving forwards,
to touch the brain, causing the apparent brain edge to appear to be
expanded backwards.
Karla tells me that this kind of shift could be related to changes in
acquisition bandwidth and/or phase-encode direction on the scanner.
Also maybe the apparent smoothness change could be related to
filtering changes on the scanner. I guess you need to talk to your
physics guys!
Cheers, Steve.
On 12 Sep 2009, at 08:54, Mehul Sampat wrote:
> Hi Steve,
> I have uploaded the file again. The reference number is 336367.
> A little more background information. This data is part of a 5 year
> longitudinal study.
> The images were collected on a 3T GE scanner. For the data from the
> first four years,
> Siena works great. All of the issues we see are after the scanner
> software upgrade from the company. That is when we compare the pre-
> and post upgrade images. We were reluctant to do any upgrades for the
> scanner but we did not have full control over this issue. (The basic
> argument was that it is hard to go 5 years or so without a scanner
> software upgrade).
>
> My feeling is that such issue would possibly come up for other
> longitudinal studies too.
> I was wondering if maybe some normalization pre-processing procedure
> like histogram matching (after skull stripping) might help to
> alleviate the problems.
>
> Thanks
> Mehul
>
>
>
>
> Thanks
> Mehul
>
> On Fri, Sep 11, 2009 at 10:47 PM, Steve Smith <[log in to unmask]>
> wrote:
>> Hi - the file you uploaded was empty - can you send it again?
>> Cheers.
>>
>>
>> On 10 Sep 2009, at 19:43, Mehul Sampat wrote:
>>
>>> Hi Steve,
>>>
>>> Thank you for your suggestions earlier.
>>>
>>> I found that after the scanner software upgrade there are intensity
>>> changes and also the SNR is better in the post-upgrade images.
>>>
>>> Also as you suggested I ran SIENA with the -d option and was looking
>>> at the intermediate outputs of Siena. I think because of the
>>> intensity
>>> changes, the BET segmentation results look different for the pre-
>>> and
>>> post software upgrade images.
>>>
>>> I was wondering if you could please give a look at the intermediate
>>> output for one case, to see if the problem with the PBVC value is
>>> being caused by the BET output ?
>>> I uploaded the all of the intermediate output for one case at
>>> http://www.fmrib.ox.ac.uk/cgi-bin/upload.cgi. The reference number
>>> is
>>> 595458.
>>>
>>> If BET is the cause for the positive PBVC value, is there a way to
>>> input different BET parameters (in the Siena script) for each of the
>>> images?
>>>
>>> Thanks
>>> Mehul
>>>
>>> On Sun, Sep 6, 2009 at 4:38 AM, Steve Smith <[log in to unmask]>
>>> wrote:
>>>>
>>>> Hi - there's a number of things that can have changed in your
>>>> scanner
>>>> upgrade that could cause this, including intensity effects and
>>>> geometrical
>>>> effects.
>>>>
>>>> In general we _strongly_ recommend against any acquisition
>>>> changes during
>>>> a
>>>> longitudinal study, as you can never be sure that you will be
>>>> able to
>>>> remove
>>>> the confounds caused by the change.
>>>>
>>>> In your case you will need to look into all the intermediate stages
>>>> (outputs) of SIENA to try to track down where this confound is
>>>> affecting
>>>> estimates. If you use the -d debug option SIENA will leave a lot
>>>> of
>>>> intermediate stages for you to look at.
>>>>
>>>> Cheers.
>>>>
>>>>
>>>> On 5 Sep 2009, at 08:20, Mehul Sampat wrote:
>>>>
>>>>> Hi FSL Experts,
>>>>>
>>>>> I am using Siena to compute PBVC for MS patients.
>>>>> In the past we would see a mean -0.4% PBVC value for two scans
>>>>> a year
>>>>> apart.
>>>>>
>>>>> Recently, our GE scanner went through a software upgrade.
>>>>> I am now comparing two scans (pre- and post- upgrade) using
>>>>> Siena.
>>>>> (These pre- and post-upgrade images are also a year apart)
>>>>> Strangely for many of the subjects, the PBVC values are now
>>>>> positive
>>>>> (which is not what we expect for these patients).
>>>>>
>>>>> I tried Siena version 2.6 and version 2.4 and for both versions
>>>>> I see
>>>>> positive PBVC value.
>>>>> Has anyone experienced similar issues with Siena ? If so any
>>>>> suggestions to overcome this issue would be very helpful.
>>>>>
>>>>> Also any suggestions on debugging the intermediate steps would
>>>>> also be
>>>>> very helpful.
>>>>> Thank you,
>>>>> Mehul
>>>>>
>>>>
>>>>
>>>>
>>>> ---------------------------------------------------------------------------
>>>> Stephen M. Smith, Professor of Biomedical Engineering
>>>> Associate Director, Oxford University FMRIB Centre
>>>>
>>>> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
>>>> +44 (0) 1865 222726 (fax 222717)
>>>> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>>>>
>>>> ---------------------------------------------------------------------------
>>>>
>>>
>>
>>
>> ---------------------------------------------------------------------------
>> Stephen M. Smith, Professor of Biomedical Engineering
>> Associate Director, Oxford University FMRIB Centre
>>
>> FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
>> +44 (0) 1865 222726 (fax 222717)
>> [log in to unmask] http://www.fmrib.ox.ac.uk/~steve
>> ---------------------------------------------------------------------------
>>
>
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|
